Entropy to the Mean: Winning Streaks Carry a Rising, Measurable Cost: Evidence from College Basketball and Six Additional Competitive Substrates

The hot hand literature asks whether success predicts success. This paper asks the inverse: as a team carries a winning streak into successive games, does the quality of its performance deteriorate as entering streak depth increases, and is that deterioration visible in raw basketball events rather than only in expectation residuals? The regression frame includes all eligible team-games. Entering streak depth is zero when the team carries no winning streak into the game and positive otherwise; the game that ends a streak remains in the frame at the depth carried into that game. Using 72,638 Division-I team-game rows across eight college basketball seasons on a fully repaired panel with point-in-time expectation snapshots from three independent rating systems, we estimate the effect of entering win-streak depth on opponent-adjusted margin residual within team-season fixed effects, with standard errors clustered on team-season. The gradient is present and adverse against every independent expectation ruler: KenPom −0.2797 (SE 0.0235), BPI −0.2865 (SE 0.0233), Torvik −0.1041 (SE 0.0217); Mean(3) −0.2234, Median(3) −0.2432; all intervals clear. The gradient is not a KenPom peculiarity, not explained by simple cumulative-overperformance reversion on the primary KenPom and BPI rulers, and not a within-season calendar artifact. A within-team-season joint model of depth against cumulative streak overperformance leaves depth intact on the two rulers with large gradients (KenPom: 97.9% retained; BPI: 108.7%) while cumulative overperformance collapses to insignificance; on Torvik the decomposition inverts, and the underlying reason remains open. Adding season × week fixed effects retains 97.7% (KenPom) to 101.4% (Torvik) of the flagship gradient. The paper’s central new result is an expectation-free bridge. On essentially the same games, we remove the expectation systems from the outcome entirely and estimate 18 objective basketball channels on entering streak depth with focal and opponent team-season fixed effects and season × week fixed effects. Thirteen of eighteen primary channels clear the standing dark gate; ten move adversely, including aggregate offensive efficiency; three — turnover control, foul rate, and raw fouls — improve. On the restricted 72,650-row expectation-covered comparison run, twelve of eighteen clear the dark gate with every direction preserved and the major production channels sharpening. Cross-substrate evidence is asymmetric. Tennis and soccer provide confirmatory evidence under substrate-appropriate frozen instruments; NHL did not clear its frozen holdout gate; NFL and CFB were stopped at prospective power boundaries before their sealed holdouts were opened; and MLB returned a well-powered null on its primary process test plus feasibility and boundary results on three other adjudicated designs. The duration-burden interpretation is supported. The hypothesis that discipline gains reflect active resistance to duration-linked deterioration is supported as an interpretation, with the plain-English framing “teams are trying to escape entropy” preserved alongside it. The causal mechanism remains open, and the entropic nature of what is being resisted is not established.