Regression to the mean
Also called Regression towards the mean, Reversion to the mean, Lucky streak.
When a result is part skill and part luck, an extreme result is usually followed by a more ordinary one, whatever you do.
Francis Galton noticed it in 1886: tall parents had children who were tall on average, but less tall than their parents, and short parents had children taller than themselves. He called it regression towards mediocrity. Nothing pulls the children back; the parents were extreme partly by chance, and chance does not run in families.
Daniel Kahneman tells, in Thinking, Fast and Slow (2011), of flight instructors who were sure that praise made cadets worse and shouting made them better. Both were regression: an unusually good or bad landing is usually followed by a more ordinary one.
Whenever results mix skill and luck, the most extreme ones are the luckiest or unluckiest. Expect them to fall back towards the average, and judge on more than one result before you praise, blame or change course.
Why it matters
If you read too much into one extreme result, you credit or blame the wrong things. The skill behind it stays; the luck does not come back on demand.
Away from the game
Praise after a best-ever quarter and blame after a worst-ever one both seem to work backwards, because the next quarter is usually closer to normal either way. Judge people, teams and plans on several results, not one.
Try it yourself
Your best salesperson had a record month, far above her usual. What should you expect next month?
3 questions, a minute or two. Nothing to finish first.
Want more? Practise Weigh odds.
Play it out: the Which bag? lab, about five minutes against bots.
Learn and practise it
- Skill: Weigh odds