What matters
A 40-plus-game season, 40-minute games and deep roster turnover make samples small. Minutes and availability dominate, and early-season numbers carry heavy uncertainty. The PBE model does not make calls on teams with fewer than three games of current-season history and flags three to five games as limited history.
Important data
| Input | Why it matters |
|---|---|
| Minutes and availability | Master variables, as in the NBA |
| Current-season history | Thin early; the model flags it |
| Pace and usage | Opportunity scalers |
| Market prices | Measured against, not used as inputs |
Common metrics
| Metric | What it measures | How to read it |
|---|---|---|
| WinBA Score | Published composite of production, win rate, winning-output share and court share | Descriptive, not a pick |
| Box impact | PTS + 1.2 × REB + 1.5 × AST | Component of WinBA |
| Brier / log loss | Holdout scoring rules on the model card | Compare with baselines and the market |
Model concepts
The card reports holdout Brier and log loss for its v1 model, a calibration slope, and a direct comparison with the de-vigged market on priced games — where the market predicted outcomes better. It therefore describes PBE Edge as disagreement with the market, not proven value.
Live intelligence
WNBACast follows possessions, score and box score. The same minutes-first logic as the NBA applies live.
PropBetEdge WNBA tools
Worked example
- Model log loss on priced holdout games is higher than the de-vigged market's.
- So the average model disagreement with the market is not, by itself, evidence of value.
- Useful reading: where the model disagrees, look for the reason; judge disagreements by CLV and graded results over a real sample.
Glossary
Related lessons
Brier & log loss
How do you grade a probability, not a pick?
Reading a model
I am looking at a model output. What should I check, in what order?
WinBA Score
What does a WinBA Score measure, and what does it not?
Calibration
When a model says 70%, does it happen 70% of the time?
Closing line value
Was the price I took better than where the market finished?