Model Weights & Custom Rankings

Tune the combination win-probability model and build your own ranking rule

Combination Model Weights

The Home WinProb on each game card is a weighted blend of the individual models below. Weights are normalized across the enabled models, so only their relative sizes matter.

Loading model status...
How the individual model probabilities are combined.
Std dev used to convert predicted margins into win probability for the Ridge and Random Forest models. Leave blank for auto (validation residuals, ≈15-16 pts).
Working...

Custom Ranking Rule Builder

Build your own Top 25: pick season efficiency/four-factors metrics, weight them, and choose whether higher or lower is better. Each metric is min-max normalized across FBS teams and the weighted sum becomes the composite score.

Season of advanced stats used for the preview table.
Working...