Skip to content

NHL Model Performance

Published forecasts archived before puck drop and graded against official final scores. Each game uses its last pregame revision. Log loss and Brier score judge the home win probability (a coin flip scores 0.693 and 0.250); total MAE and bias judge the model total in goals. The comparison with the closing line follows.

Games107
Log loss0.6615coin flip 0.693
Brier0.2350coin flip 0.250
Accuracy58.9%
Total MAE1.89goals
Total bias-0.07goals

Against the closing line

The close is the last partner-book quote from the 15 minutes before puck drop. The moneyline is converted to a home win probability with the vig removed. Model and market are scored on the same games: only those with a close, so these figures differ from the full-sample ones above. Lower is better.

Model log loss0.704219 games
Market log loss0.6896closing moneyline
Gap to market+0.0146lower is better
Model total MAE1.4319 games
Market total MAE1.45closing total
Gap to market-0.01goals

Pure-model log loss: 0.7300 on the same 19games. The published forecast blends in the morning market price, so its error and the close's are correlated.

Calibration

Observed home win rate against the predicted probability in ten bins. Points above the diagonal mean the model was too pessimistic about the home side, below it too confident.