From 2010 on, player WAR includes a play‑by‑play add‑on: same‑season luck‑adjusted box‑prior RAPM (offense and defense), counting only what it says beyond the player’s box‑score WAR. It is zero‑sum within each team, so team totals still match the box build; it moves credit between teammates. 2026 uses the first 182 of 330 games; earlier seasons are box‑score only.
From 2010 on, season WAR includes a play‑by‑play add‑on: same‑season luck‑adjusted box‑prior RAPM (offense and defense), counting only what it says beyond the player’s box‑score WAR. It is zero‑sum within each team, so team totals still match the box build; it moves credit between teammates. 2026 uses the first 182 of 330 games; earlier seasons are box‑score only.
Team Totals
Proj W = Team WAR + replacement wins (12.2% of the schedule — the NBA’s 10‑in‑82, ≈5.4 wins in a 44‑game season) · Δ = Proj − Actual (green ≤ 2, red ≥ 5). Actual records loaded for all seasons, 1997–2026. Since 2010 player WAR includes a zero‑sum‑within‑team play‑by‑play RAPM add‑on, so these team totals are unchanged by it.
Projections cover players on 2026 rosters and Spotrac contracts; 2027 draftees and expansion-draft rosters (Cleveland 2028, Detroit 2029, Philadelphia 2030) aren’t projected. GP reflects the CBA schedule (50 games in 2027–28, 52 from 2029). Players without contract data appear here but are omitted from the Surplus tab. Projected WAR includes a play‑by‑play add‑on: luck‑adjusted, box‑prior RAPM (offense and defense, shrunk toward each player’s box prior; free‑throw luck removed, and for defense opponents’ 3‑point luck too) from every possession of the last three seasons, counting only what it says beyond the box‑score projection. In 2013–26 backtests it trimmed team‑margin error 2.2% (smaller and less certain than the NBA’s 8.5%, so it is weighted conservatively). The 2026 play‑by‑play covers the first 182 of 330 games.
Lineup Lab
Pick a 2027 team and its five, choose who sits, and drop in any player in the league. Each factor is projected from the five players’ individual stats plus how their archetypes play together, fit on 3,799 real WNBA lineups from 1997–2026. Values are lineup differentials (team minus opponent).
Base lineup
Incoming player
Best replacements
Why it changes
Lineups side by side
Methodology
How WNBA WAR is built, what it is tested against, and where it is weakest.
Overview
WAR (wins above replacement) estimates how many wins a player added beyond what a freely available replacement player would have produced in the same minutes. The WNBA build uses the same method as the NBA page, re-fit for the WNBA: each player gets an offensive and a defensive half, built in four steps:
- Turn her individual offensive and defensive ratings (points produced and allowed per 100 possessions) into a win rate with a Pythagorean formula.
- Weight that win rate by the share of her team’s possessions she is responsible for, and by her share of the team’s games and minutes.
- Subtract what a replacement player would produce in the same playing time.
- Scale the season so league WAR plus replacement wins equals the actual number of wins.
From 2010 on, a play-by-play adjustment (RAPM) moves credit between teammates. It never changes a team’s total.
Data
- Player seasons, 1997–2026: Basketball-Reference WNBA advanced stats (games, minutes, ORtg, DRtg, USG%, AST% and style rates), plus FiveThirtyEight’s archive of the same tables for 1997–2019. Traded players appear only as a combined line.
- Play-by-play, 2010–2026: stats.wnba.com event data (via the shufinskiy/nba_data archive), parsed into five-player stints. It covers 99.9%+ of game minutes. 2026 covers the first 182 of 330 games, where the public archive stops.
- Lineups, 1997–2026: Basketball-Reference team lineup pages (top 10 lineups per team-season), for the Lineup Lab.
- Forward-looking: Spotrac contracts; the CBA schedule (50 games in 2027–28, 52 from 2029) and expansion (Cleveland 2028, Detroit 2029, Philadelphia 2030).
Box-score WAR
The ratings are Dean Oliver’s individual offensive and defensive ratings. Each half becomes a win rate against the league-average rating L, with a Pythagorean exponent fit on 2006–26 WNBA team ratings vs win%:
- Credit usage = shot usage + assist usage (0.517 + 0.478 × AST%), rescaled so each team’s minutes-weighted average is 20%.
