season 2026

Model

GSSM — runs scored/surrendered model. OLS of the actual home run margin on four home/away team averages, converted to a win probability and then calibrated on the model's own out-of-sample history. Multiple models share this machine, differing only in what feeds the four averages:

E[home margin] = b₀ + b₁·GFH(home) + b₂·GAH(home) + b₃·GFA(away) + b₄·GAA(away)
P(home win) = P( N(E[margin], SE) > 0 ),  SE = √(SSE / (n−2))

Each model refits nightly on all final games of the season (minimum 60). Predictions are stored pre-game and scored out-of-sample as results arrive — the Accuracy page compares the models honestly.

Fit history

Fitted (UTC)ModelGamesb₀ GFHGAHGFAGAA SEIn-sample accBrier
2026-07-30 02:04 xwoba 1619 -0.599 1.009 -2.633 6.022 -2.453 4.631 52.3% 0.249
2026-07-30 02:04 runs 1619 -0.662 1.172 -1.095 -1.090 1.150 4.515 57.6% 0.241
2026-07-30 02:01 runs 1619 -0.662 1.172 -1.095 -1.090 1.150 4.515 57.6% 0.241
2026-07-30 01:54 runs 1619 -0.662 1.172 -1.095 -1.090 1.150 4.515 57.6% 0.241