# Calcutta Live > Calcutta Live is a 2026 member-member golf Calcutta auction analysis dossier paired with a live, real-time bidding and scoring tool used on auction night and at the course. Calcutta Live combines two things: a phone-first **live tool** (adaptive auction assistant, live standings, shootout, score entry, sales, and a printable bid sheet) and a full **analysis dossier** of 15 reports plus two capstone documents (a one-pager and a six-pager that distill the entire project) covering the valuation methodology, the statistical and Monte-Carlo models behind every win probability, the bidding strategy, what actually predicts performance, how handicap markdowns work, competitor/game-theory intel, the model-corrections log, portfolio-level risk/P&L analysis ("can we actually win money?"), and an investor-grade memo. The site has no authentication — every view and report is public. ## Live Tool - [Calcutta Live (app)](https://calcutta.high.green/): Real-time auction bidding assistant + live scoring — tabs for Auction, Targets, Standings, Shootout, Scores, Sales, Print, and Reports. ## Start Here — Capstone (the whole project, distilled) - [One-Pager — Auction-Night Action Sheet](https://calcutta.high.green/one-pager): The whole project on one screen — the thesis, the buy card with walk-aways, the hard rules, the competitors, and the single mental model: the market prices the card, we price the player. - [Six-Pager — The Whole Thesis](https://calcutta.high.green/six-pager): The comprehensive six-page synthesis — the honest verdict, what predicts performance, the edge (sandbaggers/markdowns/match-play aptitude), valuation and the buy card, competitor dynamics, and risk/sizing. ## Methodology & Models - [How It Works (Methodology)](https://calcutta.high.green/reports/methodology): The valuation model end-to-end: how P(win), fair value, and max bids are built. - [Statistical Analysis](https://calcutta.high.green/reports/statistics): Player and team distributions, handicap effects, and the figures behind the numbers. - [Monte Carlo Analysis](https://calcutta.high.green/reports/montecarlo): Simulation of the net better-ball match play that drives every win probability. - [Model Corrections](https://calcutta.high.green/reports/corrections): What changed in the model: $2k cap-anchor price reshape, the 10-stroke handicap cap, match-play aptitude, sign fixes. ## Strategy & Intel - [What Predicts Performance](https://calcutta.high.green/reports/what_predicts): The sticky finding: the market prices the card, we price the player — money predicts the floor, not the ceiling. - [Auction Strategy](https://calcutta.high.green/reports/strategy): Synthesized bidding strategy for our format — pots, pacing, and the adaptive engine. - [Cap Patrol & Psychology](https://calcutta.high.green/reports/cappatrol): Reading Cap Patrol player metrics and the psychology of who beats the net. - [How Markdowns Work](https://calcutta.high.green/reports/markdowns): How handicap markdowns actually work — USGA Exceptional Score Reduction + Cap Patrol's 5 criteria (sourced), separated from our inference. - [Competitor Strategy](https://calcutta.high.green/reports/competitor): Rival bidders (Russ/Reynolds/Flammia/Perry), the cap-anchor room, and the auction-night exploitation playbook. - [Matchup Analysis](https://calcutta.high.green/reports/matchups): Round-by-round matchup simulations across all 20 flights and pairings. - [Track Records (2021–2025)](https://calcutta.high.green/reports/history): Five years of member-member history: who shows up and how flights have played. ## Can We Win? (Portfolio Risk) - [Can We Win? — Portfolio Backtest](https://calcutta.high.green/reports/portfolio_backtest): Backtests the strategy across past years to ask the bottom-line question: does it make money? - [Forward P&L Simulation](https://calcutta.high.green/reports/portfolio_montecarlo): Monte Carlo of this year's portfolio P&L — the distribution of outcomes we should expect. - [Risk & Stress Test](https://calcutta.high.green/reports/portfolio_stresstest): Downside and stress scenarios: what happens to the portfolio when things go wrong. ## For Investors - [For an Investor (Mark Cuban)](https://calcutta.high.green/reports/investor): Investor-grade evaluation — bear case first, every number sourced, the honest expected value and the ask. ## Machine-readable - [Reports catalog (JSON)](https://calcutta.high.green/reports/index.json): Structured {title, url, description, topic} catalog of the live tool + all 15 reports. - [Full LLM corpus](https://calcutta.high.green/llms-full.txt): This index followed by the full readable text of every report. - [Sitemap](https://calcutta.high.green/sitemap.xml): XML sitemap of the root + all 15 clean report URLs. ============================================================================ # FULL REPORT TEXT ============================================================================ ============================================================================ # One-Pager - Auction-Night Action Sheet Source URL: https://calcutta.high.green/one-pager ============================================================================ ← App One-Pager (PR/FAQ) Six-Pager → 2026 Member-Member Calcutta · Plan & FAQ Spend the Whole Bankroll, With Discipline. A great night, the edges captured, the whole $2,500 deployed. The plan On auction night we will deploy the full $2,500 across the 2026 Member-Member Calcutta. A held reserve does not make money; it sits in a pocket. We will spend that money in two tranches. A core (~55–60%) goes to the teams the room misprices, bought under hard walk-aways. A stretch (~40–45%) goes to upside swings and fun picks; the core is where the edge lives, the stretch is entertainment spend we expect to be mildly −EV . The market prices the card; we price the player. We do not know who wins a flight. We know which floors are mispriced, and we buy those with discipline while a cap-anchored room overpays for ceilings. The FAQ Is this +EV? Barely, and only on the core. Within-flight outcomes are near-random. The documented edge is buying mispriced floors (sandbaggers, players whose card understates them) below value. The optimistic model puts the core slate near +24%; a pessimistic-but-plausible stack lands at −18%. A thin edge wrapped in fat variance. We size and walk away accordingly. Then why spend it all — why not hold a reserve? A 25% held reserve doesn't make money; it sits in a pocket. So we deploy the whole $2,500: the core stays disciplined, and the stretch turns the leftover into action all night. What's the edge? Individuals, not teams. Beating your billing doesn't stick to teams (year-over-year r = −0.23, they regress), but a short list of players outscores their card every year — Estes (+6.7 / +6.2), Keister (+4.9 / +8.3), Copeland (+4.1 / +4.2). A blind two-year residual test, an eyewitness sandbagger read, and the simulation point at the same names . We buy those players' teams below value, never at the $2,000 cap. What could go wrong? The biggest threat is an efficient market : if the room prices fairly, there is no floor to harvest. The half-buyback is the owner's option against us (he reclaims winners, leaves us the duds). Even assuming the model is right, the forward sim shows ~39% chance to lose money and ~20% chance to lose half the stake. The headline edge rests on ~2 sandbagger teams. The core has hard caps; the stretch is labeled as fun. The plan: full $2,500, Core + Stretch Remember the half-buyback math: effective cost is half the hammer and we own 50%, so $2,500 of effective spend buys roughly $5,000 of hammer capacity. We frame every allocation below in effective dollars. Core — ~$1,400 (55–60%) · the edges, hard walk-aways Pre-set the max per lot in writing; never raise it live. The closer a lot drifts to $2,000, the more certain we fold. Lot Team Walk-away Why it's core #63 Barbaree + Williford STEAL $700–900 Eyewitness sandbagger (Williford shot 73 in a 4-club event). Uncontested lane; private info. #94 Wright + Copeland BUY ~$900 Copeland won Flight 5 in 2025; beats his billing every year. No war premium. #104 Wood + Estes BUY hard cap ~$1,100 Biggest edge, only triple-confirmed team. The Sharp contests it; pay the max, do not chase past it. #76 Nodar + Heslep BUY ~$600 2nd in flight 2025 (30.5); top model P(win). Contested by the Sharp; at/below max only. #20 Knapp + Keister IF CHEAP only if cheap Keister beats his billing both years. Comes early; buy at/below value, then go dark. Stretch — ~$1,100 (40–45%) · −EV entertainment Upside swings, not the edge. The action lives here. Lot Team Lane Note #85 Vaniman + Hatz UPSIDE swing High match-play aptitude, largest non-capped +edge on the board. Contested. #5 Martin + Torres UPSIDE swing Large modeled edge, high aptitude. Early lot if the price is sane. — A cheap longshot in a rich flight POP $200–400 One or two lottery tickets where the pot is fat. Pure variance. — A home / fun pick ACTION your call The team you want to sweat all night. The room (types, not names) The Overconfident Veteran — knows everyone, overpays for marquee names, drives teams to the $2k cap. He pushes the cap-outs we avoid. Let him win them. The Sharp — value-hunter, day-trader mindset. Our competition for the good plays; he's on Wood + Estes. Discipline beats him; a war does not. The Loyalist — bids up older / higher-handicap players he trusts. The 10-stroke cap makes it the worst value on the board. Concede it. The Wild Card — unpredictable; spikes a price out of nowhere. Set the max before the lot and hold it. The hard rules (still in force) Avoid the ~18 cap-outs (16 of 18 are −EV at $2,000). Never chase to the round number. Fade cheap negative-aptitude teams; cheap is not value (Gallagher + Hansell: positive listed edge, clutch −7). Pace it: spend lighter early, push late, because four of five core targets land in the back third when the room is tapped out. The edge is thin; the walk-away is the controllable lever — worth ~13 ROI points. The one mental model The market prices the card. We price the player. The edge is thin. We do not know who wins a flight; we know which floors are mispriced. The core is the bet; the stretch is the fun. Final live numbers are on the board. Read the Six-Pager → ← Back to Calcutta Live ============================================================================ # Six-Pager - The Whole Thesis Source URL: https://calcutta.high.green/six-pager ============================================================================ ← App The Six-Pager One-Pager → 2026 Member-Member Calcutta · Memo The Whole Thesis Six sections: the claim, the evidence with numbers, the implication. The thesis in one breath: we will deploy the full $2,500 while capturing the genuine edges. The market prices the card; we price the player . Net match play makes flights near-lotteries, so we do not try to predict winners. We hunt the one thing the data supports — mispriced floors , teams the room writes off whose players beat their handicap. We split the bankroll into a disciplined core that holds the edge and a stretch that is entertainment. The edge is thin. We size for thin and spend the rest on fun. Final live numbers are on the board. Six sections: - The game and the verdict - What predicts performance — the player, not the card - The edge — sandbaggers, markdowns, match-play aptitude - The valuation — the cap anchor and the buy card - The room — four types, and the night's playbook - The plan — full $2,500, core + stretch Section 1 The Game and the Verdict Within-flight outcomes are near-random, and no price rule reliably beats the field. The base case sets up the rest. The mechanics, verified. Each flight keeps ninety percent of its pot — the winner takes seventy percent of that, the runner-up thirty — and the remaining ten percent is skimmed to the shootout. Every dollar of edge must come from out-selecting the room, not from the game itself. We tested whether any selection strategy clears that drag, running a full battery across both years, thirty-three one-team-per-flight picks each. Every positive strategy's confidence interval includes large losses, and the two years flip signs. Buying human-intel sandbaggers was the best performer at a combined +77%, but on only seven picks with a confidence interval running from −45% to +192%. Buying mid-price returned +4%. Buying favorites lost 15% combined. The most robust rule was a "don't": the cheapest longshot in a rich flight lost 78%. A Monte-Carlo of fifty thousand random portfolios turned a profit 32% of the time, so a two-year positive result cannot be called skill. Three independent statistical tests agreed that net results here are almost entirely noise. Year-over-year correlation of team finishes is r = −0.12 ; teams that overperform regress, implying a skill share of variance near zero. Within a flight, locked handicap explains just 0.6% of points variance, because the net format equalizes ability. A handicap model run out-of-sample scores worse than a flat one-in-six guess; even the strongest 2026 team is only about a 43% favorite in the simulator, and that precision is a fact about the model, not the world. Implication. We do not pay for "who wins a flight." We treat flights as near-lotteries and hunt the one place an edge can live — price , not outcome. Full reports: Portfolio Backtest · Statistical Analysis · Monte Carlo Section 2 What Predicts Performance — The Player, Not the Card The billing predicts the floor, not the ceiling, and beating your billing sticks to individuals, not teams. The billing barely predicts anything. Within a flight, the auction price explains only three to five percent of the variance in points; each full step up the price ladder is worth +0.5 points against a real spread that runs from roughly 16 to 34. "Billed for 20, scored 25" is the norm. The residual — what you score over your billing — is almost the entire story, and most of it is noise. But the residual is not all noise. Money predicts the floor and not the ceiling: the cheapest third of teams finish dead last about forty percent of the time, stable across both years, while the priciest third does so about nineteen percent of the time. The market spots duds ; it is a coin flip on champions . Informative floor, uninformative ceiling. At the team level, beating your billing does not stick at all (r = −0.23); last year's hero reverts. When we split teams into players and asked who outscored their billing in both 2024 and 2025, a short list fell out. Estes beat his by +6.7 then +6.2. Keister by +4.9 then +8.3. Copeland by +4.1 then +4.2. This residual test is purely statistical — it knows nothing about who plays golf — and the names it surfaces are the same players flagged independently by an eyewitness sandbagger read and by the simulation plus prior results. An eyewitness, a simulation, and a blind two-year residual test point at the same handful of players. Player 2024 over billing 2025 over billing On the buy card? Estes +6.7 +6.2 ✓ Wood + Estes Keister +4.9 +8.3 ✓ Knapp + Keister Copeland +4.1 +4.2 ✓ Wright + Copeland Bachstein +4.1 +4.8 (Hancock + Bachstein) Implication. We do not bet on teams that overperformed, because they regress. We bet on the individuals who reliably outscore their billing. The market reprices teams every year; a few players beat their card season after season. Caveat: two years and a short list — with about 120 players, a few will beat their billing twice by chance. The cross-method convergence is what makes it credible, not the residual test alone. Treat it as a lean, not a certainty. Full report: What Predicts Performance . Section 3 The Edge — Sandbaggers, Markdowns & Match-Play Aptitude The valuation model uses GHIN handicaps, so it is blind to sandbagging, temperament, and the format. Three overlays restore what it misses, and they converge on the same names. The first overlay is the human read on sandbaggers. Williford, of Barbaree + Williford in flight 8, is confirmed : he was watched shooting a 73 in a four-club tournament, which puts roughly scratch real ability behind eight or nine strokes of phantom handicap. He is a flight favorite, not the model's listed 12%. Copeland, of Wright + Copeland in flight 5, is a known sandbagger and also a hothead — temperament that adds variance — and he won flight 5 in 2025 with a 29.5. We buy him with that volatility caveat attached. The second overlay is the mechanics of markdowns, which are real and documented, and which we are careful not to conflate with our own inference. The USGA's Exceptional Score Reduction docks a differential seven to ten strokes better than your index by one stroke off each of your last twenty, and ten-plus better by two; tournament rounds count. A third-party algorithm called Cap Patrol flags sandbaggers across roughly 1,100 clubs using dozens of data points. In our field, the selective-posting tell is dead — 225 of 228 players post every round — so hidden ability surfaces as form and, most usefully, as clutch , a trait-stable, pressure-relevant axis independent of recent form. Positive clutch on an unremarkable index means a player better than his number when it counts, which means the room underprices his team. Our mapping of clutch onto specific players is our edge work: defensible and cross-validated, but inference, not a published ranking. The third overlay is match-play aptitude, the format correction. Handicaps are built for medal play, but we score net better-ball match play, which rewards birdies — winning a hole outright — and grants blow-up immunity, since you can only lose a hole by one. Using hole-level GHIN data for 215 of 240 players, we credit birdie rate and blow-up rate into a single matchplay_aptitude score: positive means a player should outperform his handicap, negative marks a steady medal grinder who underperforms. This is our trap detector. A cheap team with negative aptitude is not value. Implication. The buy filter is two-sided: positive edge AND positive aptitude . A cheap, positive-edge team with negative aptitude — Gallagher + Hansell, +edge on paper but clutch −7 and a chronic underperforming record — is a trap. We let it go and take Wood + Estes instead. Full reports: Cap Patrol & Psychology · How Markdowns Work · Track Records 2021–2025 Section 4 The Valuation — The Cap Anchor and the Buy Card The $2,000 cap is a behavioral magnet. About eighteen hyped teams race to it and are overpriced there, so value concentrates in the non-capped mid-tier that also passes the aptitude filter. The round-number cap reshapes the whole board into a barbell. It pulls the most-hyped, lowest-handicap teams into a spike at exactly $2,000 — eighteen of them, $36,000 of an $83,700 pool — while the rest of the field keeps a tail below about $1,500. At those cap-anchored prices, sixteen of the eighteen cap-out teams carry negative EV edge . We never start a war there. Two model corrections sharpen the rest of the picture: a 10-stroke handicap-difference cap off the low player, which strips win-probability from very-high-handicap "bomber" teams (concentrated in flight 20, where Barnes + Smelcer drops −5.3%), and the match-play aptitude overlay above. Value lives where hype does not. From there the buy card derives itself. A team earns a place only when it clears the two-sided filter, sits below the cap, and ideally has a result behind it. Five teams clear that bar cleanly, and they are the disciplined core of the night. Lot Team Walk-away Derivation #63 Barbaree + Williford STEAL $700–900 Eyewitness sandbagger; low competition; private info. #94 Wright + Copeland BUY ~$900 Result-confirmed flight winner; beats billing both years; positive clutch tail. #104 Wood + Estes BUY hard cap ~$1,100 Triple-confirmed (intel + model + residuals). Do not chase past the cap; the Sharp is on it. #76 Nodar + Heslep BUY ~$600 Top model P(win) in flight; 2nd in 2025. Contested; at/below max. #20 Knapp + Keister IF CHEAP only if cheap Keister beats his billing both years; early lot. Never the cap-anchored $2,000. Implication. Six of the thirteen originally-flagged targets retain positive EV under the reshape; the others are traps at $2,000. We buy the floor the room mispriced and let the room overpay for the ceiling it cannot predict. Full reports: Model Corrections · Methodology · Matchup Analysis Section 5 The Room — Four Types, and the Night's Playbook We buy the same teams for less than a cap-anchored room will pay. Each type of bidder fishes a known pond. Four archetypes describe the room. The Overconfident Veteran knows everyone and overpays for marquee, low-handicap names, driving them to the $2,000 cap. He pushes the exact cap-outs we avoid, so we let him win the names and never become the underbidder who "saves" him money. The Loyalist bids up older and higher-handicap players he trusts, spreading across forty-plus high-handicap teams. The 10-stroke cap makes high-cappers worth less , so his inflated lane is the worst value on the board. We concede it entirely. The Sharp is a value-hunting, day-trader mind who fishes our exact pond, and the model flags him on Wood + Estes, Wright + Copeland, and Nodar + Heslep. Against him we do not tip our hand and we do not start a war; we set a hard max. The Wild Card is unpredictable and can spike any price out of nowhere, weighted toward the already-hyped lots; we set our number before the lot and hold it. Three properties of the room turn those types into a plan. It is cap-anchored , which we invert — the closer a lot drifts to $2,000, the more certain we are to fold. It is semi-unbudgeted , with no FOMO governor, which means it supplies both the overpays we skip and the steals we take. And its chasers are predictable : the expensive board is well covered, which leaves the quiet middle and the late lots open. Because the auction order is published, we pace against it with full look-ahead. Early, through lot 60, the board is cap-out heavy and cash-rich, so we let the room spend itself; our only early must-buy is Knapp + Keister at #20, taken quietly, after which we go dark. Late, from lot 85 on, the tapped-out discount appears, and three of our five core targets land there — #76, #94, #104 — where our edge and the room's fatigue coincide. Implication. We pre-commit a max per lot in writing before it opens and never raise it live. We let the Overconfident Veteran, the Loyalist, and the Wild Card overpay; we concede the public wars; and we buy quiet mid-tier teams late. This is a behavioral read, not a fitted model; the type tags are heuristics. Update them live. The day-of read on which marquee names the Overconfident Veteran is hot for this year is the highest-value live update. Full reports: Competitor Strategy · Auction Strategy Section 6 The Plan — Full $2,500, Core + Stretch We deploy the entire $2,500, split into a disciplined core that holds the edge and a stretch that is entertainment. The math. The optimistic model puts the core slate near +24% on stake, but that survives only if every optimistic assumption holds at once; each plausible adverse assumption erases roughly the whole edge, and a pessimistic-but-plausible stack lands at −18% ROI . The biggest single threat is an efficient market : if the room prices fairly, there is no floor to harvest. The half-buyback is the owner's option against us: he reclaims the winners (we keep 50%) and leaves us the duds (we keep 100%), so we price every team as if we keep all of the losers and half of the winners. Even assuming the model is right, the forward Monte-Carlo shows a ~39% chance of losing money, a ~20% chance of losing half the stake, and a ~7% chance of a total wipeout. The headline edge rests on about two sandbagger teams. The old advice was to hold a 20–25% reserve and optimize to a number. We deploy the full $2,500 instead; the reserve earns nothing sitting in a pocket, and a great night is part of the objective. The discipline concentrates into the core. The buyback math: effective cost is half the hammer and we own 50%, so $2,500 of effective spend buys roughly $5,000 of hammer capacity. The core — about 55–60%, the edges with hard walk-aways Roughly $1,400 goes to the five teams in Section 4: Barbaree + Williford (#63, the uncontested sandbagger), Wright + Copeland (#94), Wood + Estes (#104, hard cap near $1,100, and we do not chase the Sharp past it), Nodar + Heslep (#76), and Knapp + Keister (#20 if cheap). Pre-set maxes, honored walk-aways, no favorites, none of the eighteen cap-outs. Discipline is worth about thirteen ROI points and is the defense against the winner's curse, since we win the teams we like precisely because we value them above the room. The stretch — about 40–45%, −EV entertainment The remaining ~$1,100 goes to the high-aptitude watch teams — Vaniman + Hatz, Martin + Torres — which carry upside but more competition; a cheap longshot or two in a rich flight; and a fun or home pick. This is −EV spend by design. The edge does not live here; the action does. Tranche ~Effective $ Holdings Label Core EDGE ~$1,400 (55–60%) #63, #94, #104, #76, #20 — hard walk-aways Where the edge lives. Disciplined. Stretch FUN ~$1,100 (40–45%) Vaniman + Hatz, Martin + Torres, a longshot, a home pick −EV. Entertainment. The discipline carries through: the near-lottery nature, "we price the player not the card," avoiding the eighteen cap-outs, the match-play-aptitude filter, and the pacing — spend lighter early, push late. The edge is thin. We spend the whole bankroll rather than optimize to a reserve. Bottom Line This is a near-zero-edge, high-variance game. We do not know who wins a flight. Our one durable lever is buying mispriced floors below value, with discipline, while a cap-anchored room overpays for ceilings it cannot predict. The rest goes to a great night, labeled as such. The market prices the card. We price the player. Hold the core to its walk-aways, enjoy the stretch for what it is, and let the room supply both the overpays and the steals. Final live numbers are on the board. Appendix — risk of ruin (model-optimistic forward sim) Metric (model-optimistic) Value P(lose money) ~39% P(lose ≥ 50% of stake) ~20% P(total wipeout) ~7% Under uniform (efficient-market) probabilities, every slate loses money. Under buyback adverse selection the mean goes to ~$0 with a 52% chance of losing money. Full reports: Risk & Stress Test · Forward P&L Simulation · Portfolio Backtest · Investor Memo ← The One-Pager ← Back to Calcutta Live ============================================================================ # How It Works (Methodology) Source URL: https://calcutta.high.green/reports/methodology ============================================================================ How We're Playing the 2026 Member-Member Calcutta A plain-English guide for the buy team — the partners pooling money with Nick. No data-science background required. By the end you should understand exactly what we're buying, why we trust the numbers, and how we'll behave in the room Wednesday night. --- 1. The big idea (in three sentences) We treat the Calcutta like buying stocks: every team has a fair price (what its shot at the flight prize money is actually worth), and we hunt for teams selling below that fair price. We get our fair prices by combining four things — the inside knowledge of a longtime member, what teams actually did in past years, a computer that simulates the matches tens of thousands of times, and a form/clutch algorithm as a tiebreaker — stacked in that order of trust. Then we spend our bankroll with discipline: a few strong teams in the flights where the money actually pools, one team per flight, and cash held back for late bargains. --- 2. How a team makes money (the payout rules) This is the single most important thing to understand, because it shapes everything else. The Calcutta money is not one big shared pool. It splits two ways: - 90% of each flight's money stays inside that flight. Whoever owns the flight winner gets 70% of that flight's pot; the runner-up gets 30%. - 10% of the total event pot is skimmed off the top to fund a separate shootout among flight winners. What that means for us: the value of a team comes almost entirely from how its own flight does — both how likely your team is to finish 1st or 2nd, and how much money that particular flight attracts at auction. A flight that draws a lot of bidding has a big pot to pay out; a "no bueno" flight with thin bidding pays out little even if you win it. > We deliberately ignore the shootout. It's only 10% of the money, it's an overall-winner crapshoot across the whole field, and chasing it would distort our bids. We value teams on flight money only. If we happen to win the shootout, that's a free bonus, not part of the plan. What the flights pay (projected 2026) The flights are seeded by handicap — Flight 1 is the lowest-handicap (best) players, Flight 20 the highest. Money concentrates in the low-numbered flights: those teams cost more at auction, so those flights build bigger pots. Here's roughly how the pots stack up: | Flight tier | Projected pot (each) | Winner gets (70% of 90%) | Runner-up gets (30% of 90%) | |---|---|---|---| | Top (Flight 1) | ~$9,700 | ~$6,100 | ~$2,600 | | Upper-middle (Flights 2–6) | ~$5,400–$7,500 | ~$3,400–$4,700 | ~$1,500–$2,000 | | Middle (Flights 7–13) | ~$3,200–$4,900 | ~$2,000–$3,100 | ~$870–$1,300 | | Bottom (Flights 14–20) | ~$1,200–$2,900 | ~$760–$1,800 | ~$320–$780 | The total projected pot across all 20 flights is about $84,000. Final numbers depend on what actually sells in the room — these are projections to value against, not guarantees. Takeaway: winning a low-flight is worth several times more than winning a high-flight. That's why we concentrate on where the money pools. --- 3. How we estimate each team's chance to win its flight Here's the format every flight plays: six teams, round-robin. Each team plays the other five in a 9-hole, two-man, net better-ball match (best ball of the partners, on each hole, with handicap strokes). You earn match points (a point per hole won, a bonus for winning the match), and the team with the most total points across all five matches wins the flight. Second-most is runner-up. We can't know in advance who wins — golf is too random for that. So instead of guessing, we let a computer play the whole event out about 40,000 times (technically up to 60,000 in the main model). Each simulated run: 1. Draws a realistic round for every player, based on their actual scores over the last 12 months (pulled from GHIN). Hot players, streaky players, steady players — each gets a scoring spread that matches their real recent golf. 2. Adds the things that move teammates together — a hard or easy weather day lifts the whole flight; partners sharing a cart, conditions, and nerves move together too. 3. Plays all five matches hole by hole, applies the handicap strokes correctly (off the low player in each foursome), and tallies the points. 