SetProb uses a multi-factor statistical model to calculate win probabilities for every NBA game. Each team receives a probability score and the two always sum to 100%.

The model considers each team's rating history, current form, and matchup conditions to determine who has the statistical edge. Probabilities are recalculated fresh for every game. No human input, no gut feel.

Every game is assigned a confidence tier based on how strongly the model favors one team.

These boundaries reflect where the model's accuracy actually separates into distinct tiers, not a fixed assumption carried over from another sport. Strong Signal picks have been right far more often than picks near the 58% line, so each tier means something different.

Every SetProb NBA prediction is logged and resolved against the actual game result, the same way our tennis predictions are tracked. You'll be able to see exactly how the model has performed, broken down by confidence tier and game type.

Picks below 58% are excluded from the main accuracy stats because at that range the game is essentially too close to call. Preseason games are tracked but reported separately, since teams rest starters and the results do not change team ratings.

The NBA model is newer than our tennis model and its accuracy history is still building as the season progresses. All numbers on the dashboard are based on real resolved games.

Use the My Picks button in the top right to save games you want to follow. Add either team from any matchup and share your picks list at the end of the day.