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 11
13 games · 1 of them land in the coin-flip band, where the model has no usable call
- INDatHOUThu, Nov 19, 8:15 PMHOU 63%HOU -2.0Medium confidence
- JAXatNYGSun, Nov 22, 1:00 PMNYG 66%NYG -4.2High confidence
- NOatCHISun, Nov 22, 1:00 PMCHI 67%CHI -3.2High confidence
- TENatDALSun, Nov 22, 1:00 PMDAL 66%DAL -3.9High confidence
- BALatCARSun, Nov 22, 1:00 PMCAR 57%CAR -2.1Low confidence
- MIAatBUFSun, Nov 22, 1:00 PMBUF 64%BUF -4.5Medium confidence
- ARIatKCSun, Nov 22, 1:00 PMKC 61%KC -3.7Medium confidence
- TBatDETSun, Nov 22, 1:00 PMDET 51%DET -1.7Coin flip, no call
- NYJatLACSun, Nov 22, 4:05 PMLAC 57%LAC -2.5Low confidence
- PITatPHISun, Nov 22, 4:25 PMPHI 65%PHI -3.3High confidence
- LVatDENSun, Nov 22, 4:25 PMDEN 63%DEN -2.5Medium confidence
- MINatSFSun, Nov 22, 8:20 PMSF 59%SF -6.6Low confidence
- CINatWASMon, Nov 23, 8:15 PMWAS 66%WAS -4.0High confidence
Calibrated Logistic Regression (core + market features) · walk-forward by season (train on all prior seasons, test on the next)