4. Records who won the flight and who came 2nd. Do that 40,000+ times and you get a clean answer: "This team won the flight in 43% of simulations, came 2nd in 16%." Those percentages are the team's win and runner-up probabilities. They're not a hunch — they're the box score of 40,000 imagined tournaments. Why this format produces a lot of upsets: 9 holes is short. There aren't enough holes for the better team to grind out its edge, so even strong favorites usually land around 20–25% to win a six-team flight (a coin-flip field would be ~17% each). A genuinely dominant team — like Wood + Estes, who project around 43% — really stands out, because the format is built to flatten everyone else toward the pack. Fair dollar value, finally: once we have a team's chance to win and place, the math is simple — > Fair value = (chance to win × winner's share + chance to place × runner-up's share) × that flight's pot. A 43% shot at winning a ~$4,000 flight is worth real money; a 12% shot at a $2,500 flight is worth very little. Every team on our board has a fair value computed exactly this way. We then compare it to what we expect the team to actually sell for — and the gap (the "edge") is what we're shopping for. --- 4. The four data sources — stacked from most to least trusted The model above is powerful, but it only knows handicaps and posted scores. It is blind to sandbagging, to temperament, and to who has actually won when it mattered. So we layer four sources of information, and when they disagree, the higher one wins. Tier 1 (most trusted) — Jason's human knowledge Jason has decades of inside read on these members — who's a known sandbagger, who folds under pressure, who's quietly become a different golfer. The model can't see any of this, and here the human read beats the algorithm. Concrete examples driving our 2026 board: - Williford (Barbaree + Williford, Flight 8): the model, going off his ~17 index, rates this team a weak 12% — a "trap." But Nick personally watched Williford shoot 73 in a four-club tournament — that's near-scratch ability hiding behind 8–9 strokes of phantom handicap. The model is simply wrong about him because it's reading a stale, inflated number. Human read overrides: this is a real threat, not the bottom of the flight. - Copeland (Wright + Copeland, Flight 5): known sandbagger, and the result proves it — they won Flight 5 in 2025. We buy, with a caveat: Copeland's a hothead, so expect volatility. - Fades: Jason's read also tells us when to stay away — e.g., the Doyle teams (the price is inflated off a 2025 Doyle flight win) and Gallagher + Hansell (let Jason have that one; we'd rather own Wood + Estes in Flight 9). Tier 2 — Prior-year results (2021–2025 track records) What teams actually did is hard evidence the model can't capture. We pulled five years of finishes. The encouraging part: several of our 2026 targets already won their flight last year, which validates that we're pointed the right way: - Wood + Estes won their flight in 2025 — our #1 target this year. - Downey + Shearer won their flight in 2025 — a Jason target and a model bargain. - Gelinas + Chafin won their flight in 2025. - Wright + Copeland won Flight 5 in 2025 (confirming the Copeland sandbag). - Nodar + Heslep and Kessler + Brinson were runners-up in 2025 — both also flagged cheap by the model this year. When a team is both a model bargain and a recent proven winner, that's our highest-conviction buy. Tier 3 — Our simulation (the 40,000-run model from Section 3) This is the workhorse and the source of every fair-value number. It's trustworthy and unbiased, but it only knows what's in the data (handicaps + posted scores). So it anchors the board, and the two human layers above correct it where they know better. Tier 4 (least trusted) — the Cap Patrol form/clutch algorithm Cap Patrol is a third-party tool that scores each player on recent form ("playing hot right now?") and clutch ("performs under pressure?"). We use it strictly as a cross-check, never as a thesis on its own — it's "just an algo," and form especially tends to regress (a hot streak usually cools). Its useful moments are confirmation: e.g., it independently flagged Vola + Kerns (our Flight 1 favorite) as over-performing their handicap — nice corroboration from a different direction. But if Cap Patrol disagrees with Jason or with a proven result, we go with the human and the result. > One-line summary of the stack: Jason's read and the record book tell us where the model is wrong; the model tells us the price; Cap Patrol just nods along. We also run a matchup check — a round-by-round look at each favorite's six head-to-head matchups — to flag whether a favorite is robust (wins across the board, like Wood + Estes) or fragile (its lead hinges on a couple of coin-flips). In this short-format, most favorites are fragile, which is another reason we don't overpay for any one team. --- 5. The live auction strategy Knowing fair values is half the game. The other half is behavior in the room. Our plan: Buy where the money pools. Because 90% of value is local to each flight, and the low-numbered flights build the biggest pots, we concentrate there. A few flights are "no bueno" — thin pots that don't pay enough to bother (even when a target like Matt Smith is playing hot in Flight 15, the small pot makes it low-priority). We fade those. One team per flight. We never own two teams in the same flight — they'd partly cannibalize each other (only one can win), and it wastes capital. One shot per flight, our best-value team in it. Fade the cheap longshots. The crowd loves a cheap Cinderella. In a Calcutta those are usually the worst value — a 10% team in a small flight is almost never worth its price. Solid favorites and strong middles are the sweet spot. We don't "hunt upsets"; we buy value. Hold a reserve. With random auction order, a bargain can show up at any moment — especially late, when other bidders run out of cash and prices sag. We keep roughly 20–25% of our bankroll in reserve to pounce on those late deals rather than blowing it all early. The economics (this is the part to really get) - Bankroll: $2,500 for the buy team. - The half-buyback rule changes everything. By gentleman's agreement, the team itself can buy back up to half from whoever wins it at auction. We expect to give back half of every team we win. So if we win a team for $1,000, we hand half the team back, collect $500, and our effective cost is $500 for a 50% stake. Read every hammer price as half that to us — and every future payout as half too. (Win the flight for $6,000? Our 50% share is $3,000 on a $500 effective outlay.) - Target: 5–7 teams in the $800–$1,200 range (hammer price). After the half-buyback, that's roughly $400–$600 of real money per team, which fits a $2,500 bankroll with reserve to spare. - The $2,000 cap helps us. No team can sell above $2,000, and the very best teams get bought by their own owners at the cap — removing them from our market. The exploitable value lives in the strong-but-not-elite middle, which is exactly where we shop. --- 6. What the website / tool does live, in the room The numbers in this document are the prep. On auction night we run a live tool that updates as teams sell: - Tracks all 20 flight pots in real time (plus the grand total), because a team's value depends on how much money its flight is pulling — and that's only known once teams start selling. - Re-prices every unsold team after every sale, comparing live fair value to the current bid, and surfaces the best remaining values so we always know our top targets. - Sets a walk-away (max) price for each team and flags "pounce" moments — when a team is selling below its fair value and we have the reserve to grab it. - Tracks our spend and reserve so we don't overspend early or finish with idle cash. The bottom line: the tool keeps us disciplined and adaptive so we're never bidding on gut feel mid-room. --- 7. Honest limitations We're confident in this approach, but we won't oversell it: - It's a model, and golf is variance. A 43% favorite still loses its flight more than half the time. Our edge is buying lots of good value, not being certain on any one team. Expect some of our best buys to whiff — that's normal and already priced in. - The format is genuinely random. 9-hole net better-ball is near-coin-flip; most favorites are "fragile." We lean into that by spreading bets and never overpaying. - The model only knows the data. It can't see sandbaggers or chokers on its own — which is exactly why the human layers (Jason, results) sit above it. - The pot projections are estimates. We have no actual 2026 prices yet, so the flight pots are projected from 2025 (adjusted for the bigger field and higher cap). The live tool corrects these as real prices come in. - We ignore the shootout on purpose. It's a small slice and a crapshoot. Not worth distorting our bids for. --- Where the actual numbers live This document explains the method. The specific, final per-team fair values, recommended max bids, and live updates live in the bid sheet and on the auction site — those are what we'll actually act on Wednesday night. If a number here and a number there ever disagree, trust the bid sheet / site — it's the live, current source. ============================================================================ # Statistical Analysis Source URL: https://calcutta.high.green/reports/statistics ============================================================================ Independent Statistical Audit 2026 Member-Member Calcutta — Expert Statistical Review Independent statistical audit of the valuation model. All numbers are computed from the project's own data ( data/processed/ , data/raw/history/ , valuation/ ). Scripts: prep.py , models.py , figures.py , build_report.py in this folder. Reproduce with uv run python analysis/stats_expert/{prep,models,figures,build_report}.py . Executive summary The single most important empirical fact in this dataset: net match-play results in this event are almost entirely noise, not skill. Three independent tests agree: - No year-over-year persistence. Teams that finished above their flight average in 2024 did not tend to repeat in 2025. The correlation is r = -0.12 (n = 33 repeat teams; bootstrap 95% CI [-0.41, +0.17] , which straddles zero). For individual players it is r = -0.14 (n = 121). A team's finish one year tells you essentially nothing about the next. - Handicap does not predict the net outcome. Within a flight, the team's locked handicap explains 0.6% of the variance in match points (R² = 0.006, p = 0.39). That is by design - the net format neutralizes handicap - but it means handicap-driven win-probability spreads must be treated with great caution. - A leave-one-year-out handicap model loses to a coin flip. Predicting each flight's winner from handicap scores worse than a flat 1-in-6 guess on both log-loss (1.93 vs 1.79) and Brier (0.79 vs 0.76). There is no exploitable signal for who wins a flight . What this means for the auction. The live model's win probabilities are already appropriately humble - within-flight they sit at 97% of maximum entropy (nearly uniform), and even the strongest 2026 team is only a ~43% favorite in its flight. That is the right posture. But it also means edge does not come from "we know who wins." It comes from two places the data does support: - A small, real track-record signal. A properly shrunk (empirical-Bayes) player-ability estimate has a between-player SD of about 1.0 points on a points scale whose noise SD is ~4.4 - tiny, but the best signal available, and nearly uncorrelated with handicap (r = 0.05) . A 2026 field ability table built from it is in section 7. Use it as a tie-breaker, not a thesis. - The price side, done right. Historical prices are right-censored at the bid cap (14% of 2024 prices were pinned at \$1,000). OLS on censored prices attenuates slopes and understates dispersion; a Tobit model corrects both (σ inflates ~13% on 2024). The live model's hand-tuned PRICE_SPREAD = 1.85 is a crude patch for exactly this - a principled Tobit replaces the fudge. Bottom line for a sharp partner: Don't pay up for "favorites" inside a flight - the format makes flights close to a lottery, and five years of results confirm it. Spend the analytical effort on (a) not overpaying relative to the censored-corrected price curve, and (b) leaning, gently, toward teams with a genuine multi-year track record of beating their flight. Everything else is variance. 1. Critical assessment of the current methodology The live model ( valuation/value_model.py ) is a careful, well-documented hole-by-hole Monte-Carlo simulation of the confirmed format (round-robin 9-hole net better-ball, points accumulation). Its engineering is sound and its instinct - that 9-hole net match play is high-variance - is correct. Three critiques, in priority order: (a) It is calibrated to handicap-derived scoring, but the outcome is nearly handicap-independent. Every player's scoring distribution flows from GHIN differentials and the locked handicap. Yet in five years of actual results, handicap carries no information about the net finish (section 3). The simulation's win-probability spread is therefore driven by an axis the historical record says is close to noise. To its credit, the model's output is already very flat (entropy ratio 0.97), so the practical damage is limited - but any temptation to trust a "43% favorite" as if it were real should be resisted. The honest prior is much closer to 1/6, widened only slightly by track record. (b) The price model uses OLS on censored data, then re-inflates by hand. Prices are capped at the bid limit (\$1,000 in 2024, \$1,800 in 2025, \$2,000 in 2026). OLS treats a \$1,000 censored price as if the team were truly worth exactly \$1,000, which flattens the slope and shrinks the dispersion - precisely the symptom the code comments describe ("raw OLS is far too compressed"). The fix is a textbook Tobit (censored-normal MLE), fit in section 8. PRICE_SPREAD = 1.85 and PRICE_LEVEL_PER_TEAM = 700 are reasonable hand-corrections, but they are unidentified knobs; Tobit estimates the same correction from the data. (c) The elaborate variance structure (partner/field/hole correlations) is set by assumption, not fit. RHO_PARTNERS , RHO_FIELD , HOLE_NOISE_FRAC are judgment calls. Given finding section 3 - that we cannot even detect skill in the results - we certainly cannot estimate these second-order correlations from data. This isn't wrong, but it should be labeled as a reasonable prior, not an inference. The model's existing sensitivity analysis is the right way to handle it. Net assessment: the model is over-engineered relative to the information content of the data. That is not a knock on the craft - it is a statement that the data has very little to say about who wins , and a model that produces sharp-looking 43% favorites can give a false sense of precision. Its saving grace is that its outputs are nearly flat anyway. 2. Data and methods Asset What it is Used for hist_team.csv 198 team-flight RESULTS (2024-25), match points (0-50) + finishing position repeatability, Plackett-Luce, calibration backtest hist_player.csv the same exploded to 396 player-seasons, joined to locked handicap (71% matched) empirical-Bayes player abilities teams26.csv 2026 field (120 teams), locked handicaps, GHIN ids field ability table prices.csv 2024-25 auction prices, flagged censored at the per-year cap Tobit price model Response variable. Within each 6-team flight the round-robin distributes exactly 150 match points, so a team's points are zero-sum within its flight. We center points within flight-year ( pts_centered ) so "points above the flight average" is the comparable, mean-removed response. Typical within-flight points SD ~ 4.37 . Methods applied (each justified against our data, not a textbook tour): - Variance components / repeatability - measure how much signal exists before fitting anything elaborate. - Hierarchical empirical-Bayes (partial pooling) for player abilities - the modern shrinkage estimator; tells us how far to trust a player's track record. - Plackett-Luce (the multi-competitor generalization of Bradley-Terry) on within-flight finishing orders - the natural paired-comparison frame for match play, fit to actual results, with an out-of-sample test. - Proper scoring rules (log-loss, Brier) with a leave-one-year-out backtest and a reliability chart - validate, honestly, whether any predictor beats a uniform baseline. - Tobit (censored-normal MLE) vs OLS for prices - the correct estimator under a bid cap. - Bootstrap and Wilson intervals - uncertainty on the repeatability correlation and the Monte-Carlo win probabilities. We deliberately did not reach for PyMC/Stan: with this little signal, a full Bayesian hierarchy adds machinery without changing conclusions. Empirical-Bayes shrinkage gives the same partial-pooling answer in closed form. 3. Is there any signal? Variance components and repeatability Each point is a team that played in both 2024 and 2025; axes are its points relative to its flight average. If finishing well were a durable property, the cloud would slope up. It doesn't. Quantity Value Within-flight points SD 4.37 Team year-over-year repeatability r -0.119 (n=33, p=0.51) Player year-over-year repeatability r -0.143 (n=121, p=0.12) Bootstrap 95% CI on team r [-0.41, +0.17] (includes 0) Implied skill share of points variance ~ 0% Under the standard signal+noise decomposition, the between-year correlation equals the skill share, var_skill / (var_skill + var_noise) . With r ~ 0 (and not significantly different from it), the estimated skill variance is essentially zero : net match-play points behave like draws from a common distribution. A permutation test for repeatability returns p ~ 0.50 - exactly what pure noise gives. This does not mean players have no golf skill. It means the net, better-ball, 9-hole, points-accumulation format successfully equalizes the field - the point of a member-member. The residual that decides finishes is dominated by which day you caught lightning. 4. Hierarchical empirical-Bayes player abilities (partial pooling) We fit the one-way random-effects model y = mu + a_i + e_i , with player effects a_i ~ N(0, &τ;²) and residual e ~ N(0, &σ;²) , then shrink each player's mean toward zero by the optimal factor n_i / (n_i + &σ;²/&τ;²) . Component Estimate Within-player residual SD (σ) 4.24 Between-player SD (τ) 1.03 Shrinkage constant k = &σ;²/&τ;² 17.1 Mean shrinkage weight 0.087 With most players having only 1-2 seasons, the shrinkage weight is ~0.09 - i.e., we pull raw player means ~91% of the way back to the field average . This is the statistically correct response to a tiny between-player variance: believe almost none of a one-year result. The shrunk abilities collapse toward the origin: a player who "won" his flight by a mile in one year is credited with only a fraction of a point of durable ability. This is the honest version of a power ranking. The resulting 2026 field ability table is in section 7. 5. Plackett-Luce latent abilities from match results (paired-comparison frame) Match play is paired comparison, so we fit a Plackett-Luce model (Bradley-Terry generalized to full finishing orders): within each flight, the finishing order is a sequential "pick the best remaining" draw governed by latent team strengths θ. Teams in multiple flight-years share a strength; an L2 ridge regularizes toward equal strengths. Result Value Teams / races 163 / 33 In-sample log-lik (fit vs null) -181.6 vs -217.1 Out-of-sample (2024-learned θ -> predict 2025 order) -100.3 Out-of-sample with equal strengths -98.7 Learned strengths beat equal out-of-sample? No The in-sample likelihood improves - but that is overfitting : 163 free strength parameters on 33 short races will always fit the noise. The decisive test is out-of-sample, and there the strengths learned from 2024 fail to beat assuming all teams equal when predicting 2025. This corroborates section 3 from a completely different modeling family: the latent team strength match play would estimate is, here, indistinguishable from noise. That is the correct, modern way to discover a paired-comparison model has nothing to grip. 6. Validation with proper scoring rules We score three predictors of the flight winner against the 33 actual flight winners (2024-25), using multiclass log-loss and Brier . Ties for first split the winner mass. Predictor Log-loss (lower=better) Brier (lower=better) Uniform 1/6 (no-signal baseline) 1.792 0.758 Handicap softmax - best in-sample temperature 1.792 (β*=0.00) 0.758 Handicap softmax - honest leave-one-year-out 1.933 0.789 Two things stand out. First, the in-sample optimizer drives the handicap temperature to zero - the best the handicap model can do is ignore handicap and predict uniform . Second, the honest leave-one-year-out handicap model is worse than uniform on both rules. A predictor that loses to a coin flip out-of-sample carries no usable information about the winner. The reliability view is blunt: bin every historical team by its within-flight handicap rank (1 = lowest handicap = nominal favorite) and plot the realized win rate. Every bin hovers around the 1/6 baseline. The nominal favorite wins about as often as the nominal longshot. There is no monotone "favorites win more" gradient to calibrate against. (We cannot score the live 2026 Monte-Carlo model directly on history - it only produces 2026 teams - so we test the axis it relies on: handicap-implied favoritism. That axis is flat. The live model's own outputs are, appropriately, also nearly flat: entropy ratio 0.97, median flight-favorite p_win just 0.23.) 7. A defensible 2026 field "ability" table Given sections 3-6, the most defensible team ranking is not the handicap-driven win probability - it's the shrunk track record . For each 2026 team we sum its two players' empirical-Bayes abilities (section 4). This is a small signal (team-level SD ~ 0.48 points, against a ~4.4 noise SD) and should be read as a gentle lean with wide uncertainty, not a prediction. Critically, it is nearly uncorrelated with the live model's p_win (r = 0.05) and with team handicap (r = 0.09) - genuinely orthogonal information. Top of the field by historical track record (shrunk): Flight Team Team hcp EB ability Live p_win 20 Ballard + Armstrong 47.4 +1.29 14% 13 Panessa + Ratliff 23.1 +1.25 17% 9 Wood + Estes 19.3 +1.22 43% 19 Vaniman + Leingang 35.5 +1.15 12% 19 Truitt + Fetter 36.0 +1.14 11% 10 Gelinas + Chafin 19.7 +0.97 16% 4 Green + Green 11.5 +0.91 19% 5 Wright + Copeland 12.2 +0.91 20% Bottom of the field by track record (have historically under-performed their flight): Flight Team Team hcp EB ability Live p_win 5 Pearson + Ferguson 12.6 -1.12 21% 10 Shirley + Aronson 20.8 -0.93 23% 12 Collins + Minter 22.5 -0.86 20% 15 Jaillet + Jaillet 26.0 -0.81 17% 3 Embleau + Loewenthal 9.1 -0.81 18% Note the disagreements: Wood + Estes is both a track-record leader and the live model's strongest favorite - a rare case where two independent signals align (lean in). Conversely several teams the live model likes (high p_win) have negative track records - their favoritism rests on handicap, which section 6 shows is not predictive. Full table: field_ability_2026.csv . 8. The price model, done right: Tobit vs OLS Auction prices are right-censored at the bid cap . The censoring is what makes OLS misbehave. The red bar is the spike of 2024 teams pinned at the \$1,000 cap - 14% of the field . OLS treats each as a precise \$1,000 observation, biasing the slope toward zero and shrinking the fitted dispersion. 2024 (severe censoring, 15/108 capped), log-price ~ flight seed: OLS Tobit Seed slope 0.0448 0.0485 Residual σ 0.361 0.407 σ inflation (Tobit/OLS) - x1.13 Tobit recovers a steeper seed effect and 13% more dispersion - the strongest teams were worth more than the censored OLS fit implies, and the price spread is wider. This is exactly what PRICE_SPREAD = 1.85 approximates - but Tobit estimates it from the data instead of guessing. 2025 (mild censoring, 1/90), log-price ~ handicap + within-flight rank: here Tobit ~ OLS (σ ratio 0.99, slopes within 0.0002), because with almost no censoring there is nothing to correct. The honest read: use Tobit and let the data decide how much correction is needed - it self-deactivates when the cap doesn't bind (2025) and engages when it does (2024, and likely 2026 at the top). Recommendation. Replace the OLS-plus- PRICE_SPREAD -plus- PRICE_LEVEL stack with a single Tobit fit on pooled 2024-25 data with a year fixed effect and each year's cap, then project to 2026 at the \$2,000 cap. It removes two unidentified knobs and is the standard estimator for capped prices. 9. Uncertainty quantification - Win probabilities. The Monte-Carlo p_win for even the strongest 2026 team (~0.426) is estimated tightly - Wilson 95% interval [0.422, 0.430] at N=60,000 sims. But that precision is about the simulator , not the world: it answers "what does the model output," not "how often will they actually win." Section 3 says the real distribution is much closer to 1/6 with wide irreducible spread. - Repeatability / is there skill. Bootstrap 95% CI on the year-over-year team correlation is [-0.41, +0.17] - consistent with zero and even mildly negative. We can rule out a large skill component, not a tiny one. - Player abilities. Posterior SDs in player_abilities_eb.csv are large relative to the abilities themselves for nearly every player (most are within ~2 SD of zero) - formal confirmation individual rankings are unreliable. 10. Assumptions and limitations - Tiny samples. Two years of results (33 flights with a repeat team; 121 repeat players). Absence of detectable signal is not proof of absence - but it caps how confidently anyone can claim to "know who wins." - Surname linkage. Historical teams join to current handicaps by surname (71% of player-seasons matched). Common surnames take a median handicap; a few mismatches are possible and would only add noise, not create false signal. - Points vs true match outcomes. We have flight standings (points, finish), not hole-by-hole cards. Enough for finishing-order and points models, but not to estimate the live model's hole-level correlation parameters from data - those remain assumptions. - Prices != value. Tobit models price formation (what the room paid). Right tool for the auction-price curve, but it inherits the room's inefficiencies. - Format stability. Assumes the 2026 net better-ball points format behaves like 2024-25. A material format or field change would require refreshing the repeatability estimates. 