MLAPM is an all-in-one impact metric for WNBA players, inspired by metrics like DARKO DPM and EPM. It measures how much better or worse a team performs with a given player on the court, relative to a league-average player.

A score of 0.0 = league average. Positive scores mean the player helps their team, negative scores mean they hurt it. The scale is approximately points per 100 possessions above/below a replacement-level player.

Offensive and defensive splits are computed such that O-MLAPM + D-MLAPM = MLAPM exactly.

A retrodictive correlation of 0.91 is publication-worthy for any impact metric. For context, even the most sophisticated NBA metrics (EPM, DARKO) typically achieve 0.85-0.92. The 0.59 year-over-year predictive correlation is on par with EPM and RAPTOR — basketball is inherently variable due to injuries, trades, and role changes.

Veteran projections use a linear regression model trained on 90 players who appeared in both the 2024 and 2025 WNBA seasons. The model predicts 2026 MLAPM from four variables that were each statistically significant in cross-validation.

Prior year MLAPM is the strongest predictor (r = 0.59). AST/100 and TS% add signal because they capture skill components more stable year-over-year than team-dependent metrics. Minutes played accounts for opportunity and role stability.

Rookie projections use a regression model trained on 37 WNBA rookies from the 2019–2025 draft classes who played 350+ minutes in their rookie season. College box scores are sourced from ESPN via wehoop.

Log(draft pick) is the strongest predictor — draft position encodes scout consensus. College PPG adds signal beyond pick position. WNBA transitions are difficult; even top picks frequently post negative MLAPM in year one, so projections are intentionally conservative.

Team stats are calculated from game-by-game box scores via wehoop. Possessions use the standard estimate:

HOOPER measures self-creation — how often a player generates their own shot versus catching and finishing. It is a complementary metric to MLAPM, not a replacement. A player can be a high HOOPER creator with modest MLAPM, or an elite role player with low HOOPER and high MLAPM.

All three components are expressed as above/below league average. League averages: Uast% 31.8%, TS% 54.2%, Vol/36 5.5. Minimum 30 FGM to qualify. Unassisted% is sourced from 2025 WNBA play-by-play data via wehoop. Correlation with MLAPM: r = 0.46.

ADV measures how often a player forces the defense to make a decision before a shot exists — drives that collapse the paint, pull-ups that demand closeouts, fouls that punish overplay, kick-outs that create open 3s. High ADV = the engine that breaks defenses. Low ADV = catch-and-finish role player.

All rates are per 36 minutes, expressed above/below league average. Sourced from 2025 WNBA play-by-play via wehoop. Shooting fouls linked to the fouled player via adjacent free throw sequences. Kick-out assists identified as assisted catch-and-shoot jump shots (excluding pullups and turnarounds). Correlation with MLAPM: r = 0.365. Minimum 350 minutes to qualify.

Azura Stevens (LA, 2025): Full MLAPM could not be calculated due to a name encoding issue — her name is stored as "Azurá Stevens" (with accent) in the ESPN play-by-play feed, causing the RAPM join to fail. As a result, MLAPM_V5, O-MLAPM, and D-MLAPM are listed as null. Her Box BPM (+5.0) is displayed as a proxy and reflects her strong individual box score production. This will be corrected in the 2026 pipeline by normalizing all player name encodings before RAPM joins.

No metric is perfect. MLAPM has known limitations we're transparent about:

Team inflation — Despite lineup adjustment, players on elite teams (particularly Minnesota in 2025) may be slightly overrated. The 0.91 retrodictive correlation means the metric isn't perfectly capturing team vs individual contribution.

Small samples — Players with fewer than 500 minutes have less reliable scores. The 350-minute threshold catches most of these cases but some noise remains.

Defensive measurement — All impact metrics struggle to measure defense accurately. Steals and blocks are visible but most defensive value (positioning, communication, deterrence) is invisible in box scores.

Injury effects — A player returning from injury may underperform their true ability. Multi-year RAPM helps stabilize this but single-season anomalies can still distort scores.