Full Matchup Board
Confidence Calibration
Every pick this season, grouped by how confident the model said it was. "Actual" should roughly match "Predicted" if a range is well-calibrated — useful for spotting which confidence bands hold up.
A positive gap means the model was underconfident in that range (it did better than it said); a negative gap means it was overconfident. Small sample sizes at the high-confidence end make single-week swings expected — read this as a season-long signal, not a weekly one.