11. Recommendations - Treat within-flight win probabilities as close to uniform. Don't pay a premium for a "favorite" - five years of results say flights are nearly lotteries. The live model already leans this way; trust that humility over its sharpest numbers. - Use the shrunk track-record table (section 7) as the tie-breaker , not handicap. Small, but the only signal that survives an honest out-of-sample test, and orthogonal to handicap. Favor teams where track record and the model agree (e.g., Wood + Estes); discount model favorites with negative track records. - Replace the OLS+spread price stack with a pooled Tobit (year FE + per-year cap -> project to \$2,000). Same intent as PRICE_SPREAD , but estimated, identified, and self-deactivating when the cap doesn't bind. - Hunt for price edges, not outcome edges. Because outcomes are near-random, value comes from buying teams below the censored-corrected price curve, not from forecasting winners. Define bargains/traps against the Tobit price; down-weight win-prob differences. - Keep collecting results. The fastest way to settle whether any skill signal exists is more years; at 5+ the repeatability CI would tighten enough to decide it. Generated from the project's own data. Methods: variance components, empirical-Bayes partial pooling, Plackett-Luce, proper scoring rules (log-loss / Brier), Tobit censored regression, bootstrap & Wilson intervals. Scripts in analysis/stats_expert/ . ============================================================================ # Monte Carlo Analysis Source URL: https://calcutta.high.green/reports/montecarlo ============================================================================ Expert Monte-Carlo Analysis 2026 Member-Member Calcutta Advanced variance reduction · Quasi-Monte Carlo · Sobol' sensitivity · Copula tails · Nested uncertainty  |  engine reproduces value_model.py bit-for-bit · seed 20260609 Executive summary 1. Correctness & efficiency audit 2. Convergence diagnostics & error bars 3. Variance reduction (with measured VRFs) 4. Quasi-Monte Carlo convergence 5. Global sensitivity analysis (Sobol' indices) 6. Dependence modeling: copulas & the run-the-table tail 7. Two-level uncertainty propagation (headline teams) 8. Recommendation: the optimal simulation design Assumptions & limitations Executive summary This report audits the production 2026 Calcutta flight simulator ( valuation/value model.py ) as a Monte-Carlo estimator and applies a stack of advanced variance-reduction, quasi-Monte-Carlo, global-sensitivity, dependence-modeling, and nested-uncertainty techniques on our actual flights . Every number below comes from a self-contained engine ( mc engine.py ) that reproduces the production per-hole, 9-hole net better-ball round-robin bit-for-bit (validated to MC noise). Headline findings: - The simulator is correct and unbiased. Independent-stream Gelman-Rubin R-hat = 1.000 (Flight 9) and 1.000 (Flight 1); batch-means standard errors agree with the binomial prediction √(p(1-p)/N) to within sampling noise. Draws are genuinely i.i.d.; there is no hidden autocorrelation to correct. - 60k sims is more than enough for ranking; 10k–20k is enough for most decisions. A favorite at p≈0.42 has a 95% half-width of ±0.39 pts at 60k, ±0.68 pts at 20k. Hitting ±0.5 pt on a 0.25 win-prob needs ~27,587 sims. - Variance reduction buys a large effective-sample-size multiple essentially for free. On Flight 9 the best technique ( the expected-points control variate ) gives a variance-reduction factor of 2.1× — i.e. it reaches the same precision as plain Monte Carlo with ~2.1× fewer sims. Antithetic variates alone give 2.0×; the expected-points control variate gives 2.1×. - Quasi-Monte Carlo helps at small N but the gain fades — don't rely on it. Scrambled Sobol' over the 127-dim shock vector gives 1.5–2.5× lower RMSE than pseudorandom at budgets ≤8k, but its fitted convergence slope flattens to N -0.37 (vs pseudo's textbook N -0.53 ), so by ~16k the advantage washes out — the discontinuous, high-effective-dimension integrand denies QMC its O(N⁻¹) regime. The durable efficiency win is variance reduction, not QMC. - Sensitivity: player quality, not the correlation knobs, drives the answer. A Sobol'/Saltelli variance decomposition attributes ≈82% of the favorite's win-probability variance to its own players' assumed scoring level and consistency. HOLE NOISE FRAC is the leading structural knob; RHO_FIELD is essentially inert (it cancels within matches). Calibration effort belongs on the player distributions, not on the correlation constants. - Two-level (parameter) uncertainty dwarfs Monte-Carlo noise. Propagating the player scoring-distribution uncertainty gives Wood + Estes a win-probability band of 34.6%–46.3% (point 41.9%) and Vola + Kerns 16.0%–30.6% (point 23.5%). Parameter uncertainty is ~9× the MC noise at 15k sims — so spending sims past ~20k to shrink MC error is false precision . 1. Correctness & efficiency audit What the production sim does (restated for audit). For each 6-team flight it simulates the full round-robin (15 matches) hole by hole: each player's 9-hole round is built from a per-hole gross-vs-par mean (m18/18) plus four Gaussian shock layers — a flight-wide field/day shock ( RHO FIELD=0.15 ), a per-team partner shock ( RHO PARTNERS=0.35 ), a per-player own round-form shock, and i.i.d. per-hole scatter ( HOLE NOISE FRAC=0.65 ). Gross is rounded to integers and clipped to [par-2, par+7]; net better-ball is taken off the low player in each foursome; standings rank by total match points. Correctness checks (all pass): - Our independent engine reproduces production P(win) to within ±0.0005 (pure MC noise at the test size), confirming the standings/tie-break logic. - Scoring is deterministic given the shocks (verified): identical shock arrays yield identical standings, so the estimator is a clean function of the RNG — a prerequisite for antithetic/QMC/CRN. - Per-flight ΣP(win)=ΣP(2nd)=1 and P(advance)≥P(win) hold by construction. - Gelman-Rubin R-hat ≈ 1.000 across 8 independent streams on both focus flights ⇒ no between-stream disagreement beyond sampling noise. Efficiency audit. The production estimator is plain i.i.d. pseudo-random Monte Carlo. It does use a single fixed seed (reproducible) and, within a scenario sweep (rho=0 vs rho=0.35), it draws fresh randomness each call rather than reusing a common stream — so the reported partner-correlation deltas carry uncorrelated MC noise in each arm. Switching those paired comparisons to common random numbers (same shock stream, only the parameter changes) would sharpen every delta; see the Variance-Reduction and Sensitivity sections. The core sim leaves antithetic variates, control variates, and QMC entirely on the table. Flight Favorite p(win) Binomial SE @60k Batch-means SE @60k R-hat (8 streams) 9 Wood + Estes 0.419 0.00201 0.00172 1.000 1 Vola + Kerns 0.235 0.00173 0.00185 1.000 The batch-means SE (30 non-overlapping batches) tracking the binomial SE is the empirical proof that the draws are independent — if there were autocorrelation, batch-means would exceed the binomial value. 2. Convergence diagnostics & error bars For a win-probability estimate p̂ from N i.i.d. sims, the Monte-Carlo standard error is √(p(1−p)/N) and the 95% half-width is 1.96×that. The table gives the half-width for the two focus favorites across N, plus the sims required to hit a target precision. Flight 9 (Wood + Estes — clear-favorite flight) — favorite Wood + Estes , p(win)≈0.419: N sims 95% half-width on p(win) 5,000 ±1.37 pts 10,000 ±0.97 pts 20,000 ±0.68 pts 40,000 ±0.48 pts 60,000 ±0.39 pts 100,000 ±0.31 pts Flight 1 (Vola + Kerns — coin-flip flight) — favorite Vola + Kerns , p(win)≈0.235: N sims 95% half-width on p(win) 5,000 ±1.17 pts 10,000 ±0.83 pts 20,000 ±0.59 pts 40,000 ±0.42 pts 60,000 ±0.34 pts 100,000 ±0.26 pts Sims needed for a target 95% half-width (favorite p≈0.23): Target half-width Sims required ±1.0 pts 6,897 ±0.5 pts 27,587 ±0.2 pts 172,419 Recommendation on sim count. For *ranking* teams within a flight and producing the bid sheet, 10k–20k sims already pins each P(win) to ±0.4–0.6 pts, which is far below the partner-correlation and parameter-uncertainty effects the model genuinely carries. 40–60k is justified only if you want the advance-to-shootout *tail* probabilities and the cross-flight fair-value aggregates smooth to <0.2 pts. Past 60k you are polishing MC noise that is ≈10–20× smaller than the parameter uncertainty (see §7) — wasted compute. With the control-variate + antithetic stack below, 60k-equivalent precision is reachable at ~15–20k raw sims on clear-favorite flights. 3. Variance reduction (with measured VRFs) We measure each technique's variance-reduction factor (VRF) = Var(plain estimator)/Var(method estimator), estimated from 80 independent replications of an 8,000-sim estimate of the favorite's P(win). VRF = X means the method needs ~X× fewer sims for the same precision (equivalently, ESS is X× larger). Methods: antithetic variates (negate the entire 127-dim shock vector for the paired half); control variate using the favorite's expected match points (cheap, ~known mean, strongly correlated with winning); stratified sampling on the 1-D field/day shock (50 proportional strata); Latin Hypercube on the structured field+partner shocks (the low-dimensional, high-leverage part). Flight 9 (Wood + Estes — clear-favorite flight) — favorite Wood + Estes : Technique Var(estimator) VRF (×) Interpretation Plain pseudo-random (baseline) 2.914e-05 1.00 baseline Antithetic variates 1.448e-05 2.01 2.0× fewer sims Control variate (E[match pts]) 1.408e-05 2.07 2.1× fewer sims Stratified field shock 2.641e-05 1.10 1.1× fewer sims Latin Hypercube (field+partner) 2.149e-05 1.36 1.4× fewer sims Flight 1 (Vola + Kerns — coin-flip flight) — favorite Vola + Kerns : Technique Var(estimator) VRF (×) Interpretation Plain pseudo-random (baseline) 1.639e-05 1.00 baseline Antithetic variates 1.560e-05 1.05 1.1× fewer sims Control variate (E[match pts]) 1.352e-05 1.21 1.2× fewer sims Stratified field shock 1.589e-05 1.03 1.0× fewer sims Latin Hypercube (field+partner) 1.659e-05 0.99 1.0× fewer sims Reading the result. The control variate is the standout: the favorite's expected match-point total is almost a sufficient statistic for whether it wins the flight, so regressing the win indicator on it removes a large share of variance at near-zero extra cost. Antithetic variates give a solid, free boost because the win indicator is close to monotone in the aggregate shock. Stratifying/LHS the field shock helps less here than one might expect — the field/day shock largely *cancels within a match* (both teams feel it), so it drives cross-match point totals but not single-match outcomes; the bulk of the variance lives in the 120 idiosyncratic per-hole/own-form dimensions, which stratification on 1 dimension cannot touch. Common random numbers (already partly used inside a match) should additionally be applied across scenario reruns — it is the single highest-leverage change for the partner-correlation deltas the production report publishes. Recommended stack: antithetic + expected-points control variate + CRN across scenarios. The two are near-independent, so on a clear-favorite flight (Flight 9) they compound to ~3–4× — 60k-quality precision from ~15–20k raw draws. On a flat coin-flip flight (Flight 1) the gains are smaller (~1.2–1.3×): when p(win)≈1/6 across six near-equal teams the win indicator is weakly correlated with any single control and nearly symmetric, so there is less variance for these techniques to remove. The honest summary: variance reduction is a real, free win on the flights where one team separates, and a modest one where the flight is a scramble — but CRN across scenario reruns helps everywhere. 4. Quasi-Monte Carlo convergence Quasi-Monte Carlo replaces pseudo-random draws with a low-discrepancy sequence (scrambled Sobol' or Halton ) that fills the 127-dimensional unit cube more evenly. We map each point through the inverse normal CDF into the four shock blocks and measure RMSE of the favorite's P(win) vs a 250k-sim ground truth, averaged over 16 independent scramblings, across N. Flight 9 (Wood + Estes — clear-favorite flight) — favorite Wood + Estes , truth p=0.4204, dimension D=127: N Pseudo RMSE Sobol' RMSE Halton RMSE Sobol' speedup 256 0.03679 0.01758 0.01849 2.09× 512 0.02387 0.01098 0.01456 2.17× 1,024 0.01631 0.00652 0.00915 2.50× 2,048 0.00975 0.00574 0.00636 1.70× 4,096 0.00912 0.00421 0.00505 2.17× 8,192 0.00588 0.00386 0.00411 1.52× 16,384 0.00374 0.00376 0.00312 1.00× Fitted convergence rate (slope of log RMSE vs log N): pseudo N -0.53 , Sobol' N -0.37 , Halton N -0.44 . Theory: pseudo → −0.5, QMC → up to −1.0. Flight 1 (Vola + Kerns — coin-flip flight) — favorite Vola + Kerns , truth p=0.2347, dimension D=127: N Pseudo RMSE Sobol' RMSE Halton RMSE Sobol' speedup 256 0.02688 0.01161 0.01604 2.32× 512 0.01290 0.01054 0.01097 1.22× 1,024 0.00998 0.00998 0.00973 1.00× 2,048 0.00919 0.00659 0.00710 1.39× 4,096 0.00617 0.00496 0.00430 1.24× 8,192 0.00605 0.00250 0.00343 2.42× 16,384 0.00281 0.00248 0.00235 1.13× Fitted convergence rate (slope of log RMSE vs log N): pseudo N -0.45 , Sobol' N -0.42 , Halton N -0.46 . Theory: pseudo → −0.5, QMC → up to −1.0. Reading the result — a nuanced win that fades. Two facts that look contradictory until you separate *level* from *slope*: (1) at every practical budget from 256 to ~8k draws, scrambled Sobol' delivers 1.5–2.5× lower RMSE than pseudorandom on the favorite's P(win) — a real, free accuracy gain; but (2) its fitted convergence slope is not the theoretical O(N⁻¹) — it flattens to roughly N -0.42 (vs pseudo's textbook N -0.45 ), so the two curves converge and by N≈16k the Sobol' advantage has largely washed out. The cause is effective dimension : the win indicator depends on 127 standard normals (1 field + 6 partner + 12 own-form + 108 per-hole scatter), and the 108 per-hole scatter dimensions stay individually consequential (integer rounding + clipping make single holes pivotal). Sobol' front-loads its uniformity into the first coordinates (here the field + partner + own-form shocks), which is why it helps at low N; but a discontinuous, high-effective-dimension integrand denies it the smooth O(N⁻¹) regime, so the gain does not compound. Recommendation: Sobol' is a worthwhile, zero-cost drop-in *if* you operate at small N (≤4–8k) — pair it with the §3 stack. But it is not a substitute for variance reduction and brings little once you are already at 20k+. The robust efficiency lever here is the antithetic + control-variate + CRN stack, not QMC. 5. Global sensitivity analysis (Sobol' indices) We perform a Saltelli/Sobol' variance decomposition of the favorite's P(win) over five model inputs, each given a uniform prior around its production value: RHO FIELD [0.05,0.30], RHO PARTNERS [0.15,0.55], HOLE NOISE FRAC [0.50,0.80], a favorite mean shift [−1.5,+1.5] strokes (18h-equiv applied to the favorite team's own two players), and a favorite SD scale [0.85,1.15] (likewise). The two player perturbations target the favorite's own team because a field-wide shift cancels in relative standings — we want each input to measure a decision-relevant uncertainty. First-order index S1 = share of output variance explained by that input alone; total-effect ST = share including all interactions. We use common random numbers across every model evaluation, so the indices isolate parameter effects from MC noise (Jansen estimators, Sobol' base N=256 ⇒ 7×256 evals of 16k sims each). Flight 9 (Wood + Estes — clear-favorite flight) — favorite Wood + Estes (baseline p≈0.423, output variance over the prior box = 0.0017): Input First-order S1 Total-effect ST Player mean shift 0.380 0.401 Player SD scale 0.420 0.422 RHO_PARTNERS 0.029 0.034 HOLE NOISE FRAC 0.149 0.153 RHO_FIELD 0.028 0.004 Flight 1 (Vola + Kerns — coin-flip flight) — favorite Vola + Kerns (baseline p≈0.232, output variance over the prior box = 0.0018): Input First-order S1 Total-effect ST Player mean shift 0.414 0.424 Player SD scale 0.563 0.562 RHO_PARTNERS 0.013 0.004 HOLE NOISE FRAC 0.023 0.020 RHO_FIELD 0.008 0.001 Reading the result. The favorite's own player inputs dominate. On Flight 9 the mean-level and SD-scale of Wood + Estes together account for ≈82% of the win-probability variance over the prior box; on Flight 1 the favorite's level + consistency account for ≈99%. In plain terms: *how good we think the favorite's two players actually are — their scoring level and their consistency — drives the answer far more than any correlation or noise constant.* Among the three structural knobs, the leading one is HOLE NOISE FRAC (ST 0.15 on Flight 9): it sets how much of each player's variance lands as un-averaged 9-hole scatter, which is exactly what does or doesn't separate teams over a short match. RHO PARTNERS is a minor contributor (it tunes the better-ball smoothing the favorite keeps), and RHO FIELD is essentially inert for a single team's win probability because the field/day shock cancels within every match. ST≈S1 throughout ⇒ interactions are small. Implication: modeling effort and any future data collection should target the player scoring distributions (recency, shrinkage, sample depth) and — among structural assumptions — the per-hole variance fraction HOLE NOISE FRAC ; fine-tuning RHO_FIELD is wasted effort. 6. Dependence modeling: copulas & the run-the-table tail The production model couples teammates additively (a shared Gaussian partner shock). We compare that against a Gaussian copula and a heavy-tailed t-copula (df=4) on the two teammates' round-form, holding each player's marginal variance fixed and matching the realized teammate correlation (≈0.41 of round-form). Only the *joint tail dependence* changes. We read the effect on the 'run the table' tail — P(a team wins all 5 matches) — which feeds the shootout-advance probability. Flight 9 (Wood + Estes — clear-favorite flight) — favorite Wood + Estes : Dependence model Favorite P(win flight) Favorite P(win all 5) Additive shock (production) 42.1% 38.55% Gaussian copula 42.2% 38.56% t-copula (df=4) 42.5% 38.94% Flight 1 (Vola + Kerns — coin-flip flight) — favorite Vola + Kerns : Dependence model Favorite P(win flight) Favorite P(win all 5) Additive shock (production) 23.3% 19.28% Gaussian copula 23.5% 19.55% t-copula (df=4) 23.2% 19.23% Reading the result. Two findings. First, the Gaussian copula reproduces the additive model essentially exactly — the additive shared-shock construction *is* a Gaussian dependence, so this is a clean internal-consistency check that our copula machinery and the production model agree where they must. Second, switching to a t-copula (df=4) — same correlation, heavier *joint tails* so teammates boom or bust together more often — moves the run-the-table tail only modestly (a few tenths of a point on the favorite, direction depending on the flight) and leaves the flight-win probability essentially unchanged. Implication: for THIS field and format, the choice between Gaussian and t dependence is a genuinely second-order effect — the production additive structure is defensible. The sensitivity is real but small because (a) the 9-hole better-ball already injects large idiosyncratic variance that swamps the teammate tail-dependence, and (b) winning all 5 matches is dominated by *level* (how good the team is), not by the fine structure of how its two players' off-days co-move. The right takeaway is the *method*: when a tail probability (run-the-table, shootout-advance) drives real money, stress-test it under a t-copula rather than assuming the additive Gaussian is exact — here that test passes. 7. Two-level uncertainty propagation (headline teams) Single-level MC reports a P(win) as if the player distributions were known exactly. They are not: each player's mean/SD is estimated from a finite, recency-weighted sample (effective n). We run a two-level (nested / posterior-predictive) MC — outer loop resamples every player's (mean, SD) from its sampling distribution (mean SE = sd/√n eff, SD SE = sd/√(2·n eff)); inner loop runs the flight with common random numbers so the band reflects *parameter* uncertainty, not MC noise. This converts each headline P(win) into a full uncertainty band. Team Flight Point P(win) Param-uncertainty band (5–95%) Param SD MC-only SD @15k Param/MC ratio Wood + Estes 9 41.9% 34.6% – 46.3% 3.66 pts 0.41 pts 8.9× Vola + Kerns 1 23.5% 16.0% – 30.6% 4.60 pts 0.32 pts 14.2× Reading the result. Parameter uncertainty is an order of magnitude larger than Monte-Carlo noise at 15k sims. Wood + Estes is a genuine favorite, but its honest interval (34.6%–46.3%) is wide because the flight's outcome hinges on player scoring levels we only know to ±1–2 strokes. Vola + Kerns sits in a true coin-flip flight where the band (16.0%–30.6%) overlaps several rivals. Implication: publish P(win) with these bands , and stop spending sims to shrink an MC error that is already ~10–20× smaller than the irreducible parameter uncertainty. 8. Recommendation: the optimal simulation design Optimal simulation design for the 2026 Calcutta sim: 1. Sim count: 15k–20k as the production default (down from 60k). At 20k every P(win) is pinned to ±0.5 pt — far inside the parameter-uncertainty band (§7). Keep 40–60k only for the final cross-flight fair-value aggregation and the shootout-advance tails, where you want the last decimal smooth. 2. Sampler: QMC is optional, not a priority. Scrambled Sobol' gives a real 1.5–2.5× RMSE reduction at small budgets (≤8k) and is a zero-cost drop-in, so use it if you run small batches; but its slope flattens (N -0.37 vs pseudo N -0.53 ) and the gain washes out by ~16k. Do not treat QMC as a substitute for the variance-reduction stack. 3. Variance-reduction stack: antithetic variates + expected-points control variate + common random numbers across every scenario rerun (rho sweeps, re-pricing). Measured VRFs up to ~2.1× on clear-favorite flights (antithetic ~2.0× alone), smaller on coin-flip flights. CRN across scenario reruns is the single biggest win for the partner-correlation deltas the production report already publishes — adopt it there immediately. 4. Correlation model: keep the additive partner shock for the central estimate (it equals a Gaussian copula to within MC noise — verified in §6). Treat the t-copula as a stress test for tail/shootout EV, not a replacement; on this field the additive structure passes that test. 5. Always report parameter-uncertainty bands (two-level MC) alongside P(win). They are ~9–14× the MC noise and are the honest measure of what we know. 6. Sensitivity priority: Sobol' indices show the favorite's own player scoring level and consistency drive ~80–99% of its win-probability variance — far more than any correlation/noise constant. Among structural knobs, HOLE NOISE FRAC leads and RHO FIELD is inert. Spend calibration effort on the player distributions and the per-hole variance fraction, not on RHO FIELD . Assumptions & limitations - Inputs are the production inputs. We import value model.build player_dists read-only, so any bias in the player distributions (9→18 doubling, shrinkage target +2.6, SD floor 2.5) is inherited, not audited here. - Two-level MC uses a Gaussian sampling model for (mean, SD) with SEs from effective n; it does not capture model-form uncertainty (e.g. non-normal per-hole scores) or correlated estimation error across teammates who share rounds. - Sobol' priors are uniform boxes chosen around production values; indices are conditional on those ranges. Widening a range would raise that input's share. - The copula experiment matches the realized teammate correlation (~0.41 of round-form) and changes only tail dependence; it is illustrative of *sensitivity*, not a fitted dependence model (we did not estimate the empirical teammate copula from shared-round data). - QMC effective dimension is favorable here but integrand-specific; the Sobol' advantage shrinks for deep-tail quantities. All figures are for the favorite's P(win) on two representative flights, chosen to bracket a coin-flip flight and a clear-favorite flight. - Reproducibility: fixed master seed 20260609; NumPy PCG64 ( default rng ); scrambled Sobol'/Halton via scipy.stats.qmc . Full code in mc engine.py + experiments.py ; rerun with uv run python analysis/montecarlo_expert/experiments.py . Generated by analysis/montecarlo_expert/make_report.py from results.json . NumPy PCG64 · scipy.stats.qmc · compute 297.5s. No external dependencies. ============================================================================ # Model Corrections Source URL: https://calcutta.high.green/reports/corrections ============================================================================ 2026 Calcutta — Corrections & Reshape Report Generated by valuemodel.py. Summarizes the confirmed rules corrections and the $2,000 cap-anchor price reshape, then confirms which of our buy-card targets stay +edge._ 1. $2,000 cap-anchor price reshape (barbell) The $2,000 bid cap is a psychological magnet (round-number / cap anchoring). The 18 most-HYPED teams are modeled as racing to EXACTLY $2,000 (a fat spike); the rest of the field keeps a timid tail held below $1,500. Hype score = z(model price) + z(-team handicap) + a bonus for Jason targets. Total pool now $83,700 (target $82-$85k); spike = 18×$2,000 = $36,000. These are the cap-out teams — now OVERPRICED at $2,000, AVOID (their EV-based edge goes sharply negative): | Hype | Flight | Team | Hcp | P(win) | Est $ | Fair $ (EV) | Edge $ | Jason | |---:|---:|---|---:|---:|---:|---:|---:|:--:| | 3.77 | 2 | Hatcher + Berry | 8.4 | 25% | $2000 | $1894 | -106 | yes | | 3.57 | 2 | Downey + Shearer | 8.0 | 23% | $2000 | $1795 | -205 | yes | | 2.83 | 9 | Wood + Estes | 19.3 | 43% | $2000 | $1610 | -390 | | | 2.70 | 1 | Vola + Kerns | 5.0 | 23% | $2000 | $2439 | +439 | | | 2.30 | 1 | McHale + Bagley | 5.2 | 18% | $2000 | $2109 | +109 | | | 2.29 | 1 | Perry + Dye | 1.0 | 15% | $2000 | $1702 | -298 | | | 2.08 | 1 | Pilger + Bhatia | 5.1 | 17% | $2000 | $1853 | -147 | | | 2.08 | 1 | Terranova + Davis | 4.2 | 17% | $2000 | $1917 | -83 | | | 1.90 | 7 | Rekenthaler + Stafford | 15.4 | 15% | $2000 | $770 | -1230 | yes | | 1.78 | 1 | Beatty + Guiendon | 0.1 | 10% | $2000 | $1197 | -803 | | | 1.71 | 6 | Swiger + Dahlhauser | 13.5 | 25% | $2000 | $1469 | -531 | | | 1.69 | 3 | Burns + Preston | 8.5 | 19% | $2000 | $1733 | -267 | | | 1.62 | 3 | Embleau + Loewenthal | 9.1 | 18% | $2000 | $1641 | -359 | | | 1.56 | 3 | Brown + Watson | 9.1 | 17% | $2000 | $1543 | -457 | | | 1.54 | 3 | Patterson + Shirley | 10.3 | 17% | $2000 | $1594 | -406 | | | 1.50 | 6 | Knapp + Keister | 13.5 | 12% | $2000 | $830 | -1170 | yes | | 1.49 | 4 | Greenspan + Horne | 11.6 | 19% | $2000 | $1052 | -948 | | | 1.47 | 2 | Fann + Hartz | 7.9 | 15% | $2000 | $1286 | -714 | | 16/18 cap-out teams have NEGATIVE EV edge at $2,000 (over the cap-anchored price). Value concentrates in the mid-tier teams that do NOT cap out — see the +edge board below. Best value AFTER the reshape — non-capped teams with the largest +edge: | Flight | Team | Hcp | P(win) | Est $ | Fair $ (EV) | Edge $ | Jason | |---:|---|---:|---:|---:|---:|---:|:--:| | 6 | Vaniman + Hatz | 14.6 | 21% | $700 | $1315 | +615 | | | 3 | Martin + Torres | 9.6 | 16% | $900 | $1483 | +583 | | | 2 | Walsey + Walsey | 7.5 | 15% | $900 | $1295 | +395 | | | 7 | Marjoram + Peters | 15.4 | 23% | $700 | $1084 | +384 | | | 7 | Driggars + Richards | 15.4 | 20% | $600 | $977 | +377 | | | 6 | Kitchens + Taylor | 14.9 | 16% | $700 | $1055 | +355 | | | 3 | Brown + Brown | 9.0 | 12% | $800 | $1154 | +354 | | | 6 | Foresman + Benson | 14.5 | 15% | $700 | $1041 | +341 | | | 20 | Barnes + Smelcer | 58.1 | 33% | $200 | $456 | +256 | | | 4 | Green + Green | 11.5 | 19% | $800 | $1041 | +241 | | | 12 | Wright + Gadsby | 22.3 | 25% | $400 | $637 | +237 | | | 9 | Gallagher + Hansell | 19.6 | 15% | $500 | $727 | +227 | yes | | 7 | Thomas + Stricklin | 15.3 | 16% | $600 | $812 | +212 | | | 5 | Pearson + Ferguson | 12.6 | 21% | $700 | $910 | +210 | | | 10 | Fought + Fought | 20.6 | 24% | $500 | $697 | +197 | | Our buy-card (Jason) targets under the reshape | Flight | Team | Hcp | P(win) | Est $ | Fair $ (EV) | Edge $ | Cap-out? | Verdict | |---:|---|---:|---:|---:|---:|---:|:--:|:--| | 9 | Gallagher + Hansell | 19.6 | 15% | $500 | $727 | +227 | | BUY (+edge) | | 16 | Wilson + Slavis | 27.8 | 20% | $300 | $418 | +118 | | BUY (+edge) | | 10 | Gelinas + Chafin | 19.7 | 16% | $400 | $500 | +100 | | BUY (+edge) | | 20 | Ballard + Armstrong | 47.4 | 17% | $200 | $263 | +63 | | BUY (+edge) | | 13 | Panessa + Ratliff | 23.1 | 17% | $400 | $432 | +32 | | BUY (+edge) | | 16 | Wagner + Haswell | 28.1 | 20% | $400 | $413 | +13 | | BUY (+edge) | | 16 | Goodloe + Santivanez | 26.8 | 11% | $300 | $245 | -55 | | PASS (-edge) | | 11 | Perez + Marold | 21.5 | 15% | $500 | $441 | -59 | | PASS (-edge) | | 2 | Hatcher + Berry | 8.4 | 25% | $2000 | $1894 | -106 | CAP $2k | AVOID at $2k (capped/overpriced) | | 2 | Downey + Shearer | 8.0 | 23% | $2000 | $1795 | -205 | CAP $2k | AVOID at $2k (capped/overpriced) | | 8 | Barbaree + Williford | 17.0 | 12% | $700 | $493 | -207 | | PASS (-edge) | | 6 | Knapp + Keister | 13.5 | 12% | $2000 | $830 | -1170 | CAP $2k | AVOID at $2k (capped/overpriced) | | 7 | Rekenthaler + Stafford | 15.4 | 15% | $2000 | $770 | -1230 | CAP $2k | AVOID at $2k (capped/overpriced) | 6/13 Jason targets retain a positive EV edge under the reshape (i.e. are NOT priced to the cap). Capped targets are now traps at $2,000 — let the room overpay. 