August 2024. Connecticut Sun home game, third quarter, A’ja Wilson driving baseline against Alyssa Thomas. Wilson, the eventual unanimous MVP, takes a power dribble and rises for what looks like a routine finish. Thomas, eight inches shorter, two days into her third consecutive game, simply rotates her hips, gets her body inside, and forces the shot off-balance. Wilson misses. The replay shows Thomas’s positioning was textbook — feet planted, arms vertical, body angled to deny the rim without fouling. No counting statistic in any public WNBA database will credit Thomas for what she just did. No steal, no block, no rebound. Just one possession’s worth of contestation that the box score is structurally unable to record. By every measure that exists in mainstream WNBA coverage, the play did not happen. By the defensive RAPM framework that has slowly crept into the analytical conversation, Thomas’s contribution that night was the difference between an A’ja Wilson basket and a Sun defensive stop. Multiply that across forty games of a regular season and you have the analytical case for why Alyssa Thomas is, in 2026, the most underrated player in women’s basketball.
Defensive RAPM — Regularized Adjusted Plus-Minus, applied specifically to defensive performance — is the most ambitious analytical framework currently being developed for women’s basketball. The technique has existed in NBA analytics since the mid-2000s, where it became one of the canonical tools for evaluating defenders whose contributions don’t register in conventional counting stats. The WNBA version has lagged the NBA by roughly a decade, primarily because the underlying data infrastructure (full play-by-play, lineup tracking, possession-level outcomes) only stabilized for women’s basketball in the late 2010s. The 2026 state of public WNBA defensive RAPM is approximately where the NBA equivalent was in 2015 — increasingly available, methodologically robust, and almost entirely absent from mainstream coverage.
I have been writing about basketball analytics since 2014, with substantially more attention to the WNBA in the last three years, and the analytical gap that most clearly distinguishes serious WNBA writing from the rest is the one this article is going to unpack. Defensive RAPM in the WNBA — what it measures, how it works, where the public versions break, and why every serious women’s basketball evaluation in 2026 needs to engage with it, is the subject of this article.
The origin: where defensive RAPM came from
The Adjusted Plus-Minus framework originated in NBA analytics work in the mid-2000s, primarily through Dan Rosenbaum‘s pioneering papers on isolating individual player impact from team context. The basic insight: a player’s plus-minus (the team’s net rating while they’re on the floor) is contaminated by the quality of their teammates, the strength of their opponents, and the bench unit they replace. APM uses a regression framework to disentangle these effects, producing an estimate of each player’s individual contribution per 100 possessions.
The “regularized” version (RAPM) was a methodological refinement, introduced by Joe Sill in 2010 and refined by multiple subsequent researchers. The regularization addresses a technical problem: standard APM produces estimates with very high variance for players with limited minutes, which makes the results unreliable. RAPM applies a Bayesian-style prior that shrinks low-minute players toward zero, producing more stable estimates across the league.
By the late 2010s, RAPM had become a standard input to NBA all-in-one player metrics. FiveThirtyEight’s RAPTOR, ESPN’s real plus-minus, Daniel Myers’ DARKO, and various other proprietary versions all used RAPM as one of the foundational ingredients. The framework worked. The framework was, by 2020, established.
The WNBA version emerged slowly. The first public defensive RAPM work for women’s basketball appeared in the early 2020s through Her Hoop Stats and a small number of independent analyst-bloggers. The methodology was directly inherited from the NBA work; the data was slightly less robust due to smaller season volume (40 games vs 82) and more variable historical data quality. By 2024, the framework had stabilized enough that defensive RAPM for the WNBA was being published in a few public venues, though without anywhere near the visibility of the NBA versions.
How defensive RAPM works: in plain language
The basic mechanic is a regression model that uses every possession in the league as a data point. The independent variables are which players are on the floor (one row per possession, with binary indicators for each player); the dependent variable is the possession’s outcome (points scored by the offense). The regression estimates each player’s contribution to the team’s offensive output and defensive output, controlling for the other players on the floor.
Defensive RAPM specifically estimates each player’s contribution to the team’s defensive performance — points allowed per 100 possessions while that player is on the floor, isolated from the contributions of their defensive teammates. A defensive RAPM of -2.0 means the player’s presence reduces the team’s expected points allowed by 2 points per 100 possessions. A defensive RAPM of +1.5 means the player’s presence increases the team’s expected points allowed by 1.5 points per 100.
The regularization shrinks the estimates toward zero for players with limited samples. A player who has played only 100 minutes might have a defensive RAPM of -0.2 even if their raw on-floor defensive rating is much better, because the regularization is correcting for high variance in small samples. A player with 800+ minutes gets less shrinkage and a more reliable estimate.
