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 6
14 games · 4 of them land in the coin-flip band, where the model has no usable call
- SEAatDENThu, Oct 15, 8:15 PMSEA 50%SEA -2.9Coin flip, no call
- HOUatJAXSun, Oct 18, 9:30 AMHOU 54%JAX -1.9Coin flip, no call
- TENatINDSun, Oct 18, 1:00 PMIND 63%IND -4.2Medium confidence
- CHIatATLSun, Oct 18, 1:00 PMATL 51%CHI -2.2Coin flip, no call
- NOatNYGSun, Oct 18, 1:00 PMNYG 65%NYG -3.3High confidence
- BALatCLESun, Oct 18, 1:00 PMBAL 75%BAL -8.7High confidence
- NYJatNESun, Oct 18, 1:00 PMNE 76%NE -9.1High confidence
- PITatTBSun, Oct 18, 1:00 PMTB 58%TB -1.5Low confidence
- CARatPHISun, Oct 18, 1:00 PMPHI 68%PHI -7.1High confidence
- ARIatLARSun, Oct 18, 4:05 PMLAR 81%LAR -12.9High confidence
- LACatKCSun, Oct 18, 4:25 PMKC 67%KC -2.8High confidence
- BUFatLVSun, Oct 18, 4:25 PMBUF 69%BUF -7.6High confidence
- DALatGBSun, Oct 18, 8:20 PMGB 54%GB -0.7Coin flip, no call
- WASatSFMon, Oct 19, 8:15 PMSF 63%SF -5.6Medium confidence
Calibrated Logistic Regression (core + market features) · walk-forward by season (train on all prior seasons, test on the next)