2. 10-stroke handicap-difference cap (off the LOW player) A 5 and a 17 now play as 5 and 15 — the high player is capped to 10 strokes over the low man in the foursome, applied in the sim's stroke allocation. This removes strokes from legit high-handicappers, so high-handicap 'bomber' teams LOSE win probability and the low-handicap teams they face GAIN. Teams whose P(win) moves most (cap ON minus cap OFF): | Flight | Team | Locked hcp | P(win) no-cap | P(win) 10-cap | Δ | |---:|---|---:|---:|---:|---:| | 20 | Barnes + Smelcer | 58.1 | 38.7% | 33.4% | -5.3% | | 20 | Ballard + Armstrong | 47.4 | 14.5% | 16.7% | +2.2% | | 20 | King + Scales | 38.6 | 5.2% | 6.3% | +1.2% | | 20 | Bell + Harren | 44.7 | 12.5% | 13.3% | +0.9% | | 12 | Dudley + Cohen | 21.6 | 12.2% | 13.0% | +0.8% | | 5 | Pearson + Ferguson | 12.6 | 21.3% | 20.7% | -0.6% | | 20 | Pilger + Hatz | 54.2 | 24.3% | 24.9% | +0.6% | | 11 | Richardson + Loricchio | 21.2 | 21.3% | 21.9% | +0.6% | | 12 | Benson + Brosnahan | 23.0 | 16.0% | 15.4% | -0.6% | | 10 | Gelinas + Chafin | 19.7 | 16.7% | 16.2% | -0.5% | | 20 | Morge + Callahan | 41.8 | 4.8% | 5.3% | +0.5% | | 12 | Beaver + Rhyne | 22.7 | 11.2% | 10.8% | -0.5% | | 5 | Levin + Leaf | 12.9 | 18.8% | 19.2% | +0.4% | | 8 | Yancey + Whaley | 16.5 | 16.0% | 15.6% | -0.4% | | 9 | Wood + Estes | 19.3 | 42.2% | 42.6% | +0.4% | | 4 | Powers + Merrigan | 11.4 | 14.3% | 14.0% | -0.4% | Directional read: the cap only bites foursomes whose handicap SPREAD exceeds 10 strokes, so it is concentrated — only 10/120 teams move >0.5pt and 3 move >1pt. The clearest case is the very-high-handicap flight 20: capping Barnes + Smelcer's strokes drops them -5.3% (58.1 locked handicap, the widest gap in the field), redistributing that win probability to the lower-handicap teams in their flight. Elsewhere most foursomes already sit inside a 10-stroke spread, so the cap is a no-op and the net pot/value impact is small. 3. Cap Patrol trend sign — CONFIRMED capmostImproved ≈ (index − last4): POSITIVE = recent scoring better than index = HOT = BUY. Verified against caphotIndex: corr = +0.96 (strongly positive), so a higher trend lines up with a higher hotness index. The enriched bid sheet exposes this as captrendmean (positive = hot). This matches the data-sign note in auction_intel.md and overrides the older 'negative = improving' framing. 4. Match-play aptitude proxy (hole-level GHIN data) Hole-level holedetails were USABLE: 215 of 240 field players had enough hole-by-hole data (≥27 holes). Match play rewards BIRDIES (win a hole outright) and grants BLOW-UP IMMUNITY (you can only lose a hole by one). So vs a smooth medal handicap we credit (a) birdie rate and (b) blow-up rate (match play caps the damage medal scoring punishes). matchplayaptitude (team = sum of both players' z-scored proxy) is POSITIVE for teams that should OUTPERFORM their handicap in match play, NEGATIVE for steady medal grinders who UNDERPERFORM. Top over/under-performers vs handicap: Better in match play than handicap implies (birdie-makers / streaky): | Flight | Team | Hcp | MP aptitude | P(win) | Est $ | |---:|---|---:|---:|---:|---:| | 1 | Beatty + Guiendon | 0.1 | +4.50 | 10% | $2000 | | 1 | Perry + Dye | 1.0 | +3.98 | 15% | $2000 | | 1 | Vola + Kerns | 5.0 | +3.70 | 23% | $2000 | | 1 | Terranova + Davis | 4.2 | +3.39 | 17% | $2000 | | 2 | Downey + Shearer | 8.0 | +2.51 | 23% | $2000 | | 10 | Schmeelk + Stein | 20.5 | +2.32 | 11% | $500 | | 1 | Pilger + Bhatia | 5.1 | +2.08 | 17% | $2000 | | 2 | Hurst + Jordan | 7.7 | +2.00 | 12% | $1000 | Worse in match play than handicap implies (steady medal grinders): | Flight | Team | Hcp | MP aptitude | P(win) | Est $ | |---:|---|---:|---:|---:|---:| | 18 | Hoard + Woods | 31.6 | -2.27 | 9% | $200 | | 13 | Vazquez + Prokupek | 23.2 | -1.85 | 13% | $400 | | 19 | Harte + Love | 35.7 | -1.83 | 17% | $200 | | 15 | Perry + Calobrisi | 26.8 | -1.76 | 20% | $400 | | 12 | Beaver + Rhyne | 22.7 | -1.70 | 11% | $500 | | 17 | Bowles + Troisi | 29.2 | -1.57 | 16% | $300 | | 19 | Truitt + Fetter | 36.0 | -1.44 | 11% | $200 | | 9 | Kattookaran + Levitas | 19.3 | -1.44 | 13% | $500 | This is a read-through signal (column matchplayaptitude in teamvalues.csv), NOT folded into fair value — it flags teams whose match-play upside is mispriced by a medal-based handicap. Assumptions stated - Cap-out set: top 18 by hype score pinned to exactly $2,000; this is a behavioral projection (no actual 2026 prices). Hype = strength + low-handicap pull + Jason-target bonus. - Timid tail scaled so the total pool lands ~$82-$85k; non-capped teams held below $1,500. - EV fairvalue (teamvalues.csv) is the full EV (flight prize + shootout) on raw sim probs; the enriched bid sheet's fairvalue is FLIGHT-ONLY on SHRUNK probs (0.6·sim + 0.4·1/6). Edges in this report use the EV fair value vs the reshaped estprice. - 10-stroke cap applied as whole strokes off the low player, max 10. - Match-play proxy uses raw hole score vs par, excluding x-holes; min 27 holes per player. ============================================================================ # What Predicts Performance Source URL: https://calcutta.high.green/reports/what_predicts ============================================================================ What Predicted Performance — and What Actually Stuck The single most decision-useful finding of the 2026 Calcutta project. Built from 2024 + 2025 auction prices joined to actual match-play results. All numbers re-derived from the real data. The question The handicap and the betting money are a prediction of how a team will do. The actual match-play points are the result. So: did the billing predict the score, and does beating your billing stick to anyone? (E.g., a team billed for 20 points that scores 25 "beat its billing" by 5.) 1. The billing barely predicts anything Within a flight (6 teams, ~150 points split among them, so the average team scores 25), the auction price explains only 3–5% of the variance in points (R² = 0.03 in 2024, 0.05 in 2025). Each full step up the price ladder is worth just +0.5 points. The real spread of finishes is ~16 to ~34. So "billed for 20, scored 25" is the norm, not the exception — the prediction is nearly powerless, and the residual (what you score over your billing) is almost the entire story. And that residual is mostly luck: net better-ball match play off the low handicap is designed to equalize ability. 2. At the team level, beating your billing does NOT stick Year-over-year correlation of team residuals: r = −0.23 — actually negative. The team that overperformed last year tends to regress the next. "They crushed it in 2025" is not a reason to buy them in 2026. This is why no team-based betting strategy survived the backtest. 2025's biggest over- and under-performers (price → predicted points → actual): | Team | Billed | Scored | Beat billing by | |---|---:|---:|---:| | Keister + Knapp | 24 | 32.5 | +8.3 | | Gelinas + Chafin | 25 | 33.0 | +8.3 | | Kulik + DiCicco | 24 | 32.0 | +7.8 | | Armstrong + Ballard | 27 | 34.5 | +7.2 | | Nodar + Heslep | 24 | 30.5 | +6.5 | | Pearson + Ferguson | 25 | 16.0 | −9.3 | | Mulick + Thilmany | 25 | 16.5 | −8.2 | 3. But a handful of INDIVIDUALS beat their billing every year — and they're our buy card Split teams into players and ask who outscored their billing in both 2024 and 2025. A short, sticky list falls out — and it overlaps almost perfectly with the teams we'd already flagged through completely different methods: | Player | 2024 (over billing) | 2025 (over billing) | On the buy card? | |---|---:|---:|---| | Estes | +6.7 | +6.2 | yes, Wood + Estes | | Keister | +4.9 | +8.3 | yes, Knapp + Keister | | Copeland | +4.1 | +4.2 | yes, Wright + Copeland | | Bachstein | +4.1 | +4.8 | (Hancock + Bachstein) | | Flammia | +4.1 | +6.2 | — | The market reprices teams every year (partners change, last year's hero reverts), but a few individuals quietly outscore their card season after season. 4. Why this matters: three independent methods, same names This residual test is purely statistical — it knows nothing about who plays golf. Yet the names it surfaces (Estes, Copeland, Keister) are the same players we independently flagged via: - Human intel — an eyewitness sandbagger read (Copeland), insider knowledge. - The model + prior results — Wood + Estes, Knapp + Keister winning flights. When an eyewitness, a simulation, and a blind 2-year residual test all point at the same handful of players, that's no longer a backtest artifact — it's signal. Summary - Money predicts the floor, not the ceiling. The cheapest third of teams finish dead last ~40% of the time, stable across both years (priciest third: ~19%). The market reliably spots duds; it's a coin flip on champions. - Don't bet on teams that overperformed — bet on the individuals who reliably outscore their billing. The edge is at the player level, not the team level. - The one-liner: The market prices the card. We price the player. Honest caveat Two years of price data and a short list — some names on it are partly luck (with ~120 players, a few will beat their billing twice by chance). What makes it credible is the cross-method convergence, not the residual test alone. Treat it as a strong lean, not a certainty. ============================================================================ # Auction Strategy Source URL: https://calcutta.high.green/reports/strategy ============================================================================ Calcutta Auction Strategy — Synthesized Findings Research synthesis for a live, adaptive auction-assistant tool. Tailored to our format (README.md): ~120 teams in 20 flights of 6, member-member golf calcutta, **random auction order across flights, $200 start / $100 increments / $2,000 cap per team**, each team has first right of refusal to buy itself at the cap, buy up to half your own team allowed. Payouts: - 10% of the TOTAL pot → shootout: overall winner 50% / runner-up 30% / shootout finalist 20%. - 90% of EACH FLIGHT'S OWN money stays in that flight: flight winner 70% / runner-up 30%. This payout split is the single most important structural fact and drives most of the tailored conclusions below. Each section gives Takeaway → Reasoning → Implication for the live tool. --- 0. The format's defining structural quirk: a per-flight pot plus a global tax Takeaway. Money is not one big pool. 90% of a flight's dollars are redistributed only within that flight (70/30 to that flight's top two), and 10% of every dollar everywhere is skimmed into a global shootout pool (50/30/20). So a team's value has **two independent components**: - Flight component (local): P(win your flight)×0.70 + P(2nd in flight)×0.30, paid out of 0.90 × your-flight's-own pot. Depends only on the 6 teams in your flight and how much money your flight attracts. - Shootout component (global): P(reach shootout)×[0.50·P(win|in) + 0.30·P(2nd|in) + 0.20·P(3rd|in)] paid out of 0.10 × the GRAND total pot across all 20 flights. Reasoning. Generic calcutta advice assumes one shared pool where "growing the pot helps everyone who owns a team" ([unabated][u], [oddsjam][oj]). Here that's only ~10% true. Driving up prices in other flights barely helps you — it only feeds the 10% shootout slice. Driving up prices in your own flight directly grows the 90% local pot the flight-winner collects. Implication for the live tool. Maintain 21 running pot totals: one per flight, plus a grand total. Value every team as EV = flightEV(localpotf) + shootoutEV(0.10·grandtotal)`. When the user owns a team, the marginal value of bidding up another team in the same flight is materially positive (grows localpotf); bidding up a team in a different flight only grows the shootout slice (×0.10) and should almost never be done deliberately. [u]: https://unabated.com/articles/calcutta-betting-auction-guide [oj]: https://oddsjam.com/betting-education/calcutta-auction --- 1. Calcutta fundamentals & how pools/payouts shape value Takeaway. A calcutta is an open ascending (English) auction where bidders "own" entrants and share a prize pool funded entirely by the auction proceeds; payout structure dictates which finishes you're actually buying ([liveabout][la], [wikipedia][wk], [actionnetwork][an]). Reasoning. Because the pool is the sum of bids, you are buying a **claim on a fraction of a pot you and your rivals are simultaneously creating.** With a flighted + shootout structure, "winning" isn't binary — there are five distinct payout events (flight 1st, flight 2nd, shootout 1st/2nd/3rd), each a different probability and dollar slice. Implication for the live tool. Encode the payout structure as data. For each team compute and display the five component probabilities and their dollar contributions so the user sees why a team is worth what it is (e.g., "Team X: $310 flight-win, $90 flight-2nd, $140 shootout-equity"). [la]: https://www.liveabout.com/what-is-a-calcutta-in-golf-1564030 [wk]: https://en.wikipedia.org/wiki/Calcutta_auction [an]: https://www.actionnetwork.com/golf/calcutta-pool-auction-rules-payouts --- 2. Valuation — fair price = EV, and the pot is ENDOGENOUS Takeaway. Fair price = expected payout = Σ P(payout event) × dollar payout for that event. The universally repeated formula is EV = P(finish) × payout% × totalpot` ([sportsbookreview][sbr], [bettoredge][be], [oddsjam][oj]). Worked example: 12.5% chance × 15% payout slice × $1,000 pot = $18.75 fair price ([sbr][sbr]). The hard part — the pot is endogenous. "You don't know how much the pot will be in the end, which is why sophisticated bidders track the pot as it grows throughout the auction and adjust their valuations in real time" ([oddsjam][oj]). "The total pot grows and evolves as the auction goes on" ([bettoredge][be]). The very first sale "sets the relative value for every other asset auctioned after it" — if an asset goes for X and you think it's p% of the pool, implied total pool is X/p, cascading into every later bid ([unabated][u]). Reasoning. A fixed-point problem: each team's fair price depends on the total pot, but the pot is the sum of all fair prices. Early the pot is maximally uncertain (widest error bars); late it's nearly known. This asymmetry is the core reason late decisions are easier and often cheaper. For our format the endogeneity is two-layered: value depends on (a) how much the team's **own flight attracts (90% of value) and (b) the grand total** (10% shootout slice) — both unknown early, resolving at different rates under random draw order. Handicaps/seedings → win probability. Use net-score simulation: ~10k-round sims give, e.g., a 9-handicap beating a 14-handicap ~67% of the time ([golf.com sim][gs]). In a 6-team net flight, convert each team's expected net score + round-to-round SD into P(best net) and P(2nd) via Monte Carlo; chain to the shootout. Where you have a market-style line, remove the vig: divide each implied probability by the sum of all implied probabilities so they total 100% ([unabated][u]). Implication for the live tool. - Keep a live two-layer pot estimator: per-flight projected final pot and grand projected final pot. Seed from priors; after each sale blend prior with realized (Bayesian shrink toward realized as more of a flight sells). Re-price every unsold team after every sale. - Represent each team's probabilities from a handicap→net-score Monte Carlo, vig-normalized so flight probabilities sum to 1. - Display fair price with an uncertainty band (early = wide, late = tight), not a point. - Compute fair price against the projected final pot, not the current partial pot — else early teams look far too cheap. [sbr]: https://www.sportsbookreview.com/picks/ncaa-basketball/march-madness-calcutta-auction-tips-strategy/ [be]: https://www.bettoredge.com/post/calcutta-auction-explained [gs]: https://golf.com/instruction/odds-win-next-golf-match-simulator/ --- 3. "Buy your own team" / first right of refusal — the price ceiling & the half-hedge Takeaway. The owner's right to buy at the $2,000 cap, plus the agreement to buy up to half from the winning owner, anchors and caps prices and changes optimal bidding. Reasoning. - First-right-of-refusal at the $2,000 cap = a hard ceiling. No team costs more than $2,000, and a team worth near/above $2,000 will simply be self-bought. The most valuable teams (high flight-win + shootout equity) are effectively removed from the open market. Exploitable value lives in the middle tiers — consistent with "value is in the middle of the pack" ([liveabout][la], [bettoredge][be]). - The half-buyback is a hedge and an anchor. "You can buy back some or all of your own team" ([theleftrough][tlr], [livetourney][lt]). Example: team sells $400, half-buyback $200; win $2,000 → split $1,000/$1,000 ([theleftrough][tlr]). For the owner-bidder, buying half is +EV whenever half the clearing price < half the EV — lower-variance skin in the game. For an outside winner, the buyback right means you may keep only ~half the upside on a team you "won," reducing effective EV on won teams. - Game theory of defense. A rational owner of a strong team bids up to min($2,000, private EV). Rivals know this, so fighting an owner to the cap rarely pays — you'd pay full freight against someone with information and loyalty motives. Let the owner have it; hunt elsewhere. Implication for the live tool. - Cap-aware valuation: clamp every fair price at $2,000; flag teams whose uncapped EV exceeds ~$1,700 as "owner will likely self-buy at cap — deprioritize." - Buyback-adjusted EV for won teams: discount post-win upside by the expected buyback share (default: strong player reclaims ~50%, weak ~0%). Surface "effective EV after expected buyback." - Owner-mode: if the user owns the team on the block, recommend buy-half when (clearing_price/2) < (EV/2); full self-buy only when EV ≥ ~$2,000. - Treat the $2,000 cap and $200/$100 grid as hard constraints on proposed increments. [tlr]: https://theleftrough.com/calcutta-golf/ [lt]: https://www.livetourney.com/blog/what-is-a-calcutta-in-golf --- 4. Budget management across a sequential, random-order auction Takeaway. With random order you never know what's coming, so pace spend against a **projected final pot and a target portfolio share**, not against teams currently on the block. Reserve dry powder for late bargains. Reasoning & transferable theory. - Inflation factor (from auction-draft theory) = `(remaining dollars / remaining unsold value) − 1` ([fangraphs][fg]). Dollars outrun value → prices inflate, be patient; value outruns dollars (rooms tapped out) → bargains, pounce. - Pacing heuristics: "save 20% of budget for the final 50% of nominations"; don't blow >40–50% on one asset; three-bucket caps ~15/12/8% of budget per tier ([fantasypros][fp], [bettoredge][be], [sportsbookreview][sbr]). - Dollars-left-in-the-room is a tradeable signal. A near-tapped rival can't contest the next team → its clearing price falls. The documented declining-price / "afternoon effect" in sequential auctions is driven precisely by budget-constrained bidders running dry ([afternoon effect][ae], [budget-constrained sequential][bc]). - Overspend-early vs. late-bargain. Early FOMO inflates prices; "as participants run out of money or lose interest late, you might snag a [team] at a bargain" ([bettoredge][be], [unabated][u]). With random order an undervalued team can appear anytime — hold reserve to exploit it whenever it lands. Implication for the live tool. - Track per-bidder remaining budget (infer from spend if not public); maintain room-wide remainingdollars and remainingunsoldvalue; surface the live inflation factor**. - Convert bankroll into a target portfolio and a pacing curve (recommended cumulative spend vs. fraction-elapsed) with an explicit reserve (default ≥20–25%) for the late phase. - After each sale recompute "how many remaining target buys can I still afford at projected prices?" and warn if on pace to be priced out or to finish with idle cash. - Flag when key rivals are tapped → "buying power falling; upcoming teams should clear cheaper." [fg]: https://fantasy.fangraphs.com/how-to-account-for-keeper-inflation-in-your-auction-draft/ [fp]: https://www.fantasypros.com/2025/08/fantasy-football-auction-draft-advice-spending-strategy/ [ae]: https://www.researchgate.net/publication/248960050WineauctionsMoreexplanationsforthedecliningprice_anomaly [bc]: https://arxiv.org/pdf/1209.1698 --- 5. LATE-GAME / ADAPTIVE tactics (the core of the tool) Takeaway. Optimal bids shift as information is revealed. Early = bid under wide pot uncertainty and shade down for the winner's curse; late = pot and relative values are nearly known, rival budgets are visible, undervalued teams appear when the room is tapped — pounce. Reasoning — what changes over the auction ([unabated][u], [bettoredge][be]): | Phase | What's known | What to do | |---|---|---| | Early | Pot maximally uncertain; first sales set the implied pool. | Shade bids down (winner's curse + pot error). Don't overpay to "anchor." Don't let one buyer run away with the best teams, but don't fight an owner to the cap. | | Middle | Pot estimate sharpening; you see who's spending. | Overbidding here is recoverable. Drain rivals on teams you don't want. In our format, only bid up teams in your own flights to grow the 90% local pot. | | Late | Pot ~known; relative values certain; budgets visible. | Exploit auction fatigue (owners sated) and tapped-out rivals. "Late... the relative value is more certain... take advantage of both auction fatigue and FOMO" → step up on deflated bidding ([unabated][u]). | - The pot resolves → EV error bands collapse. Late bids can be made much closer to true EV (denominator fixed). The tool should tighten its fair-price band as the auction progresses and let the user bid more aggressively toward EV late, more conservatively early. - Declining-price effect is real and exploitable ([afternoon effect][ae], [budget-constrained sequential][bc]): equivalent-quality teams drawn later clear cheaper. So the same team is worth bidding up to a higher fraction of its EV late than early — and you should bank reserve precisely to capture this. - Flight-completion dynamics (format-specific). A flight's local pot is only fully known once all 6 teams sell. A team from a flight mostly sold → local pot nearly known → price tightly. A team from a flight where few have sold → uncertain local pot → wider band, more caution, but a chance to set the anchor. Track per-flight % sold and weight confidence accordingly. Implication for the live tool — make these the live recommendations: 1. After every sale, update (a) that flight's pot + grand pot, (b) selling team's realized-vs- expected price (calibrate the model), (c) rival remaining budgets, (d) the inflation factor. 2. Re-rank all unsold teams by value-over-projected-price (EV − expected clearing price); show top "best remaining values." 3. Set a walk-away (max) price per unsold team = min($2,000 cap, buyback-adjusted EV × confidence-scaled aggression). Aggression rises late (→1.0 of EV), falls early (~0.8) to absorb pot uncertainty and the winner's curse. 4. Bargain flag (POUNCE): current_bid < EV lowerband AND user has reserve AND likely contesters tapped → recommend bidding up to walk-away. 5. Reserve guard: never spend reserve earmarked for higher-value teams still expected to appear (estimate from the unsold pool). --- 6. Market efficiency & behavioral biases to exploit Takeaway. Calcuttas are inefficient in predictable ways; bet against the crowd's biases. - Favorite–longshot bias. Longshots are systematically overbet, favorites underbet; a 1/1 favorite returns ~85¢/$ vs ~63¢/$ for a 30/1 longshot ([wiki FLB][flb], [NBER][nber]). Driven by risk-love and overweighting small probabilities. **In a calcutta the cheap "Cinderella" longshots are usually the worst value; solid favorites/strong mids are the best — opposite of the popular "hunt the upset" advice. (The cap removes the very top favorites, so the sweet spot is strong-but-not-cap-level teams.) - Emotional / loyalty overbidding. "Popular players attract emotional bidding — the club champion, last year's winner, the player everyone likes... pushes prices above rational levels" ([sportsbettingdime][sbd], [pinnacle][pin]). Members overpay for friends and their own team. Fade. - Anchoring. The first sale anchors the room's sense of the pool ([unabated][u]); one aggressive opener inflates everything. Resist repricing the whole board off one sale — blend with priors. - Random order vs. flight-by-flight. Random order spreads a flight's teams across the auction, so the local pot resolves gradually/unpredictably — more early uncertainty but late "completion-certainty" bargains. Per-flight tracking is therefore essential. Implication for the live tool. Bake in a bias-correction layer: down-weight crowd-implied probabilities for longshots, up-weight favorites toward the model's net-score probabilities; detect likely emotional/loyalty premiums (own-team, prior winner, popular member) and label "likely to clear above EV — let it go." Keep anchoring in check by shrinking model updates from any single sale. [flb]: https://en.wikipedia.org/wiki/Favourite-longshot_bias [nber]: https://www.nber.org/system/files/working_papers/w15923/w15923.pdf [sbd]: https://www.sportsbettingdime.com/guides/articles/calcutta-auctions/ [pin]: https://www.pinnacleoddsdropper.com/blog/calcutta-betting --- 7. Auction-theory & parimutuel parallels worth borrowing - Winner's curse (common-value auctions). Bidders "bid more than rational agent theory prescribes" because the winner most overestimated value; remedy = "bid as if you knew your bid would win," i.e., shade your estimate downward ([Kagel & Levin][kl], [winner's curse wiki][wc], [Yale ECON 159][yale]). A team's pot-fraction value has a large common-value component (everyone estimates the same pot and field), so the curse applies. Open ascending format mitigates it (you see rivals' bids) but doesn't remove the pot-uncertainty curse. - Parimutuel pricing. Payouts come from pooled stakes net of takeout; the favorite-longshot bias transfers directly. With zero takeout the bias vanishes — a calcutta has near-zero takeout on the pool (it's redistributed), so the bias here is purely behavioral, hence exploitable ([wiki FLB][flb]). - Sealed vs. open ascending. Open ascending (English) yields outcomes closer to equilibrium and reduces the curse vs sealed bids ([Kagel & Levin][kl]). Ours is open ascending — good, the tool can read real-time signals (who's bidding, who's stopping). - Sequential / declining-price models. Budget-constrained sequential-auction models predict the afternoon effect and justify "hold reserve, buy late" ([budget-constrained sequential][bc], [afternoon effect][ae]). - Published calcutta quant tooling. Practitioner tools do "true percentage-of-pot modeling," live model-value vs. purchase-price comparison, and Monte-Carlo advancement odds with a tracked projected pot ([BettorEdge calculator][bec], [PoolGenius tools][pg]) — confirming the exact compute stack this tool should implement. [kl]: https://www.asc.ohio-state.edu/kagel.4/CVsurvey.short.PDF [wc]: https://en.wikipedia.org/wiki/Winner's_curse [yale]: https://oyc.yale.edu/economics/econ-159/lecture-24 [bec]: https://start.bettoredge.com/tools/calcutta-value-calculator [pg]: https://poolgenius.teamrankings.com/ncaa-calcutta-auction-tools/ --- 8. Sandbagging / handicap-integrity adjustment Takeaway. Inflated handicaps distort net-score win probabilities; buyers must adjust priors for likely sandbaggers ([golf.com][gsb], [golfdigest][gd]). Reasoning. "The format was tailor-made for the modern sandbagger: keep your handicap comfortably inflated... then 'discover' your swing when money is on the line" ([golf.com][gsb]). A team whose posted handicap overstates true ability is underpriced by a naive net-score model — real P(win flight) is higher than the handicap implies. Safeguards (verified current handicaps, capping partner stroke differential to ~5–8) reduce but don't eliminate it ([livetourney][lt]). Implication for the live tool. Allow a per-team "sandbag adjustment" (manual flag or data-driven: recent net scores far better than handicap; handicap rising pre-event). Apply as a downward shift to expected net score (raising win probability and EV). Surface "model says cheap, flagged likely sandbagger — true value higher" so the user exploits teams the room under-prices on stale handicaps. [gsb]: https://golf.com/lifestyle/sandbagging-golf-calcutta-cheating-rules-language/ [gd]: https://www.golfdigest.com/story/the-sandbagging-scandal-that-shook-golf --- 9. Live / Adaptive Algorithm Sketch (concrete rules for the assistant) State maintained (updated after every sale): - potflight[f] for f=1..20 and potgrand — realized $ so far. - pctsold[f]` — fraction of each flight's 6 teams sold. - projpotflight[f], projpotgrand — Bayesian blend of prior and realized, shrinking toward realized as pctsold` rises. - budgetremaining[bidder]; roomremaining = Σ. - unsoldvalue = Σ EV(team) over unsold; inflation = roomremaining/unsoldvalue − 1`. - userspent, userreserve (≥20–25% until late), usertargetportfolio. Per-team valuation (recomputed each sale): ``` Pwinf, P2ndf = MonteCarlo(net handicaps in flight f) # vig-normalized to sum to 1 Pshoot, Ps1, Ps2, Ps3 = chain(P_winf, shootout model) EVlocal = (Pwinf0.70 + P2ndf0.30) 0.90 projpot_flight[f] EVshoot = (Ps10.50 + Ps20.30 + Ps30.20) 0.10 projpot_grand EVraw = EVlocal + EV_shoot EVeff = EVraw - expectedbuybackshare postwin_upside # for teams user would own EVcapped = min(EVeff, 2000) # first-right-of-refusal ceiling band = EVcapped uncertainty(pctsold[f], fracauction_elapsed) # wide early, tight late ``` Confidence-scaled aggression (max bid): ``` aggression = lerp(0.80, 1.00, fracauctionelapsed) # shade for winner's curse early walkaway = min(2000, EVcapped aggression) if contestingrivalstapped: aggression = 1.05 (cap at EVcapped) # declining-price effect ``` Decision rules surfaced live: 1. Re-rank unsold teams by EVcapped − expectedclearingprice`; show top values. 