The output, for any given WNBA season, is a list of players ranked by their estimated defensive contribution. The top of the list reads very differently from the league’s defensive accolades. The Defensive Player of the Year voting typically goes to high-block, high-steal, high-rebound players — Brittney Griner, A’ja Wilson, Alyssa Thomas. The RAPM rankings often surface players whose contributions are less visible: smart positional defenders, communicators, players whose impact shows in the team’s overall defensive rating without showing in any individual stat line.
The critical component: what RAPM captures that counting stats miss
The single most important conceptual contribution of defensive RAPM is its ability to capture positional and communicational defense — the kind of contribution that doesn’t produce a steal or a block but that shapes the team’s overall defensive performance. Examples of what shows up in RAPM but not counting stats:
Closeout speed and discipline. A wing who consistently closes out hard on shooters but doesn’t bite on pump fakes prevents three-pointers without generating a steal. The metric captures the team-level effect.
Help-defense rotation. A center who rotates correctly to prevent dump-offs after a drive doesn’t accrue a block but reduces the team’s expected points allowed. RAPM credits the contribution.
Pick-and-roll coverage. A guard who fights through screens, a big who hedges or drops correctly, a wing who tags the roller — these coordinated defensive actions show up in team metrics but not in individual counting stats. Defensive RAPM is one of the few frameworks that credits them individually.
Defensive RAPM vs the alternatives: a comparison
| Metric | What it captures | Strengths | Weaknesses |
|---|---|---|---|
| Defensive RAPM | Individual defensive impact, regression-isolated | Credits invisible defense | Requires lots of data; lag for short careers |
| Defensive rating (Dean Oliver) | Points allowed per 100 with player on floor | Simple, widely available | Confounded by teammates |
| Steals + Blocks per 36 | Discrete defensive events | Counting-stat tradition | Misses positional and communicational defense |
| Defensive Box Plus-Minus (DBPM) | Box-score-derived defensive impact | Available historically | Limited to box-score events |
| Opponent FG%% at rim | Rim protection effectiveness | Position-specific metric | Only captures part of the position |
The honest defensive evaluation in 2026 uses multiple frames. Counting stats establish the visible production. Defensive rating establishes the team-level impact. RAPM isolates the individual contribution. The composite produces a defender profile that survives most of the position-specific biases of any single metric.
What the data needs: inputs
Defensive RAPM requires full play-by-play data with on-court tracking (which five players were on the floor for each possession), possession-level outcome data (points scored or allowed on each possession), and sufficient sample size for the regression to produce stable estimates.
The minimum sample for stable WNBA defensive RAPM is roughly 2-3 seasons of data per player. Single-season WNBA RAPM is too volatile because the season is only 40 games, which produces relatively few possessions per player. Three-season aggregations produce reasonably stable estimates for most rotation players.
Her Hoop Stats publishes defensive RAPM tables for the WNBA across recent seasons. Basketball-Reference’s WNBA section provides the underlying play-by-play data for those who want to compute it themselves. The proprietary versions inside teams use additional inputs (defensive scheme tagging, matchup data) that the public versions don’t have.
Building the analysis: a working framework
- Pull the player’s defensive RAPM across the last 2-3 seasons, weighted by minutes.
- Compare to defensive counting stats. Players whose RAPM substantially exceeds their counting-stat reputation are doing the kind of invisible defense the framework was built to credit.
- Cross-reference with on-court / off-court team defensive rating. The team’s defensive rating with the player on the floor versus off should correlate with their RAPM in direction, if not magnitude.
- Look at the matchup data when available. A player who guards multiple positions effectively is structurally more valuable than a position-specialist defender.
- Watch the games. RAPM is a strong statistical signal. The film tells you which specific actions produce the impact.
Where this gets weird: common mistakes
Single-season overconfidence. WNBA seasons are 40 games. Single-season RAPM has substantial variance. The signal stabilizes over 2-3 seasons.
Position-comparison failures. A center’s defensive RAPM is measured against centers; comparing across positions requires careful interpretation. A guard with a -2.5 defensive RAPM is doing different work than a center with the same value.
Team-context confounds. A player on a team with strong defensive scheme can appear to have stronger defensive RAPM than they would in a worse defensive system. The framework partially controls for this but doesn’t fully eliminate the team-context effect.
The “RAPM is everything” trap. Some analytical writing leans too heavily on RAPM, treating it as a definitive defensive ranking. The framework is one input. Combining with counting stats, opponent shooting data, and film evaluation produces more reliable evaluation than any single metric.
Rookie-year overcorrections. A rookie’s defensive RAPM is heavily shrunk toward zero because of the regularization. A rookie can have an outstanding individual defensive season that doesn’t show in their first-year RAPM. Subsequent seasons will adjust.
When defensive RAPM shines: use cases
Identifying undervalued defenders. The framework consistently surfaces players whose defensive contributions are not captured by counting stats. The names that appear in WNBA RAPM top-10 lists are often different from the names in Defensive Player of the Year voting, and the difference is informative.
