Understanding SHAP Values in Racing

SHAP (SHapley Additive exPlanations) values decompose each prediction into individual feature contributions. This tells you exactly how much each factor pushed the probability up or down.

For greyhound racing, common high-impact features include: recent finishing positions, track-specific win rate, trap draw advantage, and time vs par performance.

When you see a SHAP waterfall chart on GreyhoundAI, positive values (green) push the win probability higher, while negative values (red) reduce it. The base value represents the average prediction across all dogs.

This transparency helps you verify whether the model's reasoning aligns with your own racing knowledge and identify when it might be over- or under-weighting certain factors.