What Is Regression to the Mean? (Definition and Examples)

Regression to the mean is the statistical tendency for extreme outcomes to be followed by less extreme ones. In any process combining genuine skill with random variance, extreme observations oversample from the favorable end of the variance distribution. The next observation draws independently from the variance — and therefore tends to be closer to the average than the extreme observation was.

The Mechanism

Any performance = skill + random variance.

When a performance is extremely high, it reflects a high skill component AND a favorable variance draw. In the next period, the skill component stays roughly constant, but the variance draw is independent — it's as likely to be average as to be favorable. So the next performance tends to be: same skill + average variance = less extreme than the peak.

This is not a force pulling outcomes toward the center. It's the arithmetic of independent draws. Extreme observations oversample from favorable draws. Independent subsequent draws don't replicate the extreme.

The Sports Illustrated Illustration

The "Sports Illustrated cover jinx" is regression to the mean: athletes appear on the cover at performance peaks, which are disproportionately lucky — the best health, the best opponents, the best conditions coinciding. The following period draws from independent conditions and independent luck. It tends to be less extreme than the peak.

There's no curse. There's only the mathematical fact that peaks are followed, on average, by less-extreme observations.

The Hiring Implication

Hiring based on extreme past performance systematically selects for the luck component of a lucky period. The "star analyst" whose last year was the best on the desk is, in expectation, going to produce average results in the next period — because the lucky variance that contributed to their peak doesn't repeat, and the peak itself was a selected extreme.

The correction: evaluate performance over enough independent observations, across different conditions, to reduce the fraction of the observed result attributable to a single favorable draw.

For the full framework, read Fooled by Randomness: How Luck Masquerades as Skill.