2. Bargain flag (POUNCE): currentbid < EVcapped lowerband AND user_reserve sufficient AND likely contesters tapped/sated → recommend bidding up to walkaway. 3. Pass flag: uncapped EV ≥ ~$1,700 (owner self-buys at cap) OR clear emotional/loyalty/longshot premium → "let it go." 4. Own-flight lever: if user owns a team in flight f, note bidding up other teams in f grows potflight[f]` (90% local) and is mildly +EV; other flights only feed the 10% shootout slice — discourage. 5. Owner mode: if user owns the team on the block → buy-half when clearingprice/2 < EVcapped/2; full self-buy only if EVcapped ≥ ~2000`. 6. Pacing guard: project remaining target buys × expected clearing prices; warn if on pace to be priced out (overspending early) or to end with idle cash (too passive). Enforce reserve until late phase, then release it. 7. Sandbag overlay: apply handicap-integrity adjustments before valuation; surface "cheap per market, higher true value." 8. Anchor damping: update projpot` modestly from any single sale (shrinkage), strongly only as pctsold` accumulates — don't let one aggressive opener reprice the board. One-line philosophy: *Early — bid below EV under wide pot uncertainty and never chase favorites to the cap. Late — the pot is known, rivals are tapped, fatigue deflates bids; hold reserve, then strike near full EV on the best remaining values.* --- Sources - Unabated — A Guide To Calcutta Betting Auction Strategy: https://unabated.com/articles/calcutta-betting-auction-guide - BettorEdge — Calcutta Auction Guide: https://www.bettoredge.com/post/calcutta-auction-explained - BettorEdge — Calcutta Value Calculator: https://start.bettoredge.com/tools/calcutta-value-calculator - SportsbookReview — March Madness Calcutta tips/strategy: https://www.sportsbookreview.com/picks/ncaa-basketball/march-madness-calcutta-auction-tips-strategy/ - OddsJam — Calcutta Auction tips: https://oddsjam.com/betting-education/calcutta-auction - PoolGenius — NCAA Calcutta Auction Tools: https://poolgenius.teamrankings.com/ncaa-calcutta-auction-tools/ - Action Network — Golf Calcutta pool rules/payouts: https://www.actionnetwork.com/golf/calcutta-pool-auction-rules-payouts - LiveAbout — How a Calcutta works at golf tournaments: https://www.liveabout.com/what-is-a-calcutta-in-golf-1564030 - Wikipedia — Calcutta auction: https://en.wikipedia.org/wiki/Calcutta_auction - The Left Rough — What is a Golf Calcutta (buyback): https://theleftrough.com/calcutta-golf/ - LiveTourney — What is a Calcutta in Golf: https://www.livetourney.com/blog/what-is-a-calcutta-in-golf - Golfible — Calcutta Golf: https://golfible.com/calcutta-golf/ - SettleUp Golf — Calcutta tournament & payout calculator: https://settleup-golf.com/learn/calcutta-golf-tournament - Sports Betting Dime — Calcutta Auctions (behavioral): https://www.sportsbettingdime.com/guides/articles/calcutta-auctions/ - Pinnacle Odds Dropper — Calcutta Betting tips: https://www.pinnacleoddsdropper.com/blog/calcutta-betting - Golf.com — Sandbagging in golf/calcutta: https://golf.com/lifestyle/sandbagging-golf-calcutta-cheating-rules-language/ - Golf Digest — The sandbagging scandal that shook golf: https://www.golfdigest.com/story/the-sandbagging-scandal-that-shook-golf - Golf.com — Handicap match-win simulator: https://golf.com/instruction/odds-win-next-golf-match-simulator/ - Wikipedia — Favourite-longshot bias: https://en.wikipedia.org/wiki/Favourite-longshot_bias - NBER w15923 — Explaining the Favorite-Longshot Bias: https://www.nber.org/system/files/working_papers/w15923/w15923.pdf - Wharton — The Favorite-Longshot Midas: https://jacobslevycenter.wharton.upenn.edu/wp-content/uploads/2018/08/The-Favorite-Longshot-Midas.pdf - Kagel & Levin — Common Value Auctions and the Winner's Curse (survey): https://www.asc.ohio-state.edu/kagel.4/CVsurvey.short.PDF - Wikipedia — Winner's curse: https://en.wikipedia.org/wiki/Winner's_curse - Yale Open Courses ECON 159 L24 — Auctions & Winner's Curse: https://oyc.yale.edu/economics/econ-159/lecture-24 - ResearchGate — Wine auctions / declining-price (afternoon effect): https://www.researchgate.net/publication/248960050WineauctionsMoreexplanationsforthedecliningprice_anomaly - arXiv 1209.1698 — Sequential Auctions with Budget-Constrained Bidders: https://arxiv.org/pdf/1209.1698 - FanGraphs — Auction draft keeper inflation (inflation factor): https://fantasy.fangraphs.com/how-to-account-for-keeper-inflation-in-your-auction-draft/ - FantasyPros — Auction draft spending strategy: https://www.fantasypros.com/2025/08/fantasy-football-auction-draft-advice-spending-strategy/ - The Fantasy Footballers — Strategic nominations (nom-ahead/behind data): https://www.thefantasyfootballers.com/analysis/fantasy-football-auction-drafts-the-power-of-strategic-nominations-fantasy-football/ ============================================================================ # Cap Patrol & Psychology Source URL: https://calcutta.high.green/reports/cappatrol ============================================================================ Cap Patrol Golfer Metrics — Psychological Read & Buyer's Weighting A decision-oriented interpretation of Cap Patrol's per-player metrics for our 2026 member-member net better-ball match-play Calcutta. Goal: turn each metric into a single answer to the buyer's only real question — *"will this team beat or miss the net handicap our model assumes, in a money pressure match?"* — and roll it into one true-ability-vs-index adjustment the bidder reads next to the model's fair value. Data: cappatrol-re/fieldcappatrol.csv` (228 of 240 field players matched; 12 off-club guests blank). All ranges/correlations below are computed on those 228. --- TL;DR for the bidder - Two real signals, not seven. After de-duplicating, the file carries only three semi-independent axes, and two of them are the same thing: 1. FORM — caphotIndex and capmostImproved (trend) are the same signal (Pearson +0.96). hotIndex is just the field-percentile of the trend. Use one. 2. CLUTCH/ABILITY — capclutch and capability are literally the same column (capability is capclutch rounded to one decimal; max abs diff 0.05). Count once. 3. YTD drift — capytd` is a weaker, longer-window cousin of FORM. Minor tie-breaker. - capindex/capytd/caplast4` are not signals, they're the raw differentials the above are built from (trend ≈ capindex − caplast4, corr +0.95). - SIGN WARNING — the trend sign in our data is the OPPOSITE of the brief. The task said "negative capmostImproved = improving." In this* file it's reversed: the hottest players (hotIndex 98–99) have positive trend (+9 to +12) and the coldest (hotIndex 4–17) have negative trend (−6 to −14). Mechanically trend ≈ index − last4, so **positive = recent scoring better than index = playing hot/improving = BUY direction.** Every weight below uses the data's sign. If you wire this into the model, confirm the sign against hotIndex before trusting it. - Sandbagging via "selective posting" is a non-signal in this field. capturnedIn` is 100 for 225 of 228 players. Only 3 are below 100. There is no posting-behavior edge to mine here — chase the FORM + CLUTCH combination instead (that's where the hidden ability is). - The one composite (strokes the team beats its net index by): per player adj = 0.35·(hotIndex centered, scaled to ±2 strokes) + 0.35·trend + 0.25·clutch − 0.05·ytd, team = mean of the two partners. Read it as "expected strokes better (+) or worse (−) than the handicap our fair-value model already priced." Positive and cheap = BUY. --- Format lens: what actually wins money here Net better-ball match-play, handicaps off the low player, money via Calcutta (docs/auctionrules2026.md; auction psychology in research/FINDINGS.md). Three things change how we weight these metrics: - Net, off the low man. The better ball counts and strokes come off the low player. A team's ceiling is driven by its stronger partner having more good holes than expected. That makes upside form/ability of at least one partner worth more than steady mediocrity — match-play rewards birdies and blow-up immunity, not smooth medal scoring. - Pressure is the whole event. Money + match-play + a crowd is exactly the "pressure" condition the clutch literature studies. A metric that predicts behavior under pressure (CLUTCH) is more decision-relevant than one that predicts average scoring (FORM/index). - The market misprices on stale handicaps. Per research/FINDINGS.md §8, a naive net-score model underprices players whose posted index lags true ability. FORM + CLUTCH is our data-driven version of that sandbag overlay: it says "this index is stale low-ability or stale high-ability" before the room notices. --- Metric-by-metric 1. `cap_hotIndex` — FORM (momentum percentile) · range 4.1–99.6 - What it measures. Field-percentile of how much better a player is scoring right now vs. their own index. It is the percentile expression of capmostImproved` (corr +0.96) and is ~ −0.94 correlated with last4 − index. High = recent rounds beating the index. - Psychology. This is the "hot hand." Players, members, and the room all over-weight recent streaks. The honest read of the research is mixed: classic work calls the hot hand a fallacy driven by misreading small samples; the Miller–Sanjurjo correction and some baseball/basketball panel studies find a real but small effect. Net: a streak is partly signal (fitness, swing change, confidence) and mostly **due to regress toward the player's true index.** - Reliability / noise. Noisiest of the three axes. WHS computes the index from the best 8 of the last 20 differentials, so a hot streak is already baked into a falling index — meaning a chunk of "hotness" is mean-reversion waiting to happen, and 7-stroke-better rounds trigger Exceptional Score Reduction, dropping the index automatically. So a very high hotIndex on a stable index is the interesting case; a very high hotIndex on a fast-falling index is half-priced-in already. - How a buyer uses it. Treat as momentum that will partially regress — a tie-breaker, not a thesis. High hotIndex nudges "team beats its index" up, but shade it: weight it less than clutch and don't pay a premium on hotIndex alone. The real value is **high hotIndex + positive clutch + index that hasn't fully caught up** = genuinely under-handicapped right now. 2. `cap_mostImproved` / trend — FORM (raw strokes) · range −14.2 to +11.9 - What it measures. Essentially capindex − caplast4 (corr +0.95): how many strokes better the last-4 differential is than the carried index. Positive = improving / hot (data sign — see SIGN WARNING). It is caphotIndex` in raw-stroke units, not a new axis. - Psychology / reliability. Same hot-hand caveats as hotIndex. Its one advantage: it's in strokes, so it drops straight into a net adjustment without rescaling. Its disadvantage: a 4-round window is tiny and regresses hard; −14 and +12 tails are mostly small-sample noise plus reversion, not durable skill. - How a buyer uses it. Use it as the stroke-denominated form term, but cap its influence and don't double-count it with hotIndex (they're the same signal). Persistent positive trend across both partners is a mild BUY; a single partner's extreme tail is likely to revert and shouldn't move your bid much. 3. `cap_clutch` = `cap_ability` — CLUTCH / true-ability · range −19 to +33 - What it measures. Performance-under-pressure / true-ability signal; **positive = outperforms in competition vs. what the index predicts. Confirmed: capability is the same column rounded (max diff 0.05) — count it ONCE. It is independent of FORM (corr with hotIndex −0.08), which is exactly why it's valuable: it adds information the momentum signal doesn't. - Psychology — the most decision-relevant axis. The clutch/choke literature says the same high-pressure situation produces clutch for some and choking for others, and the split is a stable trait (appraisal of pressure, reliance on implicit vs. explicit knowledge, "paralysis by analysis"). That stability is what we want: unlike a hot streak, a player who shows up in competition tends to keep doing it. In a money match-play event, this is the signal that most directly predicts over/under-performance vs. handicap. - Reliability / noise. More trait-like and less reversion-prone than FORM, but the tails (±20–33) sit on modest sample sizes — treat the sign and rough magnitude as real, discount the exact number. It is also the axis most aligned with the sandbagging thesis: a positive clutch on an unremarkable index = a player who is better than his number when it counts = the room's net model underprices the team. - How a buyer uses it. Highest-conviction "beats handicap under pressure" term. Positive clutch on either partner (especially the low/strong partner, who drives the better-ball ceiling) is a genuine BUY edge the naive net model misses. Strongly negative clutch is a real FADE — a likely choker in exactly our conditions — even if FORM looks fine. 4. `cap_turnedIn` — posting completeness · range 11–100 (225/228 = 100) - What it measures. % of rounds actually posted. The classic **selective-posting sandbag tell**: leave the good "away" rounds unposted to keep the index inflated. - In THIS field it's a dead signal. 225 of 228 are at 100; only 3 are below (Nick Beatty 11, Bill Atkins 74, Scott Leprohon 91). With Cap Patrol clearly pulling near- complete histories, the selective-posting lever the brief hoped to exploit basically doesn't exist in our data. - How a buyer uses it. Don't build it into the composite. Use it only as a **manual flag: if a low-turnedIn player also* has positive trend and positive clutch, that's the textbook hidden-sandbagger pattern (inflated index + secretly improving + clutch) and is a strong BUY — but check the raw round count, because at very low completeness (Beatty's 11) the other metrics are computed on almost no data and are unreliable. None of our 3 low-post players hit that triple, so there's no actionable selective-posting edge in this field. 5. `cap_ytd` — year-to-date drift · range −14.7 to +26.6 - What it measures. Longer-window differential-vs-index for the season. Here **positive = scoring worse than index YTD (corr with last4−index` +0.27, with hotIndex −0.26), i.e. the opposite-signed, slower cousin of FORM. - How a buyer uses it. A regression anchor, not a primary signal. If FORM is screaming hot but YTD says the player has been worse than index all year, the hot streak is more likely to revert — so YTD gets a small negative weight, just enough to deflate flash-in-the-pan tails. Don't bid off it on its own. 6. `cap_index`, `cap_last4` — raw inputs, not signals - capindex is the carried handicap index (already inside our fair-value model). caplast4 is the recent differential the FORM metrics are built from. They are ingredients, not additional edges — using them again would double-count what FORM already encodes. --- Recommended COMPOSITE — "true-ability-vs-index" adjustment A single per-player number in **strokes the team is expected to beat (+) or miss (−) the net handicap our fair-value model already assumes. Designed to be read alongside* the model's fair value, not to replace it: positive-and-cheap = the room mispriced on a stale index = BUY. ``` All terms in "strokes better than carried index" (positive = BUY direction). hotstrokes = (caphotIndex - 50) / 50 * 2.0 # ±2 strokes at hotIndex 0 / 100 adjplayer = 0.35 hot_strokes # FORM (momentum) — regresses, so capped influence + 0.35 capmostImproved # FORM in raw strokes — SAME signal, splits the FORM weight + 0.25 capability # CLUTCH/true-ability — clutch & ability are ONE column - 0.05 capytd # YTD drift — small regression damper adjteam = mean(adjplayerpartner1, adjplayer_partner2) ``` Weights & reasoning. - **FORM gets 0.70 total but split across hotIndex (0.35) and trend (0.35) precisely because they are the same signal** — splitting avoids double-counting while keeping form's stroke units (trend) and its field-relative scaling (hotIndex). Form is the largest weight only because both its representations are present; treat it as one 0.70 momentum bucket that partially regresses. - CLUTCH gets 0.25 from a single column (clutch ≡ ability — counted once, by design). Lower nominal weight than form's combined 0.70, but it is the most trustworthy term (trait-stable, pressure-relevant, independent of form), so in practice it's the tie-breaker that separates "hot but will choke" from "hot and shows up for money." - YTD −0.05 is a deliberate small regression brake on hot-streak tails. If you only trust one number, trust clutch on the strong partner. - Team = mean of partners. Defensible default. If you want to lean into the better-ball ceiling, weight the stronger (low) partner ~60/40, since strokes come off the low man and the better ball is usually his — but mean is the honest, low-assumption baseline. How to act on it. - Add adjteam` (in strokes) to the team's expected net edge in the Monte-Carlo / fair-value model, then re-derive P(win flight). A +3 to +6 team that the room is pricing on a stale index is the canonical undervalued BUY; a −5 team with negative clutch is a FADE even if it looks cheap. - Keep it as a read-alongside overlay, not an auto-bid. It corrects the handicap input; the auction logic in research/FINDINGS.md (cap ceiling, buyback, pacing, declining-price) still decides the price. What this composite produces on our field (sanity check). - Top individual BUY signals: Tyler Brown (hot 98, trend +8.4, clutch +33, idx −0.4), Clark Alexander (hot 99, +10.9, clutch +17.6), Joshua Stein, James Singer (clutch +29), Stephen Christian (clutch +24) — all "better than their number under pressure." - Top team rollups: Brown + Brown, Schmeelk + Stein, Christian + Springer, and — notably — Vola + Kerns, the team the model independently flagged as a sandbagger (Cap Patrol ranked Vola #4 of 971 at hotIndex 99.6 per cappatrol-re/FINDINGS.md). The composite recovering that team from a different direction is corroboration the weighting is pointed the right way. - Clear FADEs: Loewenthal, Gallagher, Jaillet, the Walseys — cold form and deeply negative clutch (−13 to −16). These will likely miss their net handicap in a money match. --- Caveats - Tails are small-sample. The ±20–33 clutch and ±12 trend extremes ride on few rounds; trust sign and rough magnitude, not the decimal. - Form double-counts by construction — hotIndex and trend are one signal; the weighting splits one bucket, it does not add two independent pieces of evidence. - Sign convention is data-empirical, not from the brief. Re-verify against hotIndex before productionizing; if Cap Patrol flips the trend sign in a future pull, flip the weight. - Selective-posting edge is absent here (turnedIn ≈ 100 for everyone). The hidden ability in this field shows up as FORM + CLUTCH, not as posting gaps. Sources Auction/format context: research/FINDINGS.md, docs/auctionrules2026.md, cappatrol-re/FINDINGS.md (all in-repo). Sports-psychology / handicap research (light web): - Hot hand — mixed, real-but-small after correction: Hot hand (Wikipedia) (https://en.wikipedia.org/wiki/Hot_hand); Surprised by the Hot Hand Fallacy? (Miller & Sanjurjo, arXiv) (https://arxiv.org/pdf/1902.01265); Hot Hand in Actual Game Situations (arXiv) (https://arxiv.org/pdf/2006.14609); Correcting for bias in hot-hand analysis: youth golf (ScienceDirect) (https://www.sciencedirect.com/science/article/abs/pii/S0167487017307390). - Clutch vs. choke as a stable, pressure-driven trait: Choking vs. Clutch Performance, J. Sport & Exercise Psych. 31(5) 2009 (https://journals.humankinetics.com/view/journals/jsep/31/5/article-p583.xml); Reaching clutch performance (InnerDrive) (https://www.innerdrive.co.uk/blog/clutch-performance-in-sport/); Choking under pressure: NFL pressure kicks (PMC) (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6445473/). - Sandbagging / selective posting & WHS safeguards (best-8-of-20, Exceptional Score Reduction, away-score tell): Telltale Signs of a Sandbagger (MyGolfSpy) (https://mygolfspy.com/news-opinion/you-asked-telltale-signs-of-a-sandbagger/); USGA WHS FAQs (https://www.usga.org/content/usga/home-page/handicapping/world-handicap-system/world-handicap-system-usga-golf-faqs.html); Outlier identification procedure (Pope of Slope) (https://www.popeofslope.com/sandbagging/procedure.html). ============================================================================ # How Markdowns Work Source URL: https://calcutta.high.green/reports/markdowns ============================================================================ How Handicap "Markdowns" Work — and How They Map to Our Edge This page separates documented mechanics (with primary sources) from our own analysis/inference. Don't conflate the two. --- Part 1 — DOCUMENTED (sourced) A. The USGA automatic "markdown": Exceptional Score Reduction Built into the World Handicap System. When you post a round whose Score Differential is meaningfully better than your Handicap Index, the system reduces your index automatically: - Differential 7.0–9.9 better than your Index → −1.0 applied to each of your most recent 20 Score Differentials. - Differential 10.0+ better → −2.0 to each of the last 20. - The reduction is baked into your record and dilutes over time as new scores post (future non-exceptional scores don't carry the adjustment). Sources: - USGA, Exceptional Score Reduction (https://www.usga.org/content/usga/home-page/handicapping/world-handicap-system/topics/exceptional-score-reduction.html) - USGA Rules of Handicapping, Rule 5.9 — Submission of an Exceptional Score (https://www.usga.org/content/usga/home-page/handicapping/roh/Content/rules/5%209%20Submission%20of%20an%20Exceptional%20Score.htm) Takeaway: a player who fires a round far under his number gets mechanically docked — GHIN-wide, any round, tournament rounds included. This is the formal "hit." B. Cap Patrol — the club-level flagging layer A third-party algorithm (separate from USGA) that clubs use to spot sandbaggers and vanity handicaps. - Built by Bob Thurner (Cincinnati; 0.4 Index; sports-analytics background; former club president/handicap chairman). - Syncs with GHIN + course tee sheets; uses 43 data points across 5 criteria: 1. Handicap Index over the past 12 months (trend), 2. Home vs. away scoring, 3. Player "potential" (best recent rounds vs. current index), 4. % of scores turned in (selective posting), 5. Tournament finishes. - It recommends handicap adjustments up or down, and flags whom to watch. ~1,100 clubs / 620,000+ golfers. Sources: - Golf Digest, How to catch a sandbagger (https://www.golfdigest.com/story/how-to-catch-a-sandbagger-computer-algorithm-tournament-cheats) - cappatrol.com (https://cappatrol.com/) Takeaway: Cap Patrol's flag is driven by the competition-vs-casual gap plus a fast-dropping or lagging index — the textbook sandbagger fingerprint. --- Part 2 — OUR ANALYSIS / INFERENCE (not a published fact) The following is our read of the data we pulled, not documented by USGA or Cap Patrol. Treat it as a working hypothesis, not a citation. - The Cap Patrol metric we pulled called clutch (identical to ability, corr ≈ 1.0) appears to capture competition-relative performance: it ran roughly −19 to +33 across our 228-player field and is independent of form (Pearson ≈ −0.08 with hotIndex). We infer** it is close to the "performs above his card when it counts" axis — but Cap Patrol does not publish its exact weighting, so this is inference. - In our field, the selective-posting driver is effectively dead — ~225 of 228 players post 100% of rounds. So any markdown pressure here comes from tournament outperformance + the potential-vs-index gap, not unposted scores. - The players who, in our analysis, fit the "plays better than his card when money's on the line" profile — by beating their predicted points in both 2024 and 2025 — are Copeland, Estes, Keister, Bachstein, Flammia, plus Nick's eyewitness read on Williford (a 73 in a 4-club event). These names came from our residual analysis + human intel, not from a Cap Patrol markdown list. --- Bottom line The mechanics of how golfers get marked down are real and sourced (USGA Exceptional Score Reduction; Cap Patrol's 5 criteria). The mapping of those mechanics onto our clutch column and the specific named players is our edge work — defensible, cross-validated across methods, but ours, not a published ranking. To get a true "who's been marked down" list, the clean path is a fresh Cap Patrol token → sort our field by clutch/ability and cross-reference high tournament-finish history. That combination is the markdown signal. (Headless re-pull: src/calcutta/cappatrol.py.) ============================================================================ # Competitor Strategy Source URL: https://calcutta.high.green/reports/competitor ============================================================================ Competitor & Game-Theory Strategy · Internal · 10 June 2026 2026 Member-Member Calcutta — How to Exploit the Room Beating the bidders , not the golf. A behavioral map of four named rivals and a semi-unbudgeted, drunk, $2,000-cap-anchored auction — turned into where we bid, where we fold, and where we pounce. 