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 4
16 games · 4 of them land in the coin-flip band, where the model has no usable call
- PITatCLEThu, Oct 1, 8:15 PMPIT 59%PIT -4.6Low confidence
- INDatWASSun, Oct 4, 9:30 AMWAS 53%WAS -2.3Coin flip, no call
- JAXatCINSun, Oct 4, 1:00 PMCIN 67%CIN -3.8High confidence
- DALatHOUSun, Oct 4, 1:00 PMHOU 59%HOU -1.3Low confidence
- TENatBALSun, Oct 4, 1:00 PMBAL 77%BAL -9.3High confidence
- LARatPHISun, Oct 4, 1:00 PMLAR 52%LAR -1.1Coin flip, no call
- NYJatCHISun, Oct 4, 1:00 PMCHI 74%CHI -7.8High confidence
- ARIatNYGSun, Oct 4, 1:00 PMNYG 75%NYG -8.5High confidence
- GBatTBSun, Oct 4, 1:00 PMGB 50%GB -0.7Coin flip, no call
- NEatBUFSun, Oct 4, 1:00 PMBUF 61%BUF -4.3Medium confidence
- MIAatMINSun, Oct 4, 4:05 PMMIN 73%MIN -8.0High confidence
- KCatLVSun, Oct 4, 4:25 PMKC 69%KC -7.0High confidence
- LACatSEASun, Oct 4, 4:25 PMSEA 72%SEA -5.7High confidence
- DENatSFSun, Oct 4, 4:25 PMSF 59%SF -3.6Low confidence
- DETatCARSun, Oct 4, 8:20 PMDET 54%DET -2.8Coin flip, no call
- ATLatNOMon, Oct 5, 8:15 PMNO 57%NO -1.5Low confidence
Calibrated Logistic Regression (core + market features) · walk-forward by season (train on all prior seasons, test on the next)