Consistency

How steady is a player or a team from game to game? The Consistency score compares each player's game-by-game HF spread with what his average HF and minutes would predict, so it measures steadiness itself, not role. 50 is typical, higher is steadier. Small samples are pulled toward 50. Teams are scored on the spread of their game margins, adjusted for how good they are.

Min GP Min MPG include small samples

Floor/Ceiling = 10th/90th percentile HF game. Boom = games above 125% of his average, Bust = below 60%. Games under 10 minutes are ignored. 3-yr blends the last three seasons. * = fewer than 40 games. Updated 2026-10-07.

What the numbers say (2046-2049 box scores)

Raw "consistency" is mostly role. A player's coefficient of variation correlates -0.84 with his minutes and -0.82 with his average HF, so benchwarmers always look volatile and 38-minute stars always look steady. The score on this page removes that.
What is left is a real, moderately stable trait. Role-adjusted volatility repeats year to year (r = 0.52) and half a season predicts the other half (r = 0.37, ~0.54 for a full season), which is why small samples are pulled toward 50.
Shot profile drives it, not ratings. Three-point-heavy, high-volume shooters are the volatile ones (r = +0.31 with 3-point share); rebounders and shot-blockers are steady (r = -0.24 and -0.23). Ratings explain only ~4% of it, so it has to be read from game logs.
Minutes are a separate source. About 29% of a player's game-to-game HF swing is just his minutes moving (blowouts, foul trouble). Minutes spread is its own stable trait (year-over-year r = 0.58).
Playoffs: role-adjusted volatile players produce about 6% less in the playoffs than in the regular season, while steady players hold their level (p = 0.008 after controlling for average HF and minutes).
It does not predict next season's production (p = 0.39) - it is about how a player delivers, not how much he will deliver next year.
Teams: game-margin spread runs 14.6 to 21.7 points. A team's win% is predicted almost perfectly by its average margin and that spread (r = 0.98), and spread hurts good teams: the margin-times-spread term is significant (p = 0.002). "Wins from consistency" below is how many wins a team gained or lost versus having a league-average spread. Team spread is a trait too (year-over-year r = 0.47).
Upsets: in 60 playoff series the underdog won 36% of the time when it was the more volatile team and 26% when it was steadier. The direction fits the theory but the sample is too small to call it.