Regression to the mean: why extremes do not last

Measure anything that naturally bounces around, then measure it again right after an extreme reading, and the second reading tends to drift back toward normal on its own. Your blood pressure was sky-high at the clinic on a stressful morning. A week later it is lower, and you credit the new habit you started. But it was probably heading back toward your usual number anyway. That drift has a name: regression to the mean.

It happens because most measurements are part signal and part luck. A very high or very low value usually caught the luck at its peak. The signal stays about the same next time, but the luck does not repeat, so the number slides back toward the middle. Nobody did anything. That is simply how random variation behaves when you happen to start from an extreme.

This is one of the great tricks in medicine and self-help. People tend to try a remedy when they feel worst, at the peak of the pain, the flare, the bad week. They almost always feel better afterward, because the worst days are usually followed by less-bad days no matter what. The cold cure, the back treatment, the supplement, each takes credit for a recovery that was already coming. It is why testimonials are so convincing and so unreliable.

This is exactly why good studies use a control group. If the treated people improve and an untreated comparison group improves just as much, the improvement was regression to the mean, not the treatment. So when someone says they were at their worst, tried a thing, and got better, believe the getting-better part and hold the because-of-this part loosely. Better compared to what? Without a comparison, you cannot tell the cure from the calendar.