Roster construction. A team building around a star scorer needs defenders to absorb the defensive load. Defensive RAPM can identify candidates who would shore up a team’s defensive rating in ways that counting-stat scouting might miss.
Award argumentation. Defensive Player of the Year voting in the WNBA, like the NBA, has historically rewarded counting-stat-friendly profiles. Bringing RAPM into the conversation can identify players whose contributions deserve more recognition than the conventional voting captures.
Trade-deadline analysis. A team adding a player whose RAPM is meaningfully better than their counting stats has, in expectation, added more defensive value than the public narrative suggests. The framework helps separate signal from reputation.
A working example: Alyssa Thomas’s defensive career
Alyssa Thomas’s defensive career, evaluated through RAPM, is one of the cleaner case studies of the framework’s value in WNBA writing. Thomas’s counting-stat defensive profile is good but not historically elite — solid steal and rebound totals, modest block volume. By traditional metrics, she’s a strong defender. By Defensive Player of the Year voting, she’s been a regular finalist but not always the winner.
Her defensive RAPM, aggregated across 2022-2024, ranks consistently among the top 3 defenders in the league. The Connecticut Sun’s defensive rating with Thomas on the floor compared to off is one of the largest gaps among regulars in the WNBA. The team’s scheme — switch-heavy, position-flexible, communicational — depends on Thomas’s ability to defend multiple positions and rotate intelligently. The team-level impact is real and large.
The retrospective evaluations of Thomas as a defender, in mainstream WNBA coverage, have historically undersold her impact. The Defensive Player of the Year voting has rotated through more counting-stat-friendly winners. The RAPM framework is the public-facing tool that surfaces the gap between her actual contribution and her public reputation. Writing about WNBA defense in 2026 that doesn’t engage with Thomas’s RAPM profile is, in my opinion, incomplete.
The limits: what defensive RAPM cannot tell you
Defensive RAPM cannot predict individual possession outcomes. The framework is a season-level structural estimate.
Defensive RAPM cannot fully isolate the player from their defensive scheme. A defender in a switch-heavy system will have different responsibilities than one in a drop-coverage system; their RAPM reflects the scheme as much as the individual.
Defensive RAPM cannot capture matchup-specific weaknesses. A defender with a strong overall RAPM may have a specific matchup type where they’re vulnerable; the aggregate number doesn’t surface this.
Defensive RAPM cannot replace film evaluation. The framework identifies which players are having impact; the film tells you what specific actions produce the impact and what specific situations expose limitations.
One additional limit: the public WNBA RAPM infrastructure is still maturing. A few sources publish defensive RAPM consistently; many don’t. The cross-source comparability is improving but not yet equivalent to the NBA. Expect the framework’s public visibility to expand substantially over the next 3-5 seasons.
Frequently asked questions
What is a “good” defensive RAPM in the WNBA?
An elite defensive season-long RAPM is in the -3.5 to -2.0 range (remember, negative is good for defensive RAPM — the player is reducing points allowed). A strong defender is in the -2.0 to -1.0 range. League-average rotation players cluster near zero. Defenders below zero (positive defensive RAPM) are net negatives on the defensive end.
How does WNBA defensive RAPM compare to the NBA?
The methodology is identical. The absolute numbers run on similar scales. The data quality lags the NBA by approximately a decade in terms of length of available archives and consistency across seasons. Expect this gap to narrow as the WNBA’s data infrastructure continues to mature.
Why don’t broadcasts use defensive RAPM?
Two reasons. First, the metric is harder to explain to a casual viewer than counting stats. Second, the WNBA broadcast ecosystem is still building out its analytics integration. Expect more advanced metrics to appear in broadcasts as the league’s viewership grows and ad spend increases.
Where can I find defensive RAPM data?
Her Hoop Stats publishes defensive RAPM tables for the WNBA. Some independent analyst blogs publish their own versions. The proprietary versions inside teams aren’t public. Comparing across sources is valuable since methodologies can differ slightly.
Sources and further reading
- Her Hoop Stats — the leading public source for WNBA defensive RAPM and related advanced metrics.
- Basketball-Reference WNBA — underlying play-by-play data for RAPM computation.
- The Athletic WNBA coverage — Sabreena Merchant and others bringing analytical depth to mainstream coverage.
- WNBA.com/stats — official tracking data including some defensive splits.
- Just Women’s Sports — long-form coverage that increasingly integrates advanced defensive analysis.
The Thomas defensive possession that opened this article — vertical arms, hip-rotation positioning, A’ja Wilson forced into a miss with no counting stat to record it — is the kind of contribution the WNBA’s box score is structurally incapable of crediting. The defensive RAPM framework is the public tool that does. The names at the top of the WNBA RAPM rankings frequently differ from the names in the Defensive Player of the Year voting, and the difference is, in my opinion, the most informative analytical fact about how the league’s defensive performance is currently being measured. For the broader frame on the analytical vocabulary for women’s basketball, our guide to reading the WNBA honestly is the natural foundational read.
