A bottom-up WNBA model. Every player's minutes and scoring distribution is projected, then ten thousand Monte Carlo games roll those projections up into our own moneyline, spread, and total, priced against the market.
Three ways our number differs from the market's, and why each gap is worth betting.
The line is never copied from the market. Each player's minutes and per-possession scoring distribution is modeled, then summed into a team total. A small WNBA rotation means one minutes or usage miss moves the whole number, which is exactly where books are slow to react.
A bottom-up sum models offense only. We solve offensive and defensive efficiency ratios together, so the opponent's actual defense enters every projection. This corrects the team misrankings that create false edges.
Late-season WNBA is decided by what teams are playing for. A Monte Carlo of the remaining schedule gives each team a live playoff probability, so we fade teams that are already eliminated and respect the ones still fighting for a seed. That late-season motivation is what most models miss.
From player ratings to a final edge on the moneyline, spread, and total. Five stages, fully traceable.
The foundation is a rolling rating for every player, built with exponential decay on the calendar so recent games count more than old ones, and the offseason gap doesn't wash out a player's current form.
No market data ever enters a feature. Every input answers one causal question, every value is shifted to be strictly pre-game, and nothing about a game can leak into its own projection.
Scoring rate means nothing without minutes, and minutes are a coaching decision. A separate gradient-boosted model projects playing time per player-game, the single highest-leverage input in a 200-minute, short-rotation league.
The biggest exogenous signal is teammate absence. When a starter is out, those minutes and shots redistribute to specific teammates, and the model moves them up before the market fully adjusts.
Each player gets a full range of outcomes, not one projected number. We model points as an over-dispersed count, with rebounds and assists too, so the odds of a big night or a quiet one come out realistic.
That range is what lets the simulation give real win, cover, and over/under probabilities instead of one fragile guess.
Before simulating, two team-level adjustments are applied. An opponent-adjusted defensive efficiency ratio scales scoring toward the specific defense being faced, solved jointly so offense and defense are separated honestly.
A second layer prices motivation. A Monte Carlo of the remaining schedule gives every team a live playoff probability, so the model knows the difference between a dog still fighting for a seed and one already eliminated.
With every player distribution, the opponent adjustment, and the leverage layer in place, the game is simulated thousands of times. Each simulation draws every player's line, sums the two teams, and records the margin and total.
Those simulations become a probability for every market: win probability for the moneyline, a cover distribution for the spread, and a points distribution for the total. Each is compared to the book's implied price, and only the gaps that survive de-vig become plays.
We ran the model across the full 2025 WNBA season, training only on earlier games and pricing each one against the market.
Every release includes the full model output. Win probability, simulated margin and total, the opponent and leverage adjustments, and the edge versus the book. You can see how we got there before you bet.
Get WNBADaily WNBA moneyline, spread, and total plays with the full simulation behind each one, delivered to Discord.
Moneyline, spread, and total plays go out as lines become available for each slate, timed to capture opening numbers before the market moves toward the model.
In a short-rotation league, one scratch swings the line. Rosters are re-run as inactives drop, and freed minutes are reassigned to specific teammates before the number is finalized.
Every play ships with the win probability, margin, and total, the defense and leverage adjustments behind them, and the edge versus the book.
All picks land in the Whizard subscriber Discord with closing-line-value tracking, so every play is graded honestly against where the market closed.
Available on its own or as part of the All Access bundle, with every other Whizard model included.