Win probability
Game Prediction Model
- 20seasons of data2006–2025
- 5,133games trained on4,166 more held back for testing
- 64.7%called correctlyon games the model never saw
- ±2.1points of calibration driftwhen it says 70%, it happens about 70% of the time
23.2%decided by 3 points or fewer
47.5%decided by 7 points or fewer
8points, the median margin
A quarter of NFL games come down to a field goal, which is the ceiling every forecaster runs into. Predicting every game decided by more than a touchdown perfectly, and coin-flipping the rest, would still only reach 76.3%. Across 5,133 games, 2006–2025.
Week 13
14 games · 3 of them land in the coin-flip band, where the model has no usable call
- KCatLARThu, Dec 3, 8:15 PMLAR 58%LAR -3.8Low confidence
- DETatATLSun, Dec 6, 1:00 PMATL 66%ATL -3.5High confidence
- SFatNYGSun, Dec 6, 1:00 PMNYG 62%NYG -3.1Medium confidence
- CINatCLESun, Dec 6, 1:00 PMCLE 53%CLE -1.5Coin flip, no call
- LACatTBSun, Dec 6, 1:00 PMTB 63%TB -4.1Medium confidence
- JAXatCHISun, Dec 6, 1:00 PMCHI 62%CHI -3.4Medium confidence
- WASatTENSun, Dec 6, 1:00 PMTEN 52%TEN -1.6Coin flip, no call
- GBatNOSun, Dec 6, 1:00 PMNO 65%NO -4.5Medium confidence
- PHIatARISun, Dec 6, 4:05 PMARI 52%PHI -0.2Coin flip, no call
- MIAatDENSun, Dec 6, 4:05 PMDEN 65%DEN -3.4High confidence
- CARatMINSun, Dec 6, 4:25 PMMIN 58%MIN -4.1Low confidence
- BUFatNESun, Dec 6, 4:25 PMNE 58%NE -2.5Low confidence
- HOUatPITSun, Dec 6, 8:20 PMPIT 62%PIT -4.5Medium confidence
- DALatSEAMon, Dec 7, 8:15 PMSEA 61%SEA -3.5Medium confidence
Calibrated Logistic Regression (core + market features) · walk-forward by season (train on all prior seasons, test on the next)