Model the event
Historical performance and available team or player information inform estimates of possible outcomes.

Publicly documented methods, a simple pricing example, and a look back at analyst forecasts.
Sportsbooks combine statistical estimates with trading and risk management. Public disclosures describe the ingredients; they do not reveal the exact model weights or a reproducible pricing engine.
Historical performance and available team or player information inform estimates of possible outcomes.
The displayed odds include a bookmaker margin. They are not a direct statement of an outcome’s true probability.
New facts, market activity and the operator’s exposure can influence revised prices. A movement alone does not identify its cause.
Imagine two mutually exclusive outcomes with a 50% chance each. Fair decimal odds would be 2.00 on both. Hypothetical quotes of 1.91 on both imply 52.36% each.
The excess above 100% is 4.71 percentage points of overround. It is a property of the quoted prices, not a guaranteed profit rate. Removing it mathematically still does not establish the true probabilities.
Illustrative arithmetic only. No real event or sportsbook quote is used.FanDuel’s 2022 investor presentation describes quantitative models built on historical data alongside traders who manage inputs and respond to team and player news. Flutter also publicly describes proprietary pricing and risk tools.
FanDuel disclosure · 2022 ↗Flutter sportsbook technology ↗Its 2021 annual report describes proprietary sportsbook technology, trading and risk management, and the completed migration to its acquired SBTech platform. This establishes historical infrastructure, not today’s exact pricing formula.
DraftKings annual report · 2021 ↗Its published risk-management discussion describes tracking initial prices, price changes, wagers and liabilities. That is evidence of an ongoing trading process; it does not expose its private probability model.
Pinnacle risk analytics ↗The same public quote can result from different forecasts, margin allocations and risk decisions. Private model weights, customer activity and liabilities are missing. Any claimed reconstruction would need independent testing and clearly stated assumptions.
Joint outcomes also require assumptions about dependence: events involving the same players or game need not be independent. Public prices alone do not reveal those assumptions.
Sports Illustrated / MMQB winner forecasts. These are published opinions, not confirmed wagers or audited bettor records. Coverage: one NFL season’s reported records and three completed postseason events.
Correct versus incorrect winner predictions, as reported by SI on January 23, 2025. Each row covers 272 predictions. Accuracy does not measure profitability; prices and stakes are unavailable.
| Analyst | Correct–incorrect | Accuracy |
|---|---|---|
| Albert Breer | 182–90 | 66.9% |
| Conor Orr | 165–107 | 60.7% |
| Gilberto Manzano | 171–101 | 62.9% |
| John Pluym | 186–86 | 68.4% |
| Matt Verderame | 198–74 | 72.8% |
| Mitch Goldich | 186–86 | 68.4% |
Source-reported totals; individual regular-season forecasts have not been independently regraded here. No spread, total, player-prop or ROI breakdown is available from this sample.
A majority describes this panel’s opinions. Three selected games cannot establish predictive skill or represent all experts.
Eagles 55–23 Commanders
Eagles 4 · Commanders 2
Published January 23, 2025, before the event.
Published forecasts ↗Official final result ↗Chiefs 32–29 Bills
Bills 4 · Chiefs 2
Published January 23, 2025, before the event.
Published forecasts ↗Official final result ↗Eagles 40–22 Chiefs
Chiefs 7 · Eagles 2
Published February 6, 2025, before the event.
Published forecasts ↗Official final result ↗