Regression to the Mean
Regress small-sample player stats toward true talent using Bayesian shrinkage
Read the guide →Small sample sizes produce noisy stats. A player batting .400 through 50 plate appearances isn't necessarily a .400 hitter — much of that performance is luck. Regression to the mean blends the observed rate with the league average, weighted by how much data we have.
The formula: Regressed = Baseline + (Observed − Baseline) × n / (n + k), where n is the sample size and k is the regression constant. When n is small relative to k, the estimate stays close to the baseline. As n grows, it converges toward the observed rate.
Regression constants vary by stat. Volatile stats like BABIP (k=820) and batting average (k=910) need large samples to stabilize, while strikeout rate (k=60) stabilizes quickly. The constant represents the sample size needed to place 50% weight on the observed data.
For bettors: Early-season props and futures are often mispriced because the public overreacts to small samples. Use this calculator to get a better estimate of true talent before placing bets on season-long or next-game player props.