🐻⬇️🏀

Model Card

Sequential event model on made threes per 100 possessions. Positive values mean a player's minutes tend to raise their team's made-three volume.

Validation here is out of sample. From two weeks into each season, a model trained on earlier games predicts each following week without learning from that week. Predictions use the players' observed minutes; each point is the minutes-weighted predicted and actual rate in that player's games. When validation was computed, its source snapshot was verified against this published model revision.

Selected League / Season

Change league with the League selector at the top of the page.

Held-out player predictions

Plot Threes For Elo against any counting stat (one point per player, every season with a served list).

NBA 2026-2027: 184 players with at least 100 held-out minutes. Across players, predicted and actual correlate at r = 0.81 (weighted by held-out minutes). Across these player averages, weighted R² = 0.43 (relative to the weighted mean of actual held-out outcomes). Per held-out team game, squared error is 13.2% below predicting the league average. Each dot is a player (184 with the most held-out minutes are plotted; larger dots had more); on the dashed line, actual equals predicted. Select a dot for the player.

Held-out miss against playing time

Each dot is a player's actual minus predicted team threes made per 100 possessions in the player's games against their held-out minutes (log scale). Orange points are the minutes-weighted mean miss of eight equal groups of players by playing time: a model that shrinks its low-minutes players too far, or not far enough, shows a slope.

Stat RAPM or event Elo, per season

Each season serves one method for this stat: stat RAPM when its predictions followed the actual per-segment margin better on held-out games (fit before a date, scored on later segments) by more than two paired standard errors; within that the two are tied and the event Elo serves. Correlations are small in absolute terms; a segment holds about one such event.

Output Coverage

Rows in each model's current published run (or legacy rows where no run is selected).