The WNBA is a specific sport with specific rhythms. A 44-game season. Fifteen teams now, after expansion in 2025 (Golden State) and 2026 (Toronto, Portland). A top-heavy talent distribution where one All-Star's availability reshapes an entire matchup. Tight rotations where one player can swing 10+ minutes of usage. Our WNBA models are built to respect those specifics, not treat them as edge cases.
Here's some of what's under the hood.
Every team has a shot diet, and the league average tells you almost nothing about how a specific team will score on a specific night. Our models track each team's tendency and efficiency from beyond the arc, in the mid-range and paint, and at the free-throw line, but only after adjusting for the opponent's defense and the situation on the floor.
The technical names: Opponent-Adjusted Three-Point Tendency and Efficiency, Opponent-Adjusted Two-Point Tendency and Efficiency, and Opponent-Adjusted Free Throw Generation and Efficiency. In plain English: when the Liberty face an elite perimeter defense, we're not asking how many threes they usually take. We're asking how many threes they take against teams that defend the arc like this one, and how well they shoot when they do.
The free-throw piece carries more weight than people expect. Officiating crews call games differently across the league, foul tendencies vary widely by team, and the players who get to the line aren't always the players who shoot the highest percentage from the field. Modeling free-throw generation as its own signal, conditional on the opponent's foul tendencies, surfaces real value in the line.
Pace isn't a team trait you can pull from a season average. It's a negotiation between two teams, shaped by who's defending, who's tired, and what the game state demands. Our Defense and Game-Situation Adjusted Tempo Rating estimates expected possessions in a given matchup conditional on both teams' defensive tendencies and where the game is in flow, not just whether a team "plays fast" on paper.
Fatigue gets handled separately. Back-to-backs, three-games-in-five-nights stretches, and cross-country travel all produce measurable performance effects. Our Travel and Rest Adjusted Fatigue Index bakes those realities into every projection. A team coming off a coast-to-coast trip into a noon tipoff isn't the same team that beat you by 15 last week.
Three signals carry a lot of the weight in our WNBA player prop predictions.
Opponent-Adjusted Player Efficiency measures expected points added per possession by player, conditional on lineup context, opponent strength, and game situation. This is the foundation of our prop projections. A player's raw stat line doesn't tell you whether they're creating value or just accumulating volume.
Opponent-Adjusted Turnover and Assist Tendency tracks ball security and playmaking rates after adjusting for defensive pressure intensity. A guard who looks turnover-prone against a top-tier defense may be a model of efficiency against an average-pressure team. We're trying to isolate the player's true tendency from the noise of who they happened to play against.
Recent-Form Weighted Performance is a ridge-regression-based player rating that weights recent games more heavily than older ones. In a 44-game season, three weeks of new performance is a meaningful share of the sample. A short shooting slump or a hot stretch carries real signal, and our models reflect that.
Defense gets flattened in most ratings into a single number. That throws away the most useful information you have. A team that's elite at defending the three-point line is not the same as a team that's elite at protecting the rim, and the matchup math changes completely depending on who they're facing.
We model defense as several distinct signals including Opponent-Adjusted Defensive Efficiency (Threes) and Opponent-Adjusted Defensive Efficiency (Twos). The first measures a defense's true ability to prevent efficient three-point scoring, conditional on the offensive personnel they're facing. The second does the same for two-point scoring. When a high-volume three-point team meets an elite perimeter defense, our models know that's a different matchup than the same team meeting an elite rim-protecting defense, even if both defenses look similar in aggregate.
Two pieces of our WNBA models exist specifically because the W is the W.
Star-Concentration Usage Modeling addresses something every WNBA bettor learns the hard way: a single star's availability shifts team output dramatically in a league with tight rotations. When A'ja Wilson sits, the Aces don't just lose 25 points. Their entire offensive structure changes. Our usage projections account for this concentration, modeling how minutes, possessions, and shot share actually redistribute when a high-usage player is out or in a difficult matchup.
Injury-Adjusted Minute Projections does the bookkeeping. Every projection accounts for current injury reports, expected rotation changes, and known lineup adjustments. Not last week's lineup. Not the season average. In a league where rotations are tight and one player can swing 10+ minutes of usage, getting this right is the difference between a useful projection and a misleading one.
None of these features are revolutionary in isolation. These are concepts the sharpest analytics minds in basketball have been refining for years. What matters is how they're combined, how they're weighted, and whether they're calibrated to the specific sport you're modeling.
That's the work. Building for the league as it actually plays. Weighting signals based on how the season actually unfolds. Treating star availability, schedule density, and shot profile as the real inputs they are, not as afterthoughts.
This is a snapshot, not the full picture. The actual WNBA models have more layers we're not going to show on a blog post. But this is the philosophy that drives every WNBA prediction we produce.
