Team stats

What a rolling average shows

Season averages hide how a team is playing right now. A rolling average over the last 5 or 10 games tracks form as it changes: a shooting slump, an injury that reshapes the rotation, a defense that tightens after the trade deadline. The 5-game window reacts quickly but is noisy; the 10-game window is steadier but slower to notice a real change.

These are the model's inputs

These rolling values are built exactly the way the prediction models' inputs are: before each game, each team's recent box scores are condensed into rolling averages covering traditional stats, the four factors (shooting efficiency, turnovers, offensive rebounding and free-throw rate), hustle stats like deflections and contested shots, and player-tracking data. Comparing two teams here is a good way to see what the models were looking at when they made a pick.

To see which of these stats actually move the needle on winning, see the What wins games page; to look at how two stats relate across every game, try the Stat scatter.