Regression to the Mean — Thinking, Fast and Slow

Read It

The flight instructor's observation was accurate: after praise, performance often worsens; after criticism, it often improves. But the real question isn't whether rewards and punishments work—it's why we so easily mistake a statistical pattern for a causal one. Extreme performance contains luck, so the next attempt naturally drifts toward the average, regardless of what the instructor did. Misreading regression to the mean as intervention effect makes managers, teachers, and coaches overestimate their own influence and underestimate random fluctuation. The coin-toss experiment reproduced the same pattern without any feedback, proving the point: any outcome mixing skill and luck will regress, because exceptional results require exceptional luck, and luck doesn't repeat.

Draw It

Three layers get tangled in prose: performance itself, changes in performance, and causes of those changes. Regression to the mean describes only the statistical tendency of change; it says nothing about causes. But the brain craves causal stories, so "praise → decline" gets connected automatically. Separating the layers shows the instructor's conclusion skipped the middle layer entirely.

Rethink It

A service had an unusually high failure rate last quarter, and this quarter it naturally dropped. The team credits a refactor or a process change. But if nobody asks how much of that drop is just regression, luck gets mistaken for skill and random variation for replicable experience. Worse, when the failure rate rises again, the team may doubt the very intervention that seemed effective. Distinguishing statistical regression from real intervention effect is the step most often skipped in technical decisions.

Take It With You

Before crediting any intervention, ask: if we had done nothing, would the result have changed the same way? If you can't answer that, you shouldn't be assigning credit or blame yet.