ST Model Accuracy

Out-of-sample backtests of the per-game projection engine — projections vs. what actually happened, compared to a simple baseline
Per-game projections
Beat the baseline by 4% on per-game accuracy · p10–p90 coverage 81% (2025-26 backtest, 3,294 preds)
Player Accuracy
per-game rate vs Baseline · 3,294 predsskill 4%p10–p90 cov 81%
StatNModel/gBase/gSkillCov
Scoring
Points3662.542.71+6%78%
3PM3660.350.36+5%79%
Playmaking
Assists3660.690.73+6%76%
Turnovers3660.330.36+7%77%
Defense & Boards
Rebounds3660.860.88+2%86%
Steals3660.20.2-1%80%
Blocks3660.140.15+8%82%
Efficiency
FG%3660.030.04+5%84%
FT%3660.050.06+2%87%
Game Model — Out-of-Sample Backtest
backtest
SeasonGamesBrierBaseWin%Margin MAETotal MAEVegas MAEBeat%
2025-261,2130.20590.247169%11.4016.3
2024-251,2160.21130.248266%11.0316.0
2023-241,2120.21200.248066%11.0715.5
2022-231,2100.22740.243463%10.2514.9
2021-221,2210.22430.248464%11.1017.9
2020-211,0980.23170.249261%11.1324.1
2019-209490.21440.247667%10.1315.7
2018-191,2050.21380.241966%10.2515.2
2017-181,2050.21850.243764%9.9415.2
2016-171,2080.21950.243065%9.9915.1
2015-161,2120.20860.241968%9.7314.7
Overall12,9490.21700.245665%10.5516.4
The model's pre-game win probability is scored by Brier against every completed game's outcome; a Brier below the home-field Basemeans it adds signal beyond "home teams win." Margin/Total MAE measure the projected score. Vegas MAE / Beat% read — for this sport: the free odds feed carries no historical closing lines yet, so those columns populate as line history accrues.