14 Hard cap-out magnets — race to $2,000, AVOID 30 Low-competition teams — our clean water The Sharp The one rival on our biggest edge (Wood + Estes) $286 Edge on Wood + Estes — our prize & most contested lot Executive summary We are not trying to predict golf — outcomes are near-random. Our edge is buying the same teams for less than a drunk, overconfident, cap-anchored room will pay. This report turns the four rival archetypes into a map of where the money goes so we can stand in the empty water. Three forces define the room: it is cap-anchored (~14–18 hyped low-handicap "name" teams race to the $2,000 round-number cap — structurally overpriced; we never chase one); it is semi-unbudgeted & drunk (no FOMO discipline → overpays early, cheap steals late); and it has predictable chasers (the Overconfident Veteran and the Loyalist inflate avoidable lanes; the Sharp is the only one fishing our pond — and he is on our single biggest edge, Wood + Estes ). The play: let the Overconfident Veteran / the Loyalist / the Wild Card overpay to the cap on marquee and high-handicap lots; concede the public wars; and pounce on quiet mid-tier teams late that no archetype's profile fits. On the buy card, Knapp + Keister (early, #20) and Barbaree + Williford (#63) are clean ; Wood + Estes is our prize but contested by the Sharp — win it on discipline, not a war. Honesty: this is a behavioral read , not a fitted model. Chaser tags are heuristics from the intel memo, and a drunk room is high-variance by definition. Every "likely chaser" is a prior to update live in the room , not a fact. 1. Per-rival profiles Rival Tendency What they chase How we play them The Overconfident Veteran Knows everyone; overconfident → overpays . We have inside read via Jason. Marquee, low-handicap names — scratch / single-digit teams in Flights 1–3. Drives up guys he likes. Let him win the names to $2k. Never be the underbidder that "saves" him money. Use Jason to learn which names he's hot on this year; let him burn cash early. The Loyalist Bids older players / higher handicaps he trusts → inflates high-handicap teams. High-team-handicap lots (≈19+), trusted veteran high-cappers. Ignores young low studs. Concede the high-handicap flights. The 10-stroke cap lowers high-capper value, so his inflation hits the worst lots. Value lives in the young/low teams he won't touch. The Sharp Sharp day-trader — the dangerous one. Hunts value / sandbaggers, same as us. Our pond: sandbaggers, positive-edge, hidden upside. Flagged on Wood+Estes, Wright+Copeland, Nodar+Heslep . Don't tip our hand. Don't be visibly eager on shared lots. On Wood+Estes pay to our max but no war premium ; have a backup. Move with quiet confidence where we don't overlap. The Wild Card Smart wild card → unpredictable, can spike any price. No fixed lane; weighted to visible / hyped lots (piles onto cap-out fever). Don't anchor to him. Set our max before the lot; hold it. If he spikes a clean target past our number, let it go — another lot is minutes away. Chaser footprint across 120 lots: the Loyalist 59 (thin, high-hcp) · the Overconfident Veteran 47 (marquee/early) · the Sharp 44 (value) · the Wild Card 20 (hyped). The Overconfident Veteran and the Loyalist cover the loud, expensive parts of the board — which is exactly why the quiet middle and the late lots are open. 2. Team competition map (team_competition.csv) All 120 teams scored with per-rival affinity (0–3), expected competition, likely chasers, and a cap-out flag. Distribution: 40 High competition 50 Medium competition 30 Low competition (our lanes) 14 Cap-out magnets ($2k) The 14 cap-out magnets — AVOID All single-/low-double-digit "name" teams, all Flights 1–3 , all top-of-field price. This is where the Overconfident Veteran + the Wild Card burn cash to $2,000. Do not be the underbidder. Lot Flt Team Team hcp Est $ Magnet for 12 1 McHale + Bagley 5.2 1700 Overconfident Veteran, Wild Card 79 1 Perry + Dye 1.0 1700 Overconfident Veteran, Wild Card 82 1 Beatty + Guiendon 0.1 1700 Overconfident Veteran, Wild Card 4 1 Pilger + Bhatia 5.1 1600 Overconfident Veteran, Wild Card 30 1 Vola + Kerns 5.0 1500 Overconfident Veteran, Sharp 111 1 Terranova + Davis 4.2 1500 Overconfident Veteran, Wild Card 33 2 Hatcher + Berry 8.4 1400 Overconfident Veteran, Sharp 51 2 Fann + Hartz 7.9 1300 Overconfident Veteran, Wild Card 66 2 Downey + Shearer 8.0 1300 Overconfident Veteran, Sharp 5 3 Martin + Torres 9.6 1200 Overconfident Veteran, Sharp 42 3 Brown + Watson 9.1 1200 Overconfident Veteran, Sharp 46 2 Aller + Walker 7.7 1200 Overconfident Veteran, Wild Card 53 2 Hurst + Jordan 7.7 1200 Overconfident Veteran, Wild Card 100 3 Patterson + Shirley 10.3 1200 Overconfident Veteran, Wild Card Buy-card competition assessment Lot Team Flt Comp Likely chasers Cap? Est Fair Edge Read 20 Knapp + Keister 6 Low Overconfident Veteran (weak) no 800 696 −104 Clean. Jason-flagged value; no profile fits. Comes early (#20) — buy before cash commits elsewhere; don't overpay just because it's early. 63 Barbaree + Williford 8 Low Sharp (weak) no 900 620 −280 Cleanest lane on the card. Williford = human-intel sandbagger (watched shoot 73). The Sharp scores only 1. Pure private-info edge — value is hidden, not in the listed number. 76 Nodar + Heslep 8 High Overconfident Veteran, Sharp no 700 814 +114 Contested — positive listed edge draws sharp money. Win only at/below max ($800); have flight-8 alternates ready. 94 Wright + Copeland 5 High Overconfident Veteran, Sharp no 900 945 +45 Copeland = result-confirmed Flight-5 winner (2025); both room and sharp can see it. Late lot (#94) helps. No war premium. 104 Wood + Estes 9 Med Sharp no 700 986 +286 Our biggest edge — and the Sharp is on it (3/3). THE contested lot. Late (#104/120) helps if rivals tap out, but the Sharp is sober. Pay max $1,000 without flinching; do not exceed. Have a fallback. Is the Sharp on Wood + Estes? Yes — strongest possible read (3/3). It is our top edge and our top contested lot. Does the Loyalist inflate any of our targets? No. He scores 0–1 on every buy-card team; his inflation lands entirely on lots we're avoiding. Good news. 3. Exploitation playbook A. Clean lanes (no profile fits → cheap) 30 Low-competition teams; 17 with edge ≥ −$100. Mid-handicap, no-name teams: too high-cap for the Overconfident Veteran, too low-cap for the Loyalist, not flagged enough for the Sharp. Representative open water: Lot Team Flt Est $ Fair $ Edge $ Nominal chaser 21 Collins + Minter 12 500 546 +46 Loyalist (weak) 37 Thomas + Stricklin 7 700 723 +23 Overconfident Veteran (weak) 24 Driggars + Richards 7 800 808 +8 Overconfident Veteran (weak) 48 Bull + Singer 8 700 673 −27 Overconfident Veteran (weak) 59 Shirley + Aronson 10 700 663 −37 Loyalist (weak) 73 Marjoram + Peters 7 900 863 −37 Overconfident Veteran (weak) 75 Robinson + McCormack 11 600 560 −40 Loyalist (weak) Plus the buy-card pair Knapp + Keister and Barbaree + Williford , both Low. B. Where to avoid wars - All 14 cap-out magnets (table above) — the Overconfident Veteran + the Wild Card pay $2k; we can't profit there. - High-handicap flights (≈19+) where the Loyalist is strong — and the 10-stroke cap makes high-cappers worth less , so his inflated lane is doubly bad. Concede it. - Wood + Estes past our $1,000 max — the Sharp is sharp and sober; let the +$286 go rather than turn it into a −EV war. C. Using the KNOWN order to pace budget - Early board (1–60) is cap-out heavy and cash-rich (magnets at 4, 5, 12, 30, 33, 42, 46, 51, 53). Let the room spend itself. Our only early must-buy is Knapp + Keister (#20) — acquire, then go quiet. - Late board (85–120) is the drunk-and-tapped-out discount. Three of five targets land here: Nodar+Heslep (#76), Wright+Copeland (#94), Wood+Estes (#104) , plus late low-comp value (Cherof+Brocard #87, Christian+Springer #118). Hold dry powder for the back third. - Budget shape: spend < ~30% of bankroll before lot 60. Reserve the majority for #63 → #104 , where our edge and the room's fatigue coincide. D. How the room's psychology sharpens our discipline - Cap anchor = a do-not-cross line. The $2k magnet pulls everyone up; our rule is the inverse — the closer a lot drifts to $2k, the more certain we fold. Round numbers are their bias, not ours. - Semi-unbudgeted = our patience is the edge. They have no FOMO governor; we do. Pre-commit a max per lot, in writing, before it opens; never raise it live. - Pounce on quiet mid-tier teams late. When attention and cash are spent, a $500–$900 mid-tier lot with no natural chaser goes off cheap — our highest-EV behavior of the night. 4. Tie to the buy card Target Lot Timing Comp Action Knapp + Keister 20 Early Low Acquire early , at/below max ($700). Then conserve. Barbaree + Williford 63 Mid Low Cleanest edge — buy. Pure private-info (Williford). The Sharp unlikely to contest hard. Nodar + Heslep 76 Late High Buy at/below max ($800). Positive edge attracts sharps; have flight-8 backups. Wright + Copeland 94 Late High Buy at/below max ($900). Result-confirmed sandbagger; no war premium. Wood + Estes 104 Late Med Our prize. Pay max $1,000 without flinching; do not exceed. The Sharp is the threat — discipline beats him, a war doesn't. Pacing logic: Knapp early, then a long quiet stretch, then a concentrated late push (#63–#104) where four of five targets and the room's fatigue line up. Keep > ~70% of bankroll for the back two-thirds. Assumptions & uncertainty - Behavioral, not fitted. Chaser tags are heuristics from the intel memo applied to handicaps, prices, and flags — update them live. - Cap-out set is a floor, not a ceiling. We flag 14 unambiguous magnets; a drunk room can spike more. When in doubt near $2k, fold. - Loyalist proxy. "Older players he trusts" is approximated by high team handicap + a high-side individual; we lack ages, so a low-hcp team with a trusted veteran may surprise us. - The Sharp is sober and sharp. We assume he does not overpay and does find the same value — so on shared lots, a pre-set max is the only safe weapon. - Est prices are model expected clears ( est_price ); actual clears vary widely both ways. Maxes come from recommended_max_bid , not est. - Jason's inside read on the Overconfident Veteran's specific hot names this year is the highest-value live update — refresh it the day of the auction. - Near-random outcomes still hold. Beating the room on price is necessary but not, alone, a profit guarantee; it is the lever we control. Sources: data/raw/auction_intel.md , data/raw/auction_order_2026.csv , valuation/bid_sheet_enriched.csv . Analysis & classifier: analysis/competitor_strategy/build_competition.py → team_competition.csv . Internal use — 2026 Member-Member Calcutta. ============================================================================ # Matchup Analysis Source URL: https://calcutta.high.green/reports/matchups ============================================================================ 2026 Calcutta — Round-by-Round Matchup Analysis Generated by valuation/matchups/matchupmodel.py. Monte-Carlo N=40,000 sims per pairing (15 pairings x 20 flights = 300 matches simulated), dedicated fixed seed. Player scoring distributions and the 9-hole net better-ball machinery are imported read-only from valuemodel.py. This is a companion to the aggregate flight sim, not a replacement. The aggregate sim answers who wins the flight; here we open up the 6x6 head-to-head matrix behind it to see where each team's points come from, which matches actually decide the flight, and which favorites are fragile (lead hinges on coin-flips) vs robust (win across the board). The suggestedadjustment column is an interpretive lean** to apply on top of the sim's P(win) — explicitly NOT a new probability. How to read this Headline finding first: 9-hole NET better-ball off the LOW player in the foursome is a genuinely high-variance, near-coin-flip format. Across the whole field ~62% of head-to-head matchups land inside 45–55%, only ~4% reach 60/40, and the single most lopsided matchup anywhere is ~78/22. Handicaps off the low player plus only 9 holes to separate teams compress nearly everything toward 50/50. So 'fragile favorite' is the NORM here, and the bands below are calibrated to this format's realized (tight) dispersion, not a wider sport. - pAbeatsB: P(team A wins the 9-hole match) + 0.5·P(match halved), so it pairs exactly with the points split and pAbeatsB + pBbeatsA = 1`. - Strong win for A: pAbeatsB ≥ 57% (≈ top quartile of matchups in this field). Likely loss: ≤ 43%. Swing: in between. A tighter coin-flip** band (45%–55%) is used for fragility. - robustnessscore** (0–1): wide head-to-head margins AND few coin-flips ⇒ high. A team can lead on expected points yet score LOW here if that lead is built on 50/50s. - suggestedadjustment: capped at ±6%. UP = the sim may under-rate this team (favorable matchup spread / strong H2H vs the specific top rivals it must beat). DOWN** = the sim may over-rate it (nominal lead leans on winning multiple coin-flips). Interpretive, not a P(win). Inherited assumptions (from `value_model.py`) Same per-hole construction as the aggregate sim: 9 holes, integer gross-vs-par clipped to [-2,+7], net better-ball off the LOW player in the foursome, 10 match points (+1/hole, +0.5/halve, +1 to the match winner). Variance layers RHOFIELD=0.15, RHOPARTNERS=0.35, HOLENOISEFRAC=0.65. The flight-wide FIELD shock is shared by all four players in a match (it nearly cancels head-to-head, matching how valuemodel treats a single match); the PARTNER shock is per team and does not cancel. Full player-distribution assumptions are documented in valuemodel.py/flightreport.md`. Favorites at a glance — fragile vs robust | Flight | Favorite (by E[pts]) | E[pts] | Robustness | Coin-flips | % wins from swings | Sim P(win) | Call | |---:|---|---:|---:|---:|---:|---:|:--:| | 1 | Vola + Kerns | 26.6 | 0.09 | 2 | 78% | 23% | mixed | | 2 | Downey + Shearer | 26.4 | 0.07 | 2 | 100% | 23% | mixed | | 3 | Burns + Preston | 26.0 | 0.04 | 4 | 77% | 19% | FRAGILE | | 4 | Greenspan + Horne | 25.7 | 0.03 | 3 | 100% | 19% | mixed | | 5 | Pearson + Ferguson | 26.8 | 0.10 | 2 | 57% | 21% | mixed | | 6 | Swiger + Dahlhauser | 26.5 | 0.07 | 3 | 79% | 25% | mixed | | 7 | Marjoram + Peters | 26.7 | 0.10 | 1 | 79% | 23% | mixed | | 8 | Nodar + Heslep | 26.0 | 0.04 | 4 | 100% | 23% | FRAGILE | | 9 | Wood + Estes | 30.5 | 0.35 | 0 | 0% | 43% | ROBUST | | 10 | Fought + Fought | 27.1 | 0.13 | 1 | 38% | 24% | ROBUST | | 11 | Richardson + Loricchio | 25.9 | 0.04 | 3 | 100% | 22% | mixed | | 12 | Wright + Gadsby | 26.8 | 0.10 | 2 | 57% | 25% | mixed | | 13 | Christian + Springer | 26.3 | 0.06 | 3 | 79% | 22% | mixed | | 14 | Kessler + Brinson | 27.4 | 0.17 | 0 | 59% | 24% | ROBUST | | 15 | Perry + Calobrisi | 26.6 | 0.09 | 2 | 79% | 20% | mixed | | 16 | Wilson + Slavis | 26.2 | 0.05 | 4 | 78% | 20% | FRAGILE | | 17 | Poje + Burleson | 26.6 | 0.09 | 2 | 79% | 22% | mixed | | 18 | Foley + Hampy | 26.4 | 0.06 | 4 | 78% | 21% | FRAGILE | | 19 | Toole + Calvert | 26.9 | 0.11 | 1 | 79% | 22% | ROBUST | | 20 | Barnes + Smelcer | 31.2 | 0.39 | 0 | 0% | 39% | ROBUST | UP/DOWN adjustment calls (interpretive overlay) These are leans on top of the aggregate sim's P(win), capped at ±6%. They flag where the matchup structure suggests the sim may be mis-rating a team. Not a replacement for the sim. ▲ UP — the sim may UNDER-rate these (favorable matchup spread): | Flight | Team | Rank | Robustness | H2H vs top rivals | Nudge | Why | |---:|---|---:|---:|---:|---:|---| | 9 | Wood + Estes | 1 | 0.35 | 65% | +6.0% | ROBUST favorite: comparatively wide margins, 65% H2H vs top rivals | | 14 | Kessler + Brinson | 1 | 0.17 | 57% | +5.5% | ROBUST favorite: comparatively wide margins, 57% H2H vs top rivals | | 20 | Barnes + Smelcer | 1 | 0.39 | 62% | +3.8% | ROBUST favorite: comparatively wide margins, 62% H2H vs top rivals | | 10 | Fought + Fought | 1 | 0.13 | 54% | +3.7% | ROBUST favorite: comparatively wide margins, 54% H2H vs top rivals | | 7 | Marjoram + Peters | 1 | 0.10 | 54% | +3.1% | FRAGILE favorite: 1/5 coin-flips; 79% of expected wins from swings | | 19 | Toole + Calvert | 1 | 0.11 | 55% | +3.0% | ROBUST favorite: comparatively wide margins, 55% H2H vs top rivals | | 15 | Perry + Calobrisi | 1 | 0.09 | 54% | +2.9% | FRAGILE favorite: 2/5 coin-flips; 79% of expected wins from swings | | 12 | Wright + Gadsby | 1 | 0.10 | 54% | +2.8% | favorite in a coin-flip flight; lead is soft | | 17 | Poje + Burleson | 1 | 0.09 | 54% | +2.7% | FRAGILE favorite: 2/5 coin-flips; 79% of expected wins from swings | | 13 | Christian + Springer | 1 | 0.06 | 52% | +2.1% | FRAGILE favorite: 3/5 coin-flips; 79% of expected wins from swings | | 5 | Pearson + Ferguson | 1 | 0.10 | 52% | +2.0% | favorite in a coin-flip flight; lead is soft | | 2 | Hatcher + Berry | 2 | 0.07 | 52% | +1.8% | contender leaning on 2/5 coin-flips | | 5 | Wright + Copeland | 2 | 0.10 | 51% | +1.7% | | | 6 | Swiger + Dahlhauser | 1 | 0.07 | 53% | +1.5% | FRAGILE favorite: 3/5 coin-flips; 79% of expected wins from swings | | 1 | McHale + Bagley | 2 | 0.08 | 51% | +1.4% | | | 11 | Richardson + Loricchio | 1 | 0.04 | 53% | +0.8% | FRAGILE favorite: 3/5 coin-flips; 100% of expected wins from swings | | 18 | Foley + Hampy | 1 | 0.06 | 52% | +0.7% | FRAGILE favorite: 4/5 coin-flips; 78% of expected wins from swings | | 6 | Vaniman + Hatz | 2 | 0.06 | 50% | +0.7% | contender leaning on 4/5 coin-flips | | 8 | Nodar + Heslep | 1 | 0.04 | 52% | +0.6% | FRAGILE favorite: 4/5 coin-flips; 100% of expected wins from swings | | 3 | Brown + Brown | 6 | 0.18 | 40% | +0.5% | | | 16 | Wilson + Slavis | 1 | 0.05 | 52% | +0.5% | FRAGILE favorite: 4/5 coin-flips; 78% of expected wins from swings | | 18 | Hoard + Woods | 6 | 0.16 | 41% | +0.4% | | | 16 | Goodloe + Santivanez | 6 | 0.13 | 41% | +0.4% | | | 3 | Patterson + Shirley | 2 | 0.04 | 51% | +0.3% | contender leaning on 4/5 coin-flips | ▼ DOWN — the sim may OVER-rate these (lead leans on coin-flips): | Flight | Team | Rank | Coin-flips | % wins from swings | Nudge | Why | |---:|---|---:|---:|---:|---:|---| | 9 | Kattookaran + Levitas | 2 | 3 | 86% | -5.0% | contender leaning on 3/5 coin-flips | | 5 | Levin + Leaf | 3 | 4 | 100% | -3.7% | contender leaning on 4/5 coin-flips | | 20 | Ballard + Armstrong | 3 | 0 | 0% | -3.5% | contender with comparatively robust margins | | 10 | Gelinas + Chafin | 3 | 4 | 100% | -3.4% | contender leaning on 4/5 coin-flips | | 9 | Gallagher + Hansell | 3 | 3 | 63% | -3.1% | | | 14 | Higdon + Doyle | 3 | 4 | 100% | -3.1% | contender leaning on 4/5 coin-flips | | 14 | Dalmau + Pons | 2 | 3 | 100% | -2.5% | contender leaning on 3/5 coin-flips | | 12 | Collins + Minter | 2 | 3 | 100% | -2.4% | contender leaning on 3/5 coin-flips | | 6 | Kitchens + Taylor | 3 | 4 | 100% | -2.2% | contender leaning on 4/5 coin-flips | | 17 | Lal + Grover | 3 | 4 | 100% | -2.0% | contender leaning on 4/5 coin-flips | | 7 | Webber + Sevy | 3 | 4 | 100% | -1.8% | contender leaning on 4/5 coin-flips | | 2 | Fann + Hartz | 3 | 3 | 100% | -1.8% | contender leaning on 3/5 coin-flips | | 19 | Harte + Love | 3 | 3 | 100% | -1.7% | contender leaning on 3/5 coin-flips | | 15 | Alexander + Lambeth | 3 | 4 | 100% | -1.5% | contender leaning on 4/5 coin-flips | | 15 | Jaillet + Jaillet | 2 | 4 | 100% | -1.4% | contender leaning on 4/5 coin-flips | | 19 | Evans + Rowley | 2 | 3 | 100% | -1.4% | contender leaning on 3/5 coin-flips | | 1 | Terranova + Davis | 3 | 3 | 77% | -1.4% | contender leaning on 3/5 coin-flips | | 7 | Driggars + Richards | 2 | 5 | 100% | -1.3% | contender leaning on 5/5 coin-flips | | 13 | Hancher + Helms | 2 | 5 | 100% | -1.0% | contender leaning on 5/5 coin-flips | | 13 | DiCicco + Kulik | 3 | 4 | 100% | -1.0% | contender leaning on 4/5 coin-flips | | 9 | Oberholtzer + Doyle | 5 | 4 | 87% | -0.8% | | | 9 | Thilmany + Mulick | 4 | 4 | 85% | -0.8% | | | 20 | Morge + Callahan | 5 | 1 | 30% | -0.7% | | | 20 | Bell + Harren | 4 | 0 | 0% | -0.7% | | | 20 | King + Scales | 6 | 1 | 27% | -0.7% | | | 11 | Taylor + Simons | 3 | 5 | 100% | -0.6% | contender leaning on 5/5 coin-flips | | 17 | Brenner + McCormick | 2 | 4 | 100% | -0.6% | contender leaning on 4/5 coin-flips | | 16 | McBride + Hall | 3 | 4 | 77% | -0.6% | contender leaning on 4/5 coin-flips | | 12 | Benson + Brosnahan | 4 | 5 | 100% | -0.5% | | | 18 | Seeley + Seeley | 3 | 4 | 77% | -0.5% | contender leaning on 4/5 coin-flips | | 1 | Pilger + Bhatia | 4 | 3 | 100% | -0.5% | | | 3 | Martin + Torres | 3 | 4 | 77% | -0.4% | contender leaning on 4/5 coin-flips | | 12 | Reynolds + Swart | 3 | 2 | 78% | -0.4% | contender leaning on 2/5 coin-flips | | 8 | Cumming + Odea | 3 | 5 | 100% | -0.4% | contender leaning on 5/5 coin-flips | | 16 | Bhatia + Tear IV | 5 | 5 | 100% | -0.4% | | | 9 | Baggett + Vijay | 6 | 2 | 66% | -0.4% | | | 18 | Muldoon + Baribeau | 4 | 4 | 100% | -0.4% | | | 5 | Leprohon + Leprohon | 5 | 3 | 82% | -0.4% | | | 18 | Bhatia + Tear | 5 | 4 | 100% | -0.4% | | | 3 | Brown + Watson | 5 | 4 | 100% | -0.4% | | | 16 | Flammia + Gray | 4 | 4 | 100% | -0.4% | | | 6 | Foresman + Benson | 4 | 5 | 100% | -0.3% | | | 14 | Goodrick + Corry | 6 | 2 | 100% | -0.3% | | | 2 | Walsey + Walsey | 5 | 4 | 100% | -0.3% | | | 17 | Feldman + Waronker | 4 | 5 | 100% | -0.3% | | | 4 | Green + Green | 3 | 5 | 100% | -0.3% | contender leaning on 5/5 coin-flips | Per-flight detail Flight 1 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Vola | McHale | Terranova | Pilger | Perry | Beatty | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Vola + Kerns | — | 51% | 53% | 57% | 57% | 60% | 26.6 | | McHale + Bagley | 49% | — | 52% | 53% | 59% | 63% | 26.5 | | Terranova + Davis | 47% | 48% | — | 51% | 56% | 60% | 25.6 | | Pilger + Bhatia | 43% | 47% | 49% | — | 54% | 57% | 25.0 | | Perry + Dye | 43% | 41% | 44% | 46% | — | 53% | 23.8 | | Beatty + Guiendon | 40% | 37% | 40% | 43% | 47% | — | 22.5 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Vola + Kerns | 26.6 | 1 | 4 | 0 | 2 | 0.09 | -0.0% | FRAGILE favorite: 2/5 coin-flips; 78% of expected wins from swings | | McHale + Bagley | 26.5 | 2 | 3 | 0 | 3 | 0.08 | +1.4% | | | Terranova + Davis | 25.6 | 1 | 4 | 0 | 3 | 0.06 | -1.4% | contender leaning on 3/5 coin-flips | | Pilger + Bhatia | 25.0 | 0 | 5 | 0 | 3 | 0.06 | -0.5% | | | Perry + Dye | 23.8 | 0 | 4 | 1 | 2 | 0.09 | -0.3% | | | Beatty + Guiendon | 22.5 | 0 | 2 | 3 | 1 | 0.16 | +0.2% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | McHale + Bagley vs Vola + Kerns | 49% | 2 | 1 | 0.49 | | Vola + Kerns vs Terranova + Davis | 53% | 1 | 3 | 0.31 | | McHale + Bagley vs Terranova + Davis | 52% | 2 | 3 | 0.24 | Flight 2 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Downey | Hatcher | Fann | Hurst | Walsey | Aller | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Downey + Shearer | — | 51% | 53% | 56% | 56% | 57% | 26.4 | | Hatcher + Berry | 49% | — | 56% | 56% | 53% | 58% | 26.3 | | Fann + Hartz | 47% | 44% | — | 49% | 53% | 55% | 24.9 | | Hurst + Jordan | 44% | 44% | 51% | — | 50% | 56% | 24.7 | | Walsey + Walsey | 44% | 47% | 47% | 50% | — | 51% | 24.4 | | Aller + Walker | 43% | 42% | 45% | 44% | 49% | — | 23.3 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Downey + Shearer | 26.4 | 0 | 5 | 0 | 2 | 0.07 | -0.0% | FRAGILE favorite: 2/5 coin-flips; 100% of expected wins from swings | | Hatcher + Berry | 26.3 | 1 | 4 | 0 | 2 | 0.07 | +1.8% | contender leaning on 2/5 coin-flips | | Fann + Hartz | 24.9 | 0 | 5 | 0 | 3 | 0.05 | -1.8% | contender leaning on 3/5 coin-flips | | Hurst + Jordan | 24.7 | 0 | 5 | 0 | 2 | 0.06 | -0.2% | | | Walsey + Walsey | 24.4 | 0 | 5 | 0 | 4 | 0.03 | -0.3% | | | Aller + Walker | 23.3 | 0 | 4 | 1 | 1 | 0.10 | +0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Hatcher + Berry vs Downey + Shearer | 49% | 2 | 1 | 0.49 | | Fann + Hartz vs Downey + Shearer | 47% | 3 | 1 | 0.31 | | Hatcher + Berry vs Fann + Hartz | 56% | 2 | 3 | 0.22 | Flight 3 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Burns | Patterson | Martin | Embleau | Brown | Brown | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Burns + Preston | — | 50% | 51% | 52% | 53% | 60% | 26.0 | | Patterson + Shirley | 50% | — | 53% | 48% | 53% | 61% | 25.7 | | Martin + Torres | 49% | 47% | — | 51% | 51% | 59% | 25.4 | | Embleau + Loewenthal | 48% | 52% | 49% | — | 50% | 58% | 25.4 | | Brown + Watson | 47% | 47% | 49% | 50% | — | 57% | 25.0 | | Brown + Brown | 40% | 39% | 41% | 42% | 43% | — | 22.4 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Burns + Preston | 26.0 | 1 | 4 | 0 | 4 | 0.04 | +0.1% | FRAGILE favorite: 4/5 coin-flips; 77% of expected wins from swings | | Patterson + Shirley | 25.7 | 1 | 4 | 0 | 4 | 0.04 | +0.3% | contender leaning on 4/5 coin-flips | | Martin + Torres | 25.4 | 1 | 4 | 0 | 4 | 0.04 | -0.4% | contender leaning on 4/5 coin-flips | | Embleau + Loewenthal | 25.4 | 1 | 4 | 0 | 4 | 0.03 | -0.0% | | | Brown + Watson | 25.0 | 0 | 5 | 0 | 4 | 0.03 | -0.4% | | | Brown + Brown | 22.4 | 0 | 1 | 4 | 0 | 0.18 | +0.5% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Burns + Preston vs Patterson + Shirley | 50% | 1 | 2 | 0.50 | | Martin + Torres vs Burns + Preston | 48% | 3 | 1 | 0.32 | | Embleau + Loewenthal vs Burns + Preston | 48% | 4 | 1 | 0.24 | Flight 4 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Greenspan | Brown | Green | Williams | Powers | Kilfeather | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Greenspan + Horne | — | 50% | 51% | 50% | 56% | 56% | 25.7 | | Brown + Cozzo | 50% | — | 51% | 49% | 53% | 56% | 25.4 | | Green + Green | 49% | 49% | — | 53% | 52% | 52% | 25.4 | | Williams + Lemieux | 50% | 51% | 47% | — | 50% | 49% | 24.9 | | Powers + Merrigan | 44% | 47% | 48% | 50% | — | 50% | 24.4 | | Kilfeather + Otto | 44% | 44% | 48% | 51% | 50% | — | 24.2 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Greenspan + Horne | 25.7 | 0 | 5 | 0 | 3 | 0.03 | +0.2% | FRAGILE favorite: 3/5 coin-flips; 100% of expected wins from swings | | Brown + Cozzo | 25.4 | 0 | 5 | 0 | 4 | 0.03 | +0.1% | contender leaning on 4/5 coin-flips | | Green + Green | 25.4 | 0 | 5 | 0 | 5 | 0.02 | -0.3% | contender leaning on 5/5 coin-flips | | Williams + Lemieux | 24.9 | 0 | 5 | 0 | 5 | 0.01 | -0.1% | | | Powers + Merrigan | 24.4 | 0 | 5 | 0 | 4 | 0.03 | -0.1% | | | Kilfeather + Otto | 24.2 | 0 | 5 | 0 | 3 | 0.04 | -0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Brown + Cozzo vs Greenspan + Horne | 50% | 2 | 1 | 0.50 | | Greenspan + Horne vs Green + Green | 51% | 1 | 3 | 0.33 | | Greenspan + Horne vs Williams + Lemieux | 50% | 1 | 4 | 0.25 | Flight 5 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Pearson | Wright | Levin | Seeyave | Leprohon | Hancock | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Pearson + Ferguson | — | 53% | 52% | 61% | 56% | 58% | 26.8 | | Wright + Copeland | 47% | — | 54% | 58% | 58% | 57% | 26.4 | | Levin + Leaf | 48% | 46% | — | 54% | 54% | 56% | 25.5 | | Seeyave + Hernandez | 39% | 42% | 46% | — | 50% | 56% | 24.0 | | Leprohon + Leprohon | 44% | 42% | 46% | 50% | — | 48% | 23.8 | | Hancock + Bachstein | 42% | 43% | 44% | 44% | 52% | — | 23.6 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Pearson + Ferguson | 26.8 | 2 | 3 | 0 | 2 | 0.10 | +2.0% | favorite in a coin-flip flight; lead is soft | | Wright + Copeland | 26.4 | 2 | 3 | 0 | 2 | 0.10 | +1.7% | | | Levin + Leaf | 25.5 | 0 | 5 | 0 | 4 | 0.05 | -3.7% | contender leaning on 4/5 coin-flips | | Seeyave + Hernandez | 24.0 | 0 | 3 | 2 | 2 | 0.09 | -0.1% | | | Leprohon + Leprohon | 23.8 | 0 | 4 | 1 | 3 | 0.06 | -0.4% | | | Hancock + Bachstein | 23.6 | 0 | 4 | 1 | 1 | 0.10 | -0.2% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Pearson + Ferguson vs Wright + Copeland | 53% | 1 | 2 | 0.47 | | Pearson + Ferguson vs Levin + Leaf | 52% | 1 | 3 | 0.32 | | Wright + Copeland vs Levin + Leaf | 54% | 2 | 3 | 0.23 | Flight 6 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Swiger | Vaniman | Kitchens | Foresman | White | Knapp | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Swiger + Dahlhauser | — | 52% | 54% | 53% | 57% | 59% | 26.5 | | Vaniman + Hatz | 48% | — | 53% | 55% | 59% | 55% | 26.1 | | Kitchens + Taylor | 46% | 47% | — | 50% | 54% | 56% | 25.2 | | Foresman + Benson | 47% | 45% | 50% | — | 54% | 54% | 24.9 | | White + Dunning | 43% | 41% | 46% | 46% | — | 50% | 23.6 | | Knapp + Keister | 41% | 45% | 44% | 46% | 50% | — | 23.6 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Swiger + Dahlhauser | 26.5 | 1 | 4 | 0 | 3 | 0.07 | +1.5% | FRAGILE favorite: 3/5 coin-flips; 79% of expected wins from swings | | Vaniman + Hatz | 26.1 | 1 | 4 | 0 | 4 | 0.06 | +0.7% | contender leaning on 4/5 coin-flips | | Kitchens + Taylor | 25.2 | 0 | 5 | 0 | 4 | 0.04 | -2.2% | contender leaning on 4/5 coin-flips | | Foresman + Benson | 24.9 | 0 | 5 | 0 | 5 | 0.03 | -0.3% | | | White + Dunning | 23.6 | 0 | 4 | 1 | 3 | 0.07 | -0.1% | | | Knapp + Keister | 23.6 | 0 | 4 | 1 | 3 | 0.07 | -0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Vaniman + Hatz vs Swiger + Dahlhauser | 48% | 2 | 1 | 0.48 | | Kitchens + Taylor vs Swiger + Dahlhauser | 46% | 3 | 1 | 0.31 | | Kitchens + Taylor vs Vaniman + Hatz | 47% | 3 | 2 | 0.24 | Flight 7 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Marjoram | Driggars | Webber | Thomas | Rekenthale | Hock | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Marjoram + Peters | — | 53% | 56% | 55% | 56% | 58% | 26.7 | | Driggars + Richards | 47% | — | 51% | 52% | 53% | 54% | 25.5 | | Webber + Sevy | 44% | 49% | — | 50% | 52% | 53% | 24.9 | | Thomas + Stricklin | 45% | 48% | 50% | — | 51% | 53% | 24.8 | | Rekenthaler + Stafford | 44% | 47% | 48% | 49% | — | 51% | 24.3 | | Hock + Stephens | 42% | 46% | 47% | 47% | 49% | — | 23.9 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Marjoram + Peters | 26.7 | 1 | 4 | 0 | 1 | 0.10 | +3.1% | FRAGILE favorite: 1/5 coin-flips; 79% of expected wins from swings | | Driggars + Richards | 25.5 | 0 | 5 | 0 | 5 | 0.03 | -1.3% | contender leaning on 5/5 coin-flips | | Webber + Sevy | 24.9 | 0 | 5 | 0 | 4 | 0.03 | -1.8% | contender leaning on 4/5 coin-flips | | Thomas + Stricklin | 24.8 | 0 | 5 | 0 | 4 | 0.03 | -0.2% | | | Rekenthaler + Stafford | 24.3 | 0 | 5 | 0 | 4 | 0.03 | -0.3% | | | Hock + Stephens | 23.9 | 0 | 4 | 1 | 4 | 0.05 | -0.0% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Driggars + Richards vs Marjoram + Peters | 47% | 2 | 1 | 0.47 | | Marjoram + Peters vs Webber + Sevy | 56% | 1 | 3 | 0.29 | | Driggars + Richards vs Webber + Sevy | 52% | 2 | 3 | 0.24 | Flight 8 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Nodar | Yancey | Cumming | Smith | Bull | Barbaree | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Nodar + Heslep | — | 54% | 50% | 53% | 55% | 56% | 26.0 | | Yancey + Whaley | 46% | — | 50% | 57% | 52% | 52% | 25.4 | | Cumming + Odea | 50% | 50% | — | 52% | 52% | 52% | 25.4 | | Smith + Rhea | 47% | 43% | 48% | — | 52% | 51% | 24.5 | | Bull + Singer | 45% | 48% | 48% | 48% | — | 50% | 24.4 | | Barbaree + Williford | 44% | 48% | 48% | 49% | 50% | — | 24.3 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Nodar + Heslep | 26.0 | 0 | 5 | 0 | 4 | 0.04 | +0.6% | FRAGILE favorite: 4/5 