RAPM

What is RAPM?

RAPM (Regularized Adjusted Plus-Minus) estimates each player's individual contribution to point differential per 100 possessions. Unlike raw plus-minus, RAPM uses ridge regression across all lineup stints to de-confound players who frequently play together.

ORAPM (Offensive): Measures offensive contribution - how much a player helps their team score, adjusted for opponent defense quality.

DRAPM (Defensive): Measures defensive contribution - how much a player helps prevent opponent scoring. Calculated as Total RAPM minus ORAPM.

RAPM Adj.: RAPM smooths the huge volatility of raw plus-minus by pooling thousands of stints, while also deconfounding who played with whom. Even so, short samples are still noisy and should be treated cautiously.

Adj. RAPM denoises the scoreboard before the regression sees it: hot/cold three-point variance is muted by our shot expectation model, and (as of July 2026) free-throw and mid-range (10–23 ft) make/miss variance are muted at 75% of each shooter's deviation from his own historical expectation (full removal over-corrects — some above-baseline shooting is real form; 75% is the validated optimum). Shot selection is untouched — only the luck of whether the shot fell. That lets more signal show through in short samples; over longer samples adjusted and raw RAPM largely converge.

Alpha: The ridge regularization strength. Lower alpha means less smoothing (more variance), which can make the 3PT adjustment do more of the “smoothing” in a more informative way. Small-sample RAPM is still limited in reliability regardless of alpha. Default here is alpha=500.