Definition
Regression to the mean is the statistical tendency for an unusually high or low measurement to be followed by one closer to the average. When an outcome combines a stable underlying ability with a dose of luck, an extreme result usually means the luck was extreme too — and luck does not repeat.
The pattern shows up everywhere outcomes mix skill and chance. A team that wins by a freak margin tends to win by less next time. A test scored brilliantly is often followed by a more ordinary one. The next reading drifts toward the mean not because anything changed, but because the rare lucky component is unlikely to recur.
Why it matters
How it works
The mechanism is simple once outcomes are split into a stable part and a random part. Selecting the most extreme cases selects, by definition, the cases where the random part pushed hardest in one direction. On the next measurement that random push is gone or reversed, so the result moves back toward the centre — with no change in the underlying ability at all.
This produces a stubborn illusion of cause and effect. A treatment applied only to the worst cases will appear to work simply because the worst cases were going to improve anyway. Guarding against the error means measuring against a comparison group or a baseline, not against the extreme moment that triggered the action.