coin-flips; 100% of expected wins from swings | | Yancey + Whaley | 25.4 | 0 | 5 | 0 | 4 | 0.04 | -0.2% | contender leaning on 4/5 coin-flips | | Cumming + Odea | 25.4 | 0 | 5 | 0 | 5 | 0.01 | -0.4% | contender leaning on 5/5 coin-flips | | Smith + Rhea | 24.5 | 0 | 5 | 0 | 4 | 0.04 | -0.1% | | | Bull + Singer | 24.4 | 0 | 5 | 0 | 5 | 0.02 | -0.1% | | | Barbaree + Williford | 24.3 | 0 | 5 | 0 | 4 | 0.03 | -0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Yancey + Whaley vs Nodar + Heslep | 46% | 2 | 1 | 0.46 | | Cumming + Odea vs Nodar + Heslep | 50% | 3 | 1 | 0.33 | | Cumming + Odea vs Yancey + Whaley | 50% | 3 | 2 | 0.25 | Flight 9 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Wood | Kattookara | Gallagher | Thilmany | Oberholtze | Baggett | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Wood + Estes | — | 64% | 66% | 65% | 69% | 72% | 30.5 | | Kattookaran + Levitas | 36% | — | 52% | 52% | 51% | 56% | 24.8 | | Gallagher + Hansell | 34% | 48% | — | 52% | 54% | 59% | 24.7 | | Thilmany + Mulick | 35% | 48% | 48% | — | 49% | 53% | 23.9 | | Oberholtzer + Doyle | 31% | 49% | 46% | 51% | — | 55% | 23.8 | | Baggett + Vijay | 28% | 44% | 41% | 47% | 45% | — | 22.2 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Wood + Estes | 30.5 | 5 | 0 | 0 | 0 | 0.35 | +6.0% | ROBUST favorite: comparatively wide margins, 65% H2H vs top rivals | | Kattookaran + Levitas | 24.8 | 0 | 4 | 1 | 3 | 0.07 | -5.0% | contender leaning on 3/5 coin-flips | | Gallagher + Hansell | 24.7 | 1 | 3 | 1 | 3 | 0.09 | -3.1% | | | Thilmany + Mulick | 23.9 | 0 | 4 | 1 | 4 | 0.06 | -0.8% | | | Oberholtzer + Doyle | 23.8 | 0 | 4 | 1 | 4 | 0.07 | -0.8% | | | Baggett + Vijay | 22.2 | 0 | 3 | 2 | 2 | 0.15 | -0.4% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Kattookaran + Levitas vs Wood + Estes | 36% | 2 | 1 | 0.36 | | Kattookaran + Levitas vs Gallagher + Hansell | 52% | 2 | 3 | 0.24 | | Gallagher + Hansell vs Wood + Estes | 34% | 3 | 1 | 0.23 | Flight 10 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Fought | Shirley | Gelinas | Cherof | Schmeelk | True | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Fought + Fought | — | 51% | 56% | 60% | 58% | 60% | 27.1 | | Shirley + Aronson | 49% | — | 51% | 57% | 59% | 55% | 26.3 | | Gelinas + Chafin | 44% | 49% | — | 54% | 52% | 54% | 25.2 | | Cherof + Brocard | 40% | 43% | 46% | — | 53% | 55% | 24.2 | | Schmeelk + Stein | 42% | 41% | 48% | 47% | — | 52% | 23.8 | | True + Adams | 40% | 45% | 46% | 45% | 48% | — | 23.4 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Fought + Fought | 27.1 | 3 | 2 | 0 | 1 | 0.13 | +3.7% | ROBUST favorite: comparatively wide margins, 54% H2H vs top rivals | | Shirley + Aronson | 26.3 | 1 | 4 | 0 | 2 | 0.07 | -0.3% | contender leaning on 2/5 coin-flips | | Gelinas + Chafin | 25.2 | 0 | 5 | 0 | 4 | 0.04 | -3.4% | contender leaning on 4/5 coin-flips | | Cherof + Brocard | 24.2 | 0 | 4 | 1 | 3 | 0.08 | -0.3% | | | Schmeelk + Stein | 23.8 | 0 | 3 | 2 | 3 | 0.07 | -0.1% | | | True + Adams | 23.4 | 0 | 4 | 1 | 3 | 0.07 | -0.3% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Fought + Fought vs Shirley + Aronson | 51% | 1 | 2 | 0.49 | | Fought + Fought vs Gelinas + Chafin | 56% | 1 | 3 | 0.29 | | Shirley + Aronson vs Gelinas + Chafin | 51% | 2 | 3 | 0.25 | Flight 11 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Richardson | Robinson | Taylor | Perez | McLaughlin | Story | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Richardson + Loricchio | — | 55% | 50% | 56% | 51% | 53% | 25.9 | | Robinson + McCormack | 45% | — | 53% | 50% | 53% | 53% | 25.1 | | Taylor + Simons | 50% | 47% | — | 52% | 50% | 50% | 25.0 | | Perez + Marold | 44% | 50% | 48% | — | 52% | 52% | 24.7 | | McLaughlin + Mclaughlin | 49% | 47% | 50% | 48% | — | 50% | 24.7 | | Story + Chase | 47% | 47% | 50% | 48% | 50% | — | 24.5 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Richardson + Loricchio | 25.9 | 0 | 5 | 0 | 3 | 0.04 | +0.8% | FRAGILE favorite: 3/5 coin-flips; 100% of expected wins from swings | | Robinson + McCormack | 25.1 | 0 | 5 | 0 | 4 | 0.03 | -0.2% | contender leaning on 4/5 coin-flips | | Taylor + Simons | 25.0 | 0 | 5 | 0 | 5 | 0.01 | -0.6% | contender leaning on 5/5 coin-flips | | Perez + Marold | 24.7 | 0 | 5 | 0 | 4 | 0.03 | -0.1% | | | McLaughlin + Mclaughlin | 24.7 | 0 | 5 | 0 | 5 | 0.01 | -0.1% | | | Story + Chase | 24.5 | 0 | 5 | 0 | 5 | 0.02 | -0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Richardson + Loricchio vs Robinson + McCormack | 55% | 1 | 2 | 0.45 | | Richardson + Loricchio vs Taylor + Simons | 50% | 1 | 3 | 0.33 | | Robinson + McCormack vs Taylor + Simons | 52% | 2 | 3 | 0.24 | Flight 12 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Wright | Collins | Reynolds | Benson | Dudley | Beaver | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Wright + Gadsby | — | 52% | 56% | 54% | 58% | 62% | 26.8 | | Collins + Minter | 48% | — | 53% | 53% | 57% | 55% | 25.9 | | Reynolds + Swart | 44% | 47% | — | 55% | 59% | 57% | 25.6 | | Benson + Brosnahan | 46% | 47% | 45% | — | 54% | 54% | 24.7 | | Dudley + Cohen | 42% | 43% | 41% | 46% | — | 52% | 23.5 | | Beaver + Rhyne | 38% | 45% | 43% | 46% | 48% | — | 23.4 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Wright + Gadsby | 26.8 | 2 | 3 | 0 | 2 | 0.10 | +2.8% | favorite in a coin-flip flight; lead is soft | | Collins + Minter | 25.9 | 0 | 5 | 0 | 3 | 0.05 | -2.4% | contender leaning on 3/5 coin-flips | | Reynolds + Swart | 25.6 | 1 | 4 | 0 | 2 | 0.09 | -0.4% | contender leaning on 2/5 coin-flips | | Benson + Brosnahan | 24.7 | 0 | 5 | 0 | 5 | 0.04 | -0.5% | | | Dudley + Cohen | 23.5 | 0 | 3 | 2 | 2 | 0.09 | +0.0% | | | Beaver + Rhyne | 23.4 | 0 | 4 | 1 | 2 | 0.09 | -0.2% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Collins + Minter vs Wright + Gadsby | 48% | 2 | 1 | 0.48 | | Wright + Gadsby vs Reynolds + Swart | 56% | 1 | 3 | 0.30 | | Collins + Minter vs Reynolds + Swart | 53% | 2 | 3 | 0.24 | Flight 13 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Christian | Hancher | DiCicco | Panessa | Vazquez | Bartlett | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Christian + Springer | — | 51% | 52% | 54% | 57% | 58% | 26.3 | | Hancher + Helms | 49% | — | 50% | 53% | 52% | 55% | 25.5 | | DiCicco + Kulik | 48% | 50% | — | 51% | 52% | 55% | 25.4 | | Panessa + Ratliff | 46% | 47% | 49% | — | 56% | 54% | 25.1 | | Vazquez + Prokupek | 43% | 48% | 48% | 44% | — | 49% | 23.9 | | Bartlett + Florak | 42% | 45% | 45% | 46% | 51% | — | 23.8 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Christian + Springer | 26.3 | 1 | 4 | 0 | 3 | 0.06 | +2.1% | FRAGILE favorite: 3/5 coin-flips; 79% of expected wins from swings | | Hancher + Helms | 25.5 | 0 | 5 | 0 | 5 | 0.02 | -1.0% | contender leaning on 5/5 coin-flips | | DiCicco + Kulik | 25.4 | 0 | 5 | 0 | 4 | 0.03 | -1.0% | contender leaning on 4/5 coin-flips | | Panessa + Ratliff | 25.1 | 0 | 5 | 0 | 4 | 0.04 | -0.1% | | | Vazquez + Prokupek | 23.9 | 0 | 5 | 0 | 3 | 0.05 | -0.1% | | | Bartlett + Florak | 23.8 | 0 | 4 | 1 | 3 | 0.07 | +0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Hancher + Helms vs Christian + Springer | 49% | 2 | 1 | 0.49 | | DiCicco + Kulik vs Christian + Springer | 48% | 3 | 1 | 0.32 | | DiCicco + Kulik vs Hancher + Helms | 50% | 3 | 2 | 0.25 | Flight 14 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Kessler | Dalmau | Higdon | Rosenthal | Friedenber | Goodrick | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Kessler + Brinson | — | 57% | 57% | 59% | 62% | 57% | 27.4 | | Dalmau + Pons | 43% | — | 50% | 51% | 55% | 56% | 25.3 | | Higdon + Doyle | 43% | 50% | — | 52% | 52% | 51% | 24.9 | | Rosenthal + Rosenthal | 41% | 49% | 48% | — | 53% | 55% | 24.8 | | Friedenberg + Miller | 38% | 45% | 48% | 47% | — | 53% | 23.9 | | Goodrick + Corry | 43% | 44% | 49% | 45% | 47% | — | 23.7 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Kessler + Brinson | 27.4 | 2 | 3 | 0 | 0 | 0.17 | +5.5% | ROBUST favorite: comparatively wide margins, 57% H2H vs top rivals | | Dalmau + Pons | 25.3 | 0 | 5 | 0 | 3 | 0.05 | -2.5% | contender leaning on 3/5 coin-flips | | Higdon + Doyle | 24.9 | 0 | 5 | 0 | 4 | 0.03 | -3.1% | contender leaning on 4/5 coin-flips | | Rosenthal + Rosenthal | 24.8 | 0 | 4 | 1 | 3 | 0.06 | -0.2% | | | Friedenberg + Miller | 23.9 | 0 | 4 | 1 | 4 | 0.06 | -0.2% | | | Goodrick + Corry | 23.7 | 0 | 5 | 0 | 2 | 0.07 | -0.3% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Dalmau + Pons vs Kessler + Brinson | 43% | 2 | 1 | 0.43 | | Higdon + Doyle vs Kessler + Brinson | 44% | 3 | 1 | 0.29 | | Dalmau + Pons vs Higdon + Doyle | 50% | 2 | 3 | 0.25 | Flight 15 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Perry | Jaillet | Alexander | Smith | Riley | Pickell | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Perry + Calobrisi | — | 51% | 57% | 54% | 56% | 59% | 26.6 | | Jaillet + Jaillet | 49% | — | 49% | 50% | 56% | 51% | 25.3 | | Alexander + Lambeth | 43% | 51% | — | 52% | 50% | 52% | 24.9 | | Smith + Warden | 46% | 50% | 48% | — | 50% | 51% | 24.6 | | Riley + Ashton | 44% | 44% | 50% | 50% | — | 53% | 24.5 | | Pickell + Ames | 41% | 49% | 48% | 49% | 47% | — | 24.1 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Perry + Calobrisi | 26.6 | 1 | 4 | 0 | 2 | 0.09 | +2.9% | FRAGILE favorite: 2/5 coin-flips; 79% of expected wins from swings | | Jaillet + Jaillet | 25.3 | 0 | 5 | 0 | 4 | 0.02 | -1.4% | contender leaning on 4/5 coin-flips | | Alexander + Lambeth | 24.9 | 0 | 5 | 0 | 4 | 0.03 | -1.5% | contender leaning on 4/5 coin-flips | | Smith + Warden | 24.6 | 0 | 5 | 0 | 5 | 0.01 | -0.3% | | | Riley + Ashton | 24.5 | 0 | 5 | 0 | 3 | 0.04 | -0.2% | | | Pickell + Ames | 24.1 | 0 | 4 | 1 | 4 | 0.04 | -0.0% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Perry + Calobrisi vs Jaillet + Jaillet | 51% | 1 | 2 | 0.49 | | Perry + Calobrisi vs Alexander + Lambeth | 57% | 1 | 3 | 0.29 | | Jaillet + Jaillet vs Alexander + Lambeth | 49% | 2 | 3 | 0.24 | Flight 16 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Wilson | Wagner | McBride | Flammia | Bhatia | Goodloe | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Wilson + Slavis | — | 51% | 53% | 52% | 53% | 60% | 26.2 | | Wagner + Haswell | 49% | — | 52% | 54% | 52% | 60% | 26.0 | | McBride + Hall | 47% | 48% | — | 51% | 51% | 58% | 25.3 | | Flammia + Gray | 48% | 46% | 49% | — | 53% | 56% | 25.1 | | Bhatia + Tear IV | 47% | 48% | 49% | 47% | — | 53% | 24.6 | | Goodloe + Santivanez | 40% | 40% | 42% | 44% | 47% | — | 22.9 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Wilson + Slavis | 26.2 | 1 | 4 | 0 | 4 | 0.05 | +0.5% | FRAGILE favorite: 4/5 coin-flips; 78% of expected wins from swings | | Wagner + Haswell | 26.0 | 1 | 4 | 0 | 4 | 0.04 | +0.1% | contender leaning on 4/5 coin-flips | | McBride + Hall | 25.3 | 1 | 4 | 0 | 4 | 0.04 | -0.6% | contender leaning on 4/5 coin-flips | | Flammia + Gray | 25.1 | 0 | 5 | 0 | 4 | 0.04 | -0.4% | | | Bhatia + Tear IV | 24.6 | 0 | 5 | 0 | 5 | 0.03 | -0.4% | | | Goodloe + Santivanez | 22.9 | 0 | 2 | 3 | 1 | 0.13 | +0.4% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Wagner + Haswell vs Wilson + Slavis | 49% | 2 | 1 | 0.49 | | McBride + Hall vs Wilson + Slavis | 47% | 3 | 1 | 0.31 | | McBride + Hall vs Wagner + Haswell | 48% | 3 | 2 | 0.24 | Flight 17 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Poje | Brenner | Lal | Feldman | Bowles | Atkins | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Poje + Burleson | — | 51% | 57% | 54% | 56% | 59% | 26.6 | | Brenner + McCormick | 49% | — | 54% | 53% | 54% | 56% | 25.9 | | Lal + Grover | 43% | 46% | — | 52% | 52% | 51% | 24.6 | | Feldman + Waronker | 46% | 47% | 48% | — | 50% | 50% | 24.5 | | Bowles + Troisi | 44% | 46% | 48% | 50% | — | 50% | 24.3 | | Atkins + Single | 41% | 44% | 49% | 50% | 50% | — | 24.1 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Poje + Burleson | 26.6 | 1 | 4 | 0 | 2 | 0.09 | +2.7% | FRAGILE favorite: 2/5 coin-flips; 79% of expected wins from swings | | Brenner + McCormick | 25.9 | 0 | 5 | 0 | 4 | 0.04 | -0.6% | contender leaning on 4/5 coin-flips | | Lal + Grover | 24.6 | 0 | 5 | 0 | 4 | 0.04 | -2.0% | contender leaning on 4/5 coin-flips | | Feldman + Waronker | 24.5 | 0 | 5 | 0 | 5 | 0.02 | -0.3% | | | Bowles + Troisi | 24.3 | 0 | 5 | 0 | 4 | 0.03 | -0.3% | | | Atkins + Single | 24.1 | 0 | 4 | 1 | 3 | 0.05 | -0.1% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Brenner + McCormick vs Poje + Burleson | 48% | 2 | 1 | 0.49 | | Lal + Grover vs Poje + Burleson | 44% | 3 | 1 | 0.29 | | Feldman + Waronker vs Poje + Burleson | 46% | 4 | 1 | 0.23 | Flight 18 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Foley | Gutzman | Seeley | Muldoon | Bhatia | Hoard | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Foley + Hampy | — | 51% | 53% | 54% | 55% | 61% | 26.4 | | Gutzman + Slutzky | 49% | — | 51% | 53% | 53% | 58% | 25.8 | | Seeley + Seeley | 47% | 49% | — | 52% | 52% | 58% | 25.4 | | Muldoon + Baribeau | 46% | 47% | 48% | — | 50% | 56% | 24.9 | | Bhatia + Tear | 45% | 47% | 48% | 50% | — | 56% | 24.8 | | Hoard + Woods | 39% | 42% | 42% | 44% | 44% | — | 22.6 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Foley + Hampy | 26.4 | 1 | 4 | 0 | 4 | 0.06 | +0.7% | FRAGILE favorite: 4/5 coin-flips; 78% of expected wins from swings | | Gutzman + Slutzky | 25.8 | 1 | 4 | 0 | 4 | 0.04 | -0.2% | contender leaning on 4/5 coin-flips | | Seeley + Seeley | 25.4 | 1 | 4 | 0 | 4 | 0.04 | -0.5% | contender leaning on 4/5 coin-flips | | Muldoon + Baribeau | 24.9 | 0 | 5 | 0 | 4 | 0.04 | -0.4% | | | Bhatia + Tear | 24.8 | 0 | 5 | 0 | 4 | 0.04 | -0.4% | | | Hoard + Woods | 22.6 | 0 | 2 | 3 | 0 | 0.16 | +0.4% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Foley + Hampy vs Gutzman + Slutzky | 51% | 1 | 2 | 0.49 | | Seeley + Seeley vs Foley + Hampy | 47% | 3 | 1 | 0.31 | | Seeley + Seeley vs Gutzman + Slutzky | 49% | 3 | 2 | 0.24 | Flight 19 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Toole | Evans | Harte | Lee | Truitt | Vaniman | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Toole + Calvert | — | 54% | 55% | 56% | 56% | 60% | 26.9 | | Evans + Rowley | 46% | — | 51% | 49% | 56% | 56% | 25.5 | | Harte + Love | 45% | 49% | — | 51% | 56% | 55% | 25.3 | | Lee + Thompson | 44% | 51% | 49% | — | 54% | 53% | 25.1 | | Truitt + Fetter | 44% | 44% | 44% | 46% | — | 54% | 23.9 | | Vaniman + Leingang | 40% | 44% | 45% | 47% | 46% | — | 23.3 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Toole + Calvert | 26.9 | 1 | 4 | 0 | 1 | 0.11 | +3.0% | ROBUST favorite: comparatively wide margins, 55% H2H vs top rivals | | Evans + Rowley | 25.5 | 0 | 5 | 0 | 3 | 0.05 | -1.4% | contender leaning on 3/5 coin-flips | | Harte + Love | 25.3 | 0 | 5 | 0 | 3 | 0.05 | -1.7% | contender leaning on 3/5 coin-flips | | Lee + Thompson | 25.1 | 0 | 5 | 0 | 4 | 0.04 | -0.3% | | | Truitt + Fetter | 23.9 | 0 | 5 | 0 | 2 | 0.08 | -0.2% | | | Vaniman + Leingang | 23.3 | 0 | 4 | 1 | 3 | 0.08 | -0.0% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Toole + Calvert vs Evans + Rowley | 54% | 1 | 2 | 0.46 | | Toole + Calvert vs Harte + Love | 55% | 1 | 3 | 0.30 | | Evans + Rowley vs Harte + Love | 51% | 2 | 3 | 0.24 | Flight 20 Matchup matrix — P(row beats column): | Team (E[pts] rank) | Barnes | Pilger | Ballard | Bell | Morge | King | E[pts] | |---|---:|---:|---:|---:|---:|---:|---:| | Barnes + Smelcer | — | 58% | 66% | 68% | 78% | 78% | 31.2 | | Pilger + Hatz | 42% | — | 59% | 62% | 73% | 75% | 28.7 | | Ballard + Armstrong | 34% | 41% | — | 58% | 65% | 68% | 26.0 | | Bell + Harren | 32% | 38% | 42% | — | 58% | 59% | 23.6 | | Morge + Callahan | 22% | 27% | 35% | 42% | — | 55% | 20.7 | | King + Scales | 22% | 25% | 32% | 41% | 45% | — | 19.6 | Where the points come from: | Team | E[pts] | Strong wins | Swing | Likely losses | Coin-flips | Robustness | Adj | Note | |---|---:|---:|---:|---:|---:|---:|---:|---| | Barnes + Smelcer | 31.2 | 5 | 0 | 0 | 0 | 0.39 | +3.8% | ROBUST favorite: comparatively wide margins, 62% H2H vs top rivals | | Pilger + Hatz | 28.7 | 4 | 0 | 1 | 0 | 0.31 | -0.2% | contender with comparatively robust margins | | Ballard + Armstrong | 26.0 | 3 | 0 | 2 | 0 | 0.27 | -3.5% | contender with comparatively robust margins | | Bell + Harren | 23.6 | 2 | 0 | 3 | 0 | 0.22 | -0.7% | | | Morge + Callahan | 20.7 | 0 | 1 | 4 | 1 | 0.28 | -0.7% | | | King + Scales | 19.6 | 0 | 1 | 4 | 1 | 0.31 | -0.7% | | Swing matches that most decide the flight (closeness × contention): | Matchup | P(A beats B) | A rank | B rank | Leverage | |---|---:|---:|---:|---:| | Pilger + Hatz vs Barnes + Smelcer | 42% | 2 | 1 | 0.42 | | Ballard + Armstrong vs Barnes + Smelcer | 34% | 3 | 1 | 0.23 | | Pilger + Hatz vs Ballard + Armstrong | 59% | 2 | 3 | 0.20 | ============================================================================ # Track Records (2021-2025) Source URL: https://calcutta.high.green/reports/history ============================================================================ Member-Member Historical Track Record — Findings (2021–2025) Built 2026-06-09. Sources: 2021/2022/2023 JPEGs, 2024 PDF (Flight resultsMM2024.pdf`), and the pre-extracted results2025.csv`. Outputs: - data/raw/history/results2021.csv … results2025.csv (per year) - data/raw/history/resultsall.csv` (unified long table, 466 rows) - analysis/history/playertrackrecord.csv (300 surname-players) - analysis/history/team2026trackrecord.csv` (120 current teams + score) --- 1. Per-year extraction notes | Year | Source | Flights | Format | Confidence | |------|--------|---------|--------|------------| | 2021 | JPEG (table + photos) | 15 (6 teams each) | "Team" col is pairing order (e.g. 1-1), finish = Total points | Good; several low-res cells re-cropped & verified | | 2022 | JPEG | 15 | M1/M2/M3/Day1/M4/M5/Day2/Total; finish by Total | Good | | 2023 | JPEG | 15 | same as 2022 | Good | | 2024 | PDF, 3 pages | 18 (6 each) | Pos. explicit, R1–R5 + Points; ties shown T1/T5 | High (rendered 400 dpi) | | 2025 | pre-extracted | 15 | year,flight,pos,team,points | given | Position convention. For 2021–2023 the printed row order is the pairing/seed order, NOT the finish. pos was derived by ranking the Total column descending within each flight (ties share the better position, next skipped — matching 2025 and the 2024 PDF T1/T5 style). 2024 and 2025 carry an explicit finish position. Field size grew: 15 flights x 6 = 90 teams in 2021–2023 and 2025; 18 flights (108 teams) in 2024. Flight numbering runs low = strongest (lowest handicap). Because field size/flight count differ by year, the player table normalizes finish to finish percentile = (field − pos + 1)/field × 100, so a win ~100 regardless of year. 2025 anomaly (resolved): Vaniman + Leingang and Pilger + Bhatia each appeared in BOTH flight 2 and flight 15 of results_2025.csv. No other year duplicates either team. Vaniman + Leingang are a high-handicap pairing (flight 19 in 2026), so a flight-2 (lowest handicap) placement is implausible; the flight-15 rows are genuine. The two flight-2 rows are dropped when building resultsall.csv (logged: "2 anomaly rows dropped"). The per-year results2025.csv was left untouched (input, not ours to edit). Low-confidence / flagged transcriptions (re-cropped and double-checked; residual risk noted): - 2021 fl3: Lennon + Lennon 15.5 and Kulik + Kulik 26.5 (initially mis-read 25/19; corrected on tight crop). 2021 fl3-5 = Pearson + Walsey (won, 31.5). - 2021 fl14: row1 Fetter + Truitt 22.0, row2 Perry + Mattice 28.5 (different Perry/Fetter pairings than later years). - 2024 fl17 row2 McDaniel + Granger = 30.0 (one early read said 27.5; 30.0 confirmed on zoom). - 2024: DNF'd round 5 ("11:30 AM, Hole 9/5", 0.00 pts): Kraemer+Kraemer, Whaley+Yancey (fl11), Seeley+Fought, Mangine+Hall (fl15). Totals real but depressed by the missing round. - Spellings kept as printed: Muenchen+Muenchen (2024), Forbidussi, Coirini, Goerlich. Nothing was illegible enough to require a blank/guess; everything in the CSVs was read. --- 2. Biggest caveat — surname collisions Historical sheets give surnames only. The player track record keys on surname and pools everyone who shares it. In the 2026 field alone, 21 surnames map to ≥2 distinct people, including high-value names: - Wright -> Adam Wright (2026 fl5, w/ Copeland) vs Kyle Wright. Historical "Wright" pools Copeland's partner AND Wright+Estes (2024 fl7) AND Wright+Stroman/Goodrick. - Smith -> Don vs Matt Smith (history pools Florak/Rhea/Saccone partners). - Doyle -> Matthew (Oberholtzer's partner) vs Mason (Higdon's partner). - Also: Brown, Perry, Benson, Bhatia, Green, Leprohon, Pilger, Shirley, Taylor, Vaniman, Walsey, Jaillet, Hatz, Rosenthal, Seeley, Tear, Fought, McLaughlin. team2026trackrecord.csv has a nameambiguous column listing any ambiguous surname on a 2026 team. Treat those teams' scores as an upper bound. Pairs with stable partner names across years (Stafford+Rekenthaler, Barbaree+Williford, Downey+Shearer, Bachstein+Hancock, Ballard+Armstrong) are high-confidence. --- 3. Consistent OVERPERFORMERS (results-based ability signal) Ranked by flight wins, then top-2, then avg finish pos (5-yr window). "Move" = first-year flight number minus last-year; positive = climbed to a stronger (lower-numbered) flight = under-rated / sandbag signal. Tier 1 — repeat winners climbing into tougher flights (BUY / sandbagger signals): - Perry+Dye (Ryan Perry / Wes Dye, 2026 fl1): Perry 3 wins / 5 T2 over 5 yrs; pair won fl1 in 2021, 2024, 2025; promoted fl7->fl1. Caveat: "Perry" ambiguous; the Dye+Perry pair is genuine. - Walsey (Dustin, fl2 2026): climbed fl3->fl1, won fl1 2023/2024/2025 (Kulik+Walsey). - Bachstein + Hancock (fl5): together 5 straight yrs, avg pos 1.6, fl1 in 2023. Clean names, high confidence. Under-rated overperformer. - Copeland (+Wright) (fl5): won fl4 2021 & 2024, won fl5 2025, fl4 2nd 2022 — avg pos 1.75. Corroborates sandbagger intel. - Levin+Oxman (note: 2026 David Levin partners Leaf, not Oxman): pair won fl4-6 thrice (2021–2024), avg pos 1.75 — only partial transfer to 2026. - LeProhon+LeProhon (Billy/Scott, fl5): pooled = 4 wins / 6 T2 but conflates brothers (2022 fl9 2nd, 2023 fl8 win, 2024 fl11 win). Strong family; per-person split ambiguous. Tier 2 — proven, stable-flight winners: - Downey + Shearer (fl2): together 5 yrs, fl2 throughout, won 2022 & 2025, 2nd 2021/2023. High-confidence buy at par. - Hartz + Fann (fl2): 4 yrs, fl3-4, 2 wins, slight climb. - Pokorny + Huban (not a 2026 pair): 5 yrs, 2 wins (fl7), avg pos 2.4. - Ballard + Armstrong (fl20): 5 yrs in fl15/18 (weak handicap flights), 2 wins incl. highest score in all of 2025 (34.5). Wins reliably but small pot. Clean names. - Barbaree + Williford (fl8): won fl9 2022 & 2024 (32.0 both), 3rd 2021 — then last in fl7 2025 (18.0) after a move up. See intel. Chronic UNDERPERFORMERS / fades: - Stafford + Rekenthaler (fl7): 5 yrs, 0 wins, avg pos 4.2, finish pct 47; only cash = 2025 2nd. Results say fade; intel says buy (genuine disagreement — see §5). - Smith (Don/Matt pooled): 0 wins/5 yrs, avg pos 3.7. - Gallagher (fl9 w/ Hansell): 4 yrs, 1 win, avg pos 4.0. Matches "fade Gallagher+Hansell." - Higdon (fl14 w/ Doyle): 1 appearance, last (fl12 2025). No record. --- 4. 2026 teams with strongest track records (results-based BUY list) From team2026trackrecord.csv (top combined trackrecordscore; = name_ambiguous): | Rank | 2026 Team | Flt | Score | W/T2 | Note | |---|---|---|---|---|---| | 1 | Leprohon + Leprohon | 5 | 18.7 | 8/12 | * pooled brothers — split unclear | | 2 | Wilson + Slavis | 16 | 16.3 | 4/6 | * Wilson; Wilson+Slavis was a real 2022/2025 pair | | 3 | Wright + Copeland | 5 | 15.8 | 4/7 | * Wright; Copeland half is the real sandbag signal | | 4 | Brown + Brown | 3 | 15.5 | 0/8 | ** heavily pooled (4 Browns) — discount hard | | 5 | Perry + Dye | 1 | 14.7 | 5/8 | * Perry; Dye+Perry fl1 dynasty is genuine | | 6 | Oberholtzer + Doyle | 9 | 14.5 | 4/6 | * Doyle; won fl9 2025 (intel = FADE on price) | | 7 | Shirley + Aronson | 10 | 13.7 | 2/7 | * Shirley pooled | | 8 | Ballard + Armstrong | 20 | 12.9 | 4/6 | clean names, smallest pot | | 11 | Downey + Shearer | 2 | 12.6 | 4/8 | clean names, very consistent — best confidence/value | | 12 | Hancock + Bachstein | 5 | 12.5 | 3/9 | clean names, 9 top-2s in 5 yrs — strong | | — | Knapp + Keister | 6 | 11.1 | 4/4 | both halves won recently | | — | Panessa + Ratliff | 13 | 10.8 | 4/4 | * — won fl11 2023 | Cleanest high-confidence results buys (unambiguous names + repeat top-2): Downey + Shearer (fl2) and Hancock/Bachstein (fl5); then Ballard + Armstrong (fl20) and Knapp + Keister (fl6). The brothers/duplicate-surname teams (Leprohon, Brown+Brown, Green+Green, Walsey+Walsey) score high but carry pooling risk. --- 5. Cross-reference vs manual intel & the model | Intel claim | Results say | Verdict | |---|---|---| | Williford strong sandbagger -> BUY | Barbaree+Williford WON fl9 2022 & 2024 (32 both); last in fl7 2025. | Corroborated as proven ability; 2025 collapse adds caution (may have been re-rated). | | Copeland sandbagger, hothead -> BUY w/ caveat | Wright+Copeland won fl4 2021 & 2024, fl5 2025; avg pos 1.75. | Corroborated — consistent winner across changing flights. | | Rekenthaler+Stafford proven, 2nd fl6 2025 -> BUY | 5 yrs, 0 wins, avg pos 4.2; only cash 2025 2nd. | Partially CONTRADICTED. Multi-year record below average. Results lean FADE; intel BUY. Coin-flip, not a lock. | | Goodloe won "the whole thing ~2 yrs ago" -> proven (small pot) | Milhous+Goodloe WON fl10 2021 & fl12 2022; middling since (28/23/28). | Corroborated as flight winner (w/ Milhous, not 2026 partner Santivanez). Recent form only middling. | | The Doyles -> AVOID both | Oberholtzer+Doyle won fl9 2025, avg pos 2.2 (good). Higdon+Doyle: Higdon last in only appearance. | Corroborated as a PRICE call. Oberholtzer+Doyle genuinely good; Higdon+Doyle easy avoid. Two different Doyles (Matthew vs Mason). | | Matt Smith (Smith+Warden fl15) hot | "Smith" pooled: 0 wins, avg pos 3.7; can't isolate Matt. | Not corroborated (and not contradicted — pooling hides him). Trust the human read; data silent. | | Gallagher+Hansell -> fade; take Wood+Estes | Gallagher 1 win/4 yrs, avg pos 4.0. Wood+Estes won fl7 2025 (Estes also fl7 2024). | Corroborated. | | 2025 validation winners (Wood+Estes, Downey+Shearer, Gelinas+Chafin, Wright+Copeland, Armstrong+Ballard) | All confirmed in results_all.csv. | All corroborated. | Where results AGREE with model + intel: Downey+Shearer, Bachstein/Hancock, Ballard+Armstrong, Wood+Estes, Copeland. Where results DISAGREE / add nuance: 1. Rekenthaler+Stafford — intel BUY vs 5-yr record (0 wins, avg pos 4.2) FADE; the model also fades (−$158). Results side with the model against the human intel. Low-confidence flier. 2. Williford — proven winner (intel right) but 2025 last-place hints the handicap caught up; don't pay full "flight favorite" price blind. 