- Defensive usage is 20% for every player: there is no WNBA contest-tracking data.
- The WAR/44 columns put every season on the 2025–26 44-game schedule.
Play-by-play RAPM add-on (2010 on)
RAPM estimates each player’s offensive and defensive effect from every possession, controlling for the other nine players on the floor. It picks up what box ratings miss.
- Model: the NBA model, with ridge shrinkage toward each player’s box prior (OBPM / DBPM). λ = 1,000, lower than the NBA’s 3,000 because WNBA players log about a quarter of the possessions.
- Luck adjustment: offense with free-throw luck removed; defense with opponents’ 3-point and free-throw luck removed (shooter talent pooled over three seasons).
- Only the residual counts: what RAPM says beyond box WAR per minute. The NBA weights were 1.4–1.6× too high in a simple backtest compared with the full one, so the WNBA weights are deflated by those factors.
- Zero-sum within each team-season in the historical tables. In the projections it uses the three prior seasons, fading as in the NBA (offense 80% a year, defense 25% after year 1).
Backtest on team margin, 2013–26 (172 team-seasons): RMSE 3.73 → 3.65 (−2.2%). The direction matches the NBA result, but the small sample keeps it from being significant (out-of-sample t = 1.1).
Replacement level
Replacement level is the NBA’s win%: 10 wins per 82 games, or 12.2%. In a 44-game season that is about 5.4 wins. Team WAR + 12.2% of the schedule is the team’s implied record on the Team Totals tab.
Projections (2027–2031)
The engines, blend and intervals are the NBA system: a Marcel-style weighted history regressed to archetype with aging curves, gradient boosting on its misses, and historical comps for the range. It runs in NBA-equivalent units, then converts back (games × schedule / 82, minutes × 40 / 48). Games share is the average of the last three seasons; if that is below 65/82, it is averaged with 65/82. Each team’s 2027 minutes are filled to its court time.
One-year-ahead backtest (464 player-seasons; RMSE; ‘repeat’ = carry last season forward):
| Stat | Model RMSE | Repeat RMSE | Model r |
|---|---|---|---|
| ORtg | 8.44 | 12.15 | 0.52 |
| DRtg | 4.22 | 4.70 | 0.57 |
| USG% | 2.98 | 3.30 | 0.80 |
| OBPM | 1.20 | 1.63 | 0.59 |
Surplus value and trades
Surplus = projected WAR × $/WAR − salary, in contract seasons. $/WAR starts from the 2026 market rate: payroll above the $270K minimum divided by league WAR. It grows with the cap (7.8% a year) and is scaled by 44 / games, because longer seasons produce more WAR per player-season. The Trade Machine grades each side on year-1 WAR, contract WAR and surplus. Draft picks aren’t valued yet, and hard-cap legality isn’t checked.
Lineup Lab
Lineup models are fit on 3,799 WNBA lineups from 1997–2026, one each for the four factors and net rating. Each is modeled as the five players’ shrunk individual stats (prior-season blended) plus archetype effects and archetype-pair interactions (ridge, leave-one-season-out CV). Basketball-Reference WNBA data lacks per-100 free throws, so the FT model uses usage-weighted FT rate.
Validation
Team WAR + replacement wins vs actual wins, for teams without mid-season trades (271 team-seasons, 1997–2026): correlation 0.93, RMSE 3.2 wins in seasons averaging 34 games. As in the NBA, the spread is narrower than real records: each WAR above average goes with about 1.7 actual wins.
Limitations
- Small samples: WNBA seasons are 22–44 games, so single-season ratings, RAPM and projections are noisier than the NBA’s.
- Defensive usage is 20% for everyone; defense is mostly the team’s DRtg split by minutes, stops and rebounds.
- 2026 play-by-play covers only 182 of 330 games.
- Traded players have only combined lines.
- Projections can’t yet place 2027 draftees or expansion-draft rosters.
WAR Trade Machine
Assemble a multi‑team trade — each side is graded on year‑1 WAR, WAR over each player’s remaining contract, and surplus value (Spotrac salaries vs. $/WAR). Draft picks aren’t valued for the WNBA yet, and hard‑cap legality isn’t checked.