Box plus-minus (BPM)

Box plus-minus estimates how many points per 100 possessions a basketball player contributes above or below a league-average player, using only the statistics found in a standard box score, points, rebounds, assists, steals and turnovers, rather than play-by-play lineup data.

BPM is built by regressing box-score statistics against play-by-play-derived plus-minus data from past seasons, producing a formula, with separate versions for guards, wings and bigs, that turns box-score inputs into an estimated impact number. This differs from raw plus-minus, which needs no model but is entirely teammate- and opponent-dependent, and from true adjusted plus-minus, which uses lineup data directly but requires far more possessions to stabilize. BPM sits between the two: model-based, available for any season with box scores, but still an estimate rather than a direct measurement.

BPM feeds into value-over-replacement-player calculations and is widely used for historical comparisons, since box scores exist for seasons with no tracking data. The main pitfall is treating BPM as a precise measurement: it is a regression estimate with real uncertainty, it can undervalue defense, which box scores capture poorly beyond steals and blocks, and different sites' versions of BPM use different training data and coefficients, so values are not always directly comparable across sources.