Runs Under Predictions — Performance

Cumulative Units

Calibration Reliability

Each row is a model-probability decile. Predicted is the mean model P(zero) in the bucket; Actual is the observed share that scored zero. A well-calibrated model has the two columns track each other.

Breakdowns

By Edge Bucket

By Odds Band

By Batting Slot

By Park (min 5 picks)

Loss Diagnostics

Categorizes every losing UNDER pick by how many runs the player actually scored.