When luck plays a role, extreme outcomes are usually followed by more ordinary ones — purely by chance, no cause required. Miss this and you'll credit fixes that did nothing and blame things that mattered.
Regression to the mean: whenever chance is involved, an unusually extreme result tends to be followed by one closer to average — not because anything caused the change, but because the extreme was partly luck.
A pilot instructor noticed that trainees praised for a great landing usually did worse next time, while those screamed at for a bad one usually improved — and concluded praise hurts and criticism helps. He was fooled. A great landing is partly skill, partly luck; the next landing keeps the skill but not the lucky bounce, so it's closer to the pilot's average. Same for the terrible one. Kahneman calls this regression to the mean, and the instructor mistook a statistical certainty for a causal lesson.
Wherever an outcome mixes skill and luck, extremes don't repeat. The best-performing fund this year tends to be more ordinary next year; the worst tends to recover; the athlete on a hot streak cools; the kid who aced one test scores closer to normal on the next. Nothing 'caused' the drop or rebound — the extreme simply contained good or bad luck that didn't persist. We hate this explanation, because our minds demand a story, so we invent causes (the Sports Illustrated 'jinx,' the manager whose tough talk 'worked').
The defense is to ask, before crediting any intervention: would this have improved anyway, just by regression? If you only act on the worst cases, things will usually get better whatever you do — which makes useless remedies look effective and real ones hard to judge. To know if something actually worked, you need a comparison group, not just 'it was extreme, then it normalized.'
It's the silent reason quack remedies, harsh management, and 'we fixed it' stories seem to work — extremes normalize on their own, so without a control you'll credit causes that did nothing.
It's not 'everything averages out, so nothing you do matters.' Real skill and real interventions absolutely move the average — regression just means the LUCK component doesn't persist. The point is to separate the genuine effect from the automatic normalization, which takes a comparison, not a story.
Lock this idea into memory with a 5-minute active-recall session — the science of spaced repetition, no signup.
Try this idea free →