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 1
16 games · 5 of them land in the coin-flip band, where the model has no usable call
- NEatSEAWed, Sep 9, 8:20 PMSEA 69%SEA -5.2High confidence
- SFatLARThu, Sep 10, 8:35 PMLAR 51%LAR -3.4Coin flip, no call
- CLEatJAXSun, Sep 13, 1:00 PMJAX 76%JAX -9.0High confidence
- BUFatHOUSun, Sep 13, 1:00 PMHOU 51%BUF -1.9Coin flip, no call
- BALatINDSun, Sep 13, 1:00 PMBAL 55%BAL -3.9Low confidence
- ATLatPITSun, Sep 13, 1:00 PMPIT 60%PIT -3.4Medium confidence
- NOatDETSun, Sep 13, 1:00 PMDET 73%DET -7.2High confidence
- NYJatTENSun, Sep 13, 1:00 PMTEN 57%TEN -1.7Low confidence
- TBatCINSun, Sep 13, 1:00 PMCIN 68%CIN -5.1High confidence
- CHIatCARSun, Sep 13, 1:00 PMCHI 53%CHI -2.4Coin flip, no call
- GBatMINSun, Sep 13, 4:25 PMMIN 53%MIN -1.8Coin flip, no call
- MIAatLVSun, Sep 13, 4:25 PMLV 65%LV -3.1High confidence
- WASatPHISun, Sep 13, 4:25 PMPHI 60%PHI -3.9Low confidence
- ARIatLACSun, Sep 13, 4:25 PMLAC 80%LAC -10.9High confidence
- DALatNYGSun, Sep 13, 8:20 PMDAL 55%DAL -3.2Coin flip, no call
- DENatKCMon, Sep 14, 8:15 PMKC 58%KC -2.9Low confidence
Calibrated Logistic Regression (core + market features) · walk-forward by season (train on all prior seasons, test on the next)