3. Oberholtzer+Doyle — genuinely good (won fl9 2025), so the "avoid" is a price judgment, not ability. If the price isn't actually inflated, they're a real team. 4. Surname pooling inflates several headline scores (Brown+Brown, Leprohon, Wright). The full-name GHIN model is MORE reliable than surname-only history for those teams. --- Method notes / assumptions - trackrecordscore = Σ over a player's results of weight × (finishpct + winbonus)/100, where recent years (2024–2025) weight 2×, older 1×; win_bonus = 25 for a win, 10 for top-2. Team score = sum of both players' scores (credits two strong individuals even if never partnered). It rewards volume of strong finishes, so 5-yr veterans outrank one-hit wonders — read alongside avg finish pos / win count, not in isolation. - Finish percentile normalizes across 15- vs 18-flight years. - Ties: shared position (T1/T5) per source convention. - Scripts: analysis/history/buildyearcsvs.py and analysis/history/build_analysis.py. Re-runnable. ============================================================================ # Can We Win? - Portfolio Backtest Source URL: https://calcutta.high.green/reports/portfolio_backtest ============================================================================ Calcutta Portfolio Backtest Would any betting strategy have made money? Historical backtest of the 2024 & 2025 auctions · generated 2026-06-10 (US Eastern) Bottom line: No price-rule strategy made money with statistical confidence. A few posted positive two-year averages (mid-price +4.0% , buy-cheapest +1.3% ), but every CI runs from roughly −55% to +73% and the two years flip signs. The standout — human-intel/sandbaggers +77% , 71% hit — rests on just 7 picks (CI ≈ [−45%, +192%]) and two flight wins: a lead to track , not a proven edge. Expected ROI of a disciplined one-team-per-flight price rule is near zero; realistic single-year range −50% to +30%. The one robust, repeatable finding is a don't : buying cheap longshots in the richest flights lost −78% (−100% in 2025 — zero of six cashed). -10.9% Random portfolio mean ROI 32% Random portfolios that profited 100% Name match rate (both years) Verified mechanics - Payout structure. Within each flight, 70% of the payout pool goes to the winner and 30% to the runner-up; 10% of the pot is directed to the shootout. Per-flight payouts reconcile to 0.9×pot to the dollar. - Tie-aware payouts. Winner ties (9 flights) pool 1st+2nd money split evenly; runner-up ties (5 flights) split 2nd money. Handled explicitly — they materially move payouts. - Match rate 100% (108/108 in 2024, 90/90 in 2025) via surname-pair matching with a fuzzy fallback (caught Shirely→Shirley). Strategy battery — combined 2024 + 2025 Strategy 2024 2025 Combined 95% CI (bootstrap) Hit rate P&L /$10k Buy Human-Intel / Sandbaggers +10.0% +126.7% +76.7% [-45.0%, +192.1%] +70.8% $7,666 Rich-6: Favorite -41.4% +63.0% +10.8% [-61.6%, +99.4%] +66.7% $1,084 Buy Lowest Handicap — +10.4% +10.4% [-67.6%, +107.9%] +33.3% $1,042 Buy Mid-Price +6.0% +1.6% +4.0% [-44.7%, +57.0%] +36.7% $399 Buy Cheapest -13.0% +18.3% +1.3% [-56.6%, +71.0%] +27.8% $127 Buy Value (price/pot·6) -13.0% +18.3% +1.3% [-56.4%, +72.7%] +27.8% $127 Buy-All (baseline) -10.0% -10.0% -10.0% [-30.6%, +9.6%] +33.3% $-1,000 Buy Top Seed -31.6% +10.4% -12.5% [-60.8%, +54.5%] +30.6% $-1,249 Buy Favorite (priciest) -37.2% +12.2% -14.7% [-55.4%, +40.0%] +48.9% $-1,471 Buy Value (seed-implied) -44.7% -3.7% -26.0% [-71.7%, +42.5%] +24.4% $-2,605 Rich-6: Cheapest -55.5% -100.0% -77.8% [-100.0%, -11.0%] +8.3% $-7,775 Rich-6: Value -55.5% -100.0% -77.8% [-100.0%, -11.0%] +8.3% $-7,775 One team per flight unless noted. "Combined" = equal-dollar across all 33 picks pooled over both years. CI = across-year block bootstrap (20,000 resamples) — deliberately wide to reflect that we have only two years. Buy-all shown price-weighted; Rich-6 strategies use the 6 highest-pot flights (our 2026 slate logic applied retroactively). Monte-Carlo: random portfolios Drawing one random team per flight, 50,000 times, the combined mean ROI is -10.9% , with a 95% range of [-56%, +40%] . A random portfolio turned a profit 32% of the time. Roughly a third of random portfolios made money, so a two-year positive result from any named strategy is not evidence of an edge. 2026 slate logic, applied retroactively - Rich-6 Cheapest (cheapest team in the 6 richest flights): −55% (2024), −100% (2025), −78% combined. - Rich-6 Favorite (favorite in the 6 richest flights): −41% then +63%, +11% combined — best hit rate among rich-6 plays; the swing across years reflects variance, not skill. - Human-intel / sandbaggers (Williford, Copeland, Wood+Estes, Rekenthaler+Stafford): +10% then +127%, +77% combined , 71% hit rate — the best result in the battery. But 7 picks over two years, dominated by two flight wins (Williford 2024, Copeland & Wood+Estes 2025); CI ≈ [−45%, +192%]. Williford flipped from flight winner (2024) to 5th (2025) — exactly the sandbagger volatility the intel doc flags. Verdict for the partner - Expected ROI is near zero. Best point estimates for the stronger strategies are 0% to +5%, inside the noise. - Realistic single-year range: −50% to +30%. - No edge is statistically established. Every positive strategy's CI includes large losses. - One robust rule, and it's a "don't": never buy cheap longshots in rich flights. - Watch, don't bank: human intel/sandbaggers (+77% on 7 picks) is the only signal worth tracking forward, not yet bankable on two years of data. Data limitations - Prices exist for 2024 & 2025 only — full P&L is a two-year backtest. 2021–2023 are results-only context. - Tiny sample: 33 one-team-per-flight picks per strategy. Bands are wide by necessity. - Surname matching across files (100% here, but rests on within-year uniqueness). - 'Value = price/(pot/6)' is mathematically identical to buy-cheapest (pot/6 is constant within a flight); we added a seed-implied fair-value variant as a genuine value test. Scripts: backtest.py , run_strategies.py , make_report.py . Data: strategies_pnl.csv , panel.csv . Read-only on all source data. ============================================================================ # Forward P&L Simulation Source URL: https://calcutta.high.green/reports/portfolio_montecarlo ============================================================================ 2026 Calcutta — Portfolio P&L Monte-Carlo 50,000-draw forward simulation of Wednesday's auction. If we ran it thousands of times, what is our profit/loss distribution — and how much should we risk? +28% Best feasible slate E[ROI] (fav5, sharp probs) −4 to −14% E[ROI] if probabilities are uniform (no edge) ~$1,500 Recommended stake (~60% of $2,500 budget) 121% Budget used by the rec6 slate — over budget Executive summary / verdict - A positive-EV slate exists — but only if the model's sharp probabilities are roughly right, and only at a sane bet size. Sharp p_win : every slate +18% to +37% E[ROI]. Shrunk: +8% to +19%. Uniform 1/6: every slate loses (−4% to −14%). - The recommended 6-team slate does not fit the $2,500 budget. Half-back cost ~$3,021 ( 121% ); adverse selection ~$4,293 ( 172% ). rec5/rec8 also over. Feasible attractive slates: sand3, cheap5, fav5. - Recommended: a small value/favourite-tilted 3–5 team slate, ~$1,300–$1,900 (50–75% of budget), holding dry powder. fav5 = best single slate; sand3 = top ROI but feast-or-famine; cheap5 = lowest downside (VaR ~$630). - Risk of ruin is non-trivial: P(lose >50% of staked capital) is 15–25% (sharp/shrunk), 25–44% (uniform). Size it like a high-variance bet. Bottom line: No slate is positive-EV under the pessimistic (uniform) probabilities. If the model is even partly sharp, a 3–5 team value slate at ~$1,500 (~60% of budget) is the risk-adjusted sweet spot: positive expected P&L, lowest dollar downside, and budget held in reserve in case the model is wrong. P&L distribution by probability scenario [figure: P&L distribution by scenario] rec6 slate. The sharp distribution (navy) sits clearly right of zero; shrinking the probabilities pulls the mean toward zero; under uniform (red) the mean goes negative. Headline results — naive buyback, est-price E[ROI] is on capital actually deployed. VaR$ = 5th-percentile P&L (95% one-sided VaR). ⚠ = mean capital deployed exceeds the $2,500 budget. Slate Scenario E[ROI] E[P&L] P(profit) Median VaR$ (5th) P(lose >½ stk) Stake %budget sand3 n=3 sharp +36.5% $+496 51% $+27 $-1,411 25% $1,360 54% shrunk +19.3% $+262 45% $-80 $-1,433 32% $1,360 54% uniform -5.9% $-80 35% $-302 $-1,459 44% $1,360 54% cheap5 n=5 sharp +17.8% $+116 56% $+74 $-629 17% $655 26% shrunk +9.0% $+59 51% $+17 $-648 22% $655 26% uniform -4.5% $-30 44% $-77 $-664 29% $655 26% fav5 n=5 sharp +27.6% $+514 59% $+400 $-1,697 16% $1,863 75% shrunk +10.9% $+203 51% $+42 $-1,816 22% $1,863 75% uniform -13.9% $-258 37% $-468 $-1,898 34% $1,863 75% rec5 n=5 sharp +21.6% $+490 59% $+397 $-2,178 16% $2,267 91% shrunk +8.8% $+199 52% $+57 $-2,245 21% $2,267 91% uniform -10.2% $-231 41% $-424 $-2,302 30% $2,267 91% rec6 n=6 sharp +21.8% $+660 59% $+504 $-2,467 16% $3,021 121% ⚠ shrunk +9.5% $+286 52% $+81 $-2,832 20% $3,021 121% ⚠ uniform -8.6% $-261 41% $-478 $-2,999 29% $3,021 121% ⚠ rec8 n=8 sharp +19.6% $+672 59% $+518 $-2,513 13% $3,425 137% ⚠ shrunk +8.3% $+283 52% $+105 $-2,789 18% $3,425 137% ⚠ uniform -8.3% $-283 41% $-474 $-3,105 26% $3,425 137% ⚠ Slate comparison [figure: Slate comparison] Bar = inter-quartile range, whisker = 5th/95th pct, white dot = median, green diamond = mean. Smaller value slates (sand3, cheap5) carry less dollar downside; bigger slates add cost faster than EV. Buyback model: naive vs adverse selection [figure: Buyback effect] Adverse selection (owners buy back the strong teams, we keep the losers) raises total dollars but lowers ROI and pushes capital deployed from $3,021 to $4,293 — it makes the big slates budget-infeasible. The honest metric is ROI, which falls. Optimal number of teams & bet sizing [figure: ROI vs slate size] More teams cut variance but E[ROI] drifts down as the slate grows. Beyond ~5 teams you also blow the budget. Sweet spot: 3–5 teams. - More teams → lower variance (P(lose >½ stake) 25% → 13% sharp) but E[ROI] erodes (+36% → +20%), and you exceed budget past ~5 teams. - fav5 — best balance: +28% E[ROI] sharp, 59% P(profit), 75% of budget. The recommended single slate. - cheap5 — lowest variance and smallest dollar VaR (~$630); a base-hit portfolio. - sand3 — highest E[ROI] (+37%) but only 3 bets, so feast-or-famine (lowest P(profit)). - Sizing rule: stake ~60% of budget (~$1,500) on a 4–5 team slate; hold ~40% reserve. Caveat on EV Under the uniform scenario every slate is negative-EV (favorites hit −39% in a backtest year). If Wednesday's room prices efficiently, reality sits near uniform. All the positive EV above is contingent on our probability edge being real — the biggest risk, bigger than buyback or price noise, and the reason to bet small and keep powder dry. Also note: if we win every bidding war (pay recommended_max_bid ), rec6 E[ROI] collapses from +21.8% to +9.1% — discipline at the auction matters as much as slate choice. Assumptions Budget $2,500 partnership stake. Payout Within each flight, winner 70% / runner-up 30% of the payout pool; 10% of the pot directed to the shootout. Ties split. Buyback (naive) Every won team half bought back → cost = hammer/2, own 50%. Buyback (adverse) P(buyback)=logistic(−0.62 + 12·(p_win−1/6)); favourites ~55–70%, average ~35%. We keep full ownership of losers (adverse selection). Tunable. Probabilities Sharp p_win / p_2nd , shrunk-to-uniform versions, and uniform 1/6 (pessimistic). We do not know which is right. Prices est ~ est_price ±12% noise; maxbid = recommended_max_bid (we win every bidding war). Outcomes Winner ~ p_win; runner-up ~ p_2nd over non-winners. Buyback decided pre-outcome (independent). Flights independent. 50,000 draws/cell, seed 20260610, common random numbers across scenarios. Engine: mc_portfolio.py . Data: valuation/bid_sheet_enriched.csv , valuation/team_values.csv . Times US Eastern. ============================================================================ # Risk & Stress Test Source URL: https://calcutta.high.green/reports/portfolio_stresstest ============================================================================ Red-team risk analysis · 2026 Calcutta Portfolio Stress Test — how the partnership loses money Adversarial audit of the recommended bid portfolio. Every optimistic assumption is assumed wrong; each failure mode is quantified in dollars against a $2,500 stake. VERDICT: thin and fragile — likely −EV in a live room. Play small or not at all. The model shows the portfolio at +611 EV (+24%), but that survives only if every optimistic assumption holds at once. Each individually plausible adverse assumption erases roughly the whole edge; any two stacked turn it negative. The realistic pessimistic case is -$561 (-18% ROI), and even the model's own best case loses money 39% of the time and loses half the stake 20% of the time. The only documented edge rests on 2 sandbagger teams and a single anomalous backtest year. +611 Model baseline net EV (everything goes right) -$561 Pessimistic-but-plausible net EV 20% Chance of losing ≥50% of stake (best case) 39% Chance of losing money (best case) 1. What's known — verified against the data Re-derived from 2024-25 prices joined to real finishes, paying exact prices and exact flight payouts (winner 70% / runner-up 30% of the payout pool; 10% of the pot directed to the shootout). Strategy (1 team/flight) 2024 ROI 2025 ROI Buy the favorite (priciest) -38.2% +18.0% Buy the cheapest -1.7% -8.3% Buy everything (whole pot) -10.0% -11.1% The "favorites win" thesis hangs on one year. Favorites won 2/18 flights in 2024 (11% — worse than the 17% random baseline) but 6/15 in 2025 (40%). Price–finish rank correlation was a weak +0.259 / +0.2213. 2025 cannot be distinguished from noise — and it is the only year the strategy worked. 2. The portfolio under attack Best dollar-edge teams, one per flight, ≤$2,500 at est_price. Fair value = P(win)·0.70·0.90·pot + P(2nd)·0.30·0.90·pot , reproduced from the bid sheet. Flt Team P(win) Pot Fair Price Edge Sandbag 20 Barnes + Smelcer 30.0% $1,200 $288 $200 +88 9 Wood + Estes 32.2% $4,000 $986 $700 +286 yes 12 Wright + Gadsby 21.6% $3,300 $603 $500 +103 yes 8 Nodar + Heslep 20.5% $4,700 $814 $700 +114 15 Alexander + Lambeth 17.6% $2,700 $418 $400 +18 Baseline (everything goes right): spend $2,500 → EV return $3,111 → net +611 (+24%) . This is the number under test. Pot caveat: the bid sheet's own flight pots sum to $84,100 , higher than the ~$76k projection cited. Every fair value is proportional to the pot, so a lighter room already shrinks all edges ~10% before any other stress. 3. Failure modes, ranked by $ impact EV hit vs. the +611 baseline, each mode in isolation. Bars scaled to the largest hit. # Failure mode Mechanism EV hit Net after Severity 1 Efficient market Price = unbiased fair value -$861 -$250 2 Pot -25% (fixed price) Room spends less; our prices already committed -$778 -$167 3 Winner's curse / overpay We win bidding wars - pay our max-bid ceiling -$600 -$8 4 Buyback adverse selection Owners reclaim winners (own 50%), dump losers (own 100%) -$319 +$292 5 Sandbag doesn't convert Flagged teams regress halfway to field average -$247 +$363 6 First right of refusal Strong teams pulled at the $2,000 cap; we get the weaker pool -$141 +$470 Pot projection sweep (±30%, fixed prices) Pot vs. projection -30% -20% -10% base +10% +30% Net EV -$323 -$12 +$299 +$611 +$922 +$1,544 Our cost is fixed once we bid; the pot is not. A pot 20% light zeros the edge; 30% light is a loss. Sandbag concentration risk 64% of the portfolio's entire edge (+$390 of +$611) comes from just 2 sandbagger teams. If the room also bids them up, edge → +$221. If the sandbag doesn't convert in the 2-man net better-ball, edge → +$363. Precedent: Williford — the headline confirmed sandbagger — finished 5th in flight 7 in 2025 despite a scratch-level game. Individual ball-striking does not reliably win a team match-play points race. 4. Combined "pessimistic but plausible" Pay max-bid + pot 15% light + sandbag converts 75% + no buyback relief → net -$561 on $3,100 spent (-18% ROI). And this excludes buyback adverse selection and first-right-of-refusal, either of which deepens the loss. 5. Variance / risk of ruin 200k-sim Monte Carlo on the model-optimistic portfolio (own 100%, model correct — the best case). +$615 Mean P&L +$461 Median P&L 39% P(lose money) 20% P(lose ≥50% of stake) 7% P(total wipeout — 0 teams cash) ~$0 Mean P&L under buyback adverse selection Even assuming the model is right, we lose money 2 years in 5 and lose half the stake 1 year in 5. 5 teams on $2,500 is over-concentrated for a near-zero-edge, high-variance game. 6. Verdict & mitigations The edge is real only on paper and rests almost entirely on 2 sandbagger teams and a single anomalous backtest year (2025). Play small, or not at all. If the partnership plays anyway: - Bet smaller. Cut the stake to $1,000–1,500 . The edge is small; the variance is huge. - Only buy strictly below value — never at the ceiling. Hard walk-away $200+ below est_price . The only defense against the winner's curse (alone worth −$600). - Avoid favorites / expensive flights. Favorites lost 38% in 2024 and carry the most absolute downside. Skip anything over ~$700. - Concentrate on verified-cheap, results-backed sandbaggers — demand the result , not the reputation. Williford's 5th is the cautionary tale. - Treat the buyback as the owner's option against you. Price every team as if you keep 100% of losers and 50% of winners. - Walk-away discipline on the pot. If the room bids light early, re-anchor your max bids downward in real time. - Diversify if you must play big — more small positions cut the 7% wipeout risk. Assumptions & limitations. Fair value reproduced exactly from the bid sheet (flight-only; shootout equity ~0). Backtest covers 2024-25 only — the n=2 is itself the caveat. Name-join 99%+. MC treats flights as independent (one team/flight). Buyback (owner foresight 60%), first-refusal (20% regression), sandbag-failure (50% regression) are deliberately moderate , not worst-case — harsher defensible values make the verdict worse. All probabilities are the model's own shrunk values: we attacked the model on its own terms. Generated by backtest.py + stresstest.py + make_report.py ; figures in backtest_results.json / stress_results.json . ============================================================================ # For an Investor (Mark Cuban) Source URL: https://calcutta.high.green/reports/investor ============================================================================ Investment Memo · Confidential · 10 June 2026 2026 Member-Member Golf Calcutta — Prepared for Mark Cuban A read on a small, high-variance, information-advantaged betting partnership. Every number is sourced to our own analysis. Where it's unproven, it says so. ~0% Skill share of outcome variance (r = −0.12) n = 7 Data points the sandbagger edge rests on (CI spans negative) ~$1,300 Stake requested (of a $2,500 bankroll) 1. Summary This is a $2,500 bankroll bet into a member-member golf Calcutta — a live auction where you buy teams and collect if your team wins or places in its 6-team flight. Five years of results say flight outcomes in this 9-hole net-better-ball format are statistically indistinguishable from a lottery (year-over-year skill correlation r = −0.12, CI [−0.41, +0.17], skill share of variance ~0%). The single-Wednesday expected return is not a confident positive number — it is near break-even unless our private information edge is real, in which case the model says +8% to +24% after disciplined sizing. The ask is ~$1,200–$1,500 deployed (not the full $2,500), for an asymmetric, low-downside, repeatable bet whose asset is a proprietary information + execution pipeline, not the outcome of one auction. If you want a sure thing or a scalable fund, this is not it. 2. The bear case The worst facts, up front. # The hard fact The number B1 Outcomes are near-random. Finishing well one year does not predict the next. team r = −0.12, CI [−0.41, +0.17]; skill share ~0% B2 No price rule beats the field with confidence. Every positive backtest strategy's CI includes large losses. mid-price +4.0%, CI [−44.7%, +57.0%] B3 Favorites are a coin flip that flipped. The "market predicts winners" thesis rests on one anomalous year. −38% / −37% (2024) vs +18% / +12% (2025) B4 The entire edge rests on ~2 sandbagger teams. 64% of model edge from Wood+Estes & Wright+Gadsby. $390 / $611 B5 That sandbagger edge is statistically unproven. Best signal, least trustworthy: 7 picks, 2 years, 2 wins dominate. +76.7%, CI [−45.0%, +192.1%], n=7 B6 ~1/3 of random portfolios made money. A two-year positive result is not evidence of skill. 32% profitable B7 The buyback is the owner's option against us (adverse selection): reclaim winners, dump losers. +$308 → −$11 mean ($319 swing) B8 Winner's curse is real. We win auctions on teams we like because we value them above the room. −$600 EV → break-even B9 Risk of ruin is non-trivial — even assuming our model is right. P(lose) 38.6%; P(lose ½) 20.1%; P(wipeout) 6.7% B10 It doesn't scale. A single ~$84k pool, capped bids ($2,000), one night a year. total pot $84,100 Under a pessimistic-but-plausible stack (pay our ceiling + pot 15% light + sandbaggers convert at 75% + no buyback relief), the model's own red-team puts EV at −$561 (−18% ROI) . Under fully efficient pricing it is −$250 on $2,500 . 3. The actual edge — quantified and caveated The thesis is not "we predict winners." Section 2 proves we largely can't. It's three narrower claims: (a) An information edge no other bidder assembles in one place Four sources, stacked by trust: (1) a GHIN-driven Monte-Carlo valuation — 60k hole-by-hole sims/flight, audited as correct and unbiased (Gelman-Rubin R-hat = 1.000); (2) five years of real results (2021–25, 466 rows) — the only signal that survives an honest out-of-sample test, and orthogonal to handicap and to the model (r = 0.05); (3) a Cap Patrol form/clutch overlay, used as a cross-check only; and (4) insider human reads — the genuinely proprietary input. A partner personally watched a flagged sandbagger (Williford) shoot 73 in a four-club tournament — near-scratch ability behind ~8–9 strokes of phantom handicap GHIN cannot see. Copeland is a known sandbagger who won Flight 5 in 2025 — result-confirmed. No other bidder combines a calibrated simulator, 5 years of normalized results, a form/clutch overlay, and decades of inside knowledge. Caveat: it moved only ~2 teams enough to matter, on n=7. Thin and unproven at scale. (b) An execution edge — walk-away discipline is worth ~13 ROI points In the forward Monte-Carlo, winning bidding wars (paying our ceiling) cuts E[ROI] from +21.8% to +9.1% — a ~13-point swing; the red-team prices the same effect at −$600 EV . The single most valuable behavior is not bidding when the price crosses our number — enforced live by a tool that re-prices every unsold team after each sale. (c) The durable asset — a reusable, compounding pipeline A research-grade flight simulator (60k-quality precision at ~15–20k sims; uncertainty bands on every P(win)); a normalized 5-year results database with collision handling; a Cap Patrol overlay distilled to two real signals; and a live auction tool tracking 20 flight pots and enforcing walk-away limits. That cost is sunk and improves every year as more results land — at 5+ years the repeatability CI tightens enough to settle whether any skill signal exists. The single-year edge is thin. 4. Unit economics & the bet Bankroll $2,500. Payout within each flight: winner 70%, runner-up 30%, with 10% of the pot directed to the shootout. Buyback — team can buy back half; planning assumption is adverse selection (own 50% of winners, 100% of losers). Pot $84,100 is a projection , not a fit to 2026 prices; pot risk is asymmetric against us (a 20%-light pot zeros the edge). The disciplined buy card (results-backed, not reputation) Flt Team P(win) Fair $ Est $ Edge $ Why 9 Wood + Estes 43% 1,424 700 +724 Sandbagger and won Fl 7 in 2025 — two independent signals agree 12 Wright + Gadsby 25% 781 500 +281 Flagged sandbagger; model bargain 5 Wright + Copeland 20% 1,089 900 +189 Copeland sandbag result-confirmed (won Fl 5, 2025) 2 Downey + Shearer 23% 1,545 1,300 +245 Won Fl 2 in 2025; 5-yr consistent; clean names Hard rules (each counters a quantified failure mode): buy only $200+ below est-price (defends the −$600 winner's curse); avoid favorites/expensive flights (lost 38% in 2024); demand the result, not the reputation (Williford finished 5th in 2025 despite scratch ball-striking); price the buyback as the owner's option against you. Expected P&L range (forward Monte-Carlo, naive buyback, est-price) Scenario Assumes Slate E[ROI] E[P&L] P(profit) P(lose ½) Sharp Model probabilities roughly right +18% to +37% +$116 to +$514 51–59% 16–25% Shrunk Probabilities pulled toward uniform +8% to +19% +$59 to +$262 45–52% 22–32% Lottery Room prices efficiently; no edge −4% to −14% −$30 to −$258 35–44% 29–44% Downside: even in the model-optimistic case, P(lose money) = 38.6% and P(lose ≥½ stake) = 20.1%. Under the pessimistic stack, central EV is −$561. Why $1,200–$1,500, not $2,500: the 6-team slate doesn't fit the budget (half-back ~$3,021 = 121%; adverse ~$4,293 = 172%). The red-team's explicit instruction: cut stake to $1,000–$1,500 . Deploy ~60% (~$1,500) on a 4–5 team value-tilted slate; hold ~40% dry powder. 5. The bet profile The bet is small, asymmetric, disciplined, information-advantaged, and repeatable. Low-downside (no catastrophic-loss path; loss bounded by a stake sized to lose); asymmetric (top team +104% modeled edge; slate upside +28% to +37% vs a bounded downside); information-advantaged (eyewitness reads + 5-yr results no other bidder combines); disciplined (walk-away worth ~13 ROI points, enforced by tooling); and repeatable . The stake is in a repeatable, improving edge-finding operation in an information-poor market: size it small, play it many times, let the process and the data decide. 6. What Cuban will ask Q1. If you can't predict winners, why isn't this just gambling? We arbitrage prices , not outcomes. Outcomes are near-random, so the edge is buying below the censored-corrected fair-value curve using two sources the room lacks (eyewitness sandbagger reads + 5-yr results). Caveat: unproven at scale (B5). It is advantaged gambling, sized to survive being wrong. Q2. Your edge is 7 data points. That's noise. +76.7% but n=7, CI [−45%, +192%], 2 wins dominate. It is a lead to track forward. That is why we bet $1,200–$1,500, and why the durable thesis is the pipeline that converts more years into a real answer, not this year's 7 points. Q3. What's your real, durable moat? Ranked: (1) proprietary human intel — you can't buy "I watched him shoot 73 in a four-club event"; (2) the integrated pipeline no casual bidder replicates; (3) walk-away discipline worth ~13 ROI points. The model alone is not a moat — decades-experienced members price nearly as well. Q4. Why should I trust the sandbagger reads? Only where a result backs them. Copeland won Fl 5 in 2025; Wood+Estes won Fl 7 in 2025 — confirmed. Distrust the reputation-only ones: Williford, the headline sandbagger, finished 5th in 2025. We demand the outcome, not the game. Q5. The buyback protects your downside, right? No — it's the owner's option against us: they reclaim winners (we keep 50%), dump losers (we keep 100%). A $319 swing from +$308 to −$11 mean. We price it as a cost, not protection. Q6. What if the room is efficient and your model is just wrong? Then EV is −$250 on $2,500. That is the default , not a tail; the absence of persistence suggests the model is partly wrong. Defense: bet size, walk-away discipline, dry powder. We do not bet the budget on the model being right. Q7. What's the scale — this is a $2,500 pool? Yes. Total pot $84,100, capped bids, once a year. This is not a scalable fund. What scales is the method : the pipeline applies to any info-asymmetric small-pool auction and improves with each year of data. Q8. Give me the one-line EV. Single-year ROI between −18% (pessimistic) and +24% (model-optimistic) , center near break-even-to-slightly-positive only if the information edge is real, with a ~20% chance of losing half the stake regardless. 7. The ask & terms Term Detail Stake requested $1,200–$1,500 working capital (~50–60% of the $2,500 bankroll); rest held as dry powder What you're buying A stake in a repeatable edge-finding operation — proprietary intel + audited pipeline + walk-away discipline — not one auction's outcome Return range (1 yr) −18% (pessimistic) to +24% (model-optimistic); center ~break-even to modestly positive iff the info edge is real Max downside Bounded by the sized stake; ~20% chance of losing ~half in a year even when the model is right. No catastrophic-loss path. The real return Compounding — each year of results sharpens the skill estimate and the pipeline. The asset is the operation, not the Wednesday. Expected-value statement. This is a near-random format where outcomes show essentially zero year-over-year skill, and where the only documented edge rests on ~2 sandbagger teams across 7 unproven data points. The single-year expected value is roughly break-even — positive (+8% to +24%) only if our proprietary information edge is real, negative (−18% to −10%) if the room prices efficiently — with a ~20% chance of losing half the stake either way. This is a small, asymmetric, disciplined, information-advantaged, repeatable bet whose durable value is a proprietary, compounding data-and-execution pipeline. Size it to lose, play it many times, let the process decide. Assumptions - Budget $2,500; within each flight winner 70% / runner-up 30%, 10% of pot to the shootout; ties split. - Buyback planning assumption is adverse selection (own 50% of winners, 100% of losers). - Probabilities are the model's own values, attacked on their own terms; sharp/shrunk/uniform scenarios bracket the truth. - Pot is a 2026 projection ($84,100), not a fit to actual 2026 prices; pot risk is asymmetric against us. - Backtest covers only 2024–2025 (only years with both prices and results); n=2 is the central caveat. - Historical sandbagger strategy is n=7 with a CI spanning negative; a lead to track, not a proven edge. - All ROI/P&L ranges are model outputs, not guarantees; the simulator is audited as unbiased but inherits any bias in its player-distribution inputs. Sources: portfolio backtest, portfolio Monte-Carlo, portfolio stress-test, statistical-expert review, Monte-Carlo-expert audit, flight report & value board, methodology, manual intel, Cap Patrol psych findings, historical track-record findings — all in this repository. Numbers cited are computed in those reports from the project's own data. Prepared 2026-06-10 (US Eastern).