esc
it's a feature, not a bug
back to models
14 Statistics

regression to the mean

by Francis Galton

Core Idea

Extreme outcomes tend toward average over time. Exceptional performance is partly luck; future performance regresses toward the mean. Don't overreact to outliers.

key principles

  • 01

    Extremes don't persist

    Very good or very bad outcomes rarely repeat at the same intensity. The exceptional tends toward the ordinary.

  • 02

    Luck disguises itself as skill

    When performance has a random component, extreme results likely include extreme luck. Next time, luck normalizes.

  • 03

    Don't overreact to outliers

    The rookie sensation, the terrible quarter, the exceptional year—these often mean less than they appear.

  • 04

    Base rates matter

    The long-run average is the best predictor when you don't have strong evidence otherwise.

applications

Sports
Sophomore slump
Rookie of the year often underperforms next year. Not because they got worse—the luck component normalized.
Investing
Hot hands cool
Last year's best fund usually isn't this year's. Extreme performance includes extreme luck.
Business
Quarterly results
One great quarter doesn't mean transformation. One bad quarter doesn't mean crisis. Regression happens.
Hiring
Interview performance
The exceptional interview candidate may be experiencing exceptional interview luck. Expect regression.
Related Models
second-order thinking
Howard Marks
circle of competence
Warren Buffett

The Statistical Reality

Galton discovered regression studying height: very tall parents tend to have children shorter than themselves (though still tall). Very short parents have taller children. Extreme values regress toward the population mean.

This isn’t about causation—it’s about the structure of variation. When outcomes combine skill and luck, extreme results have extreme luck components that won’t repeat.