← Back to Paradoxes

Regression to the Mean

Extreme performances don't last—they regress toward average. This creates powerful illusions: we think praise makes people worse and punishment makes them better, that the "Sports Illustrated jinx" is real, and that placebos actually work. Francis Galton discovered this in 1886—and we still fall for it every day.

📈 Performance Over Time

--
After Highs
--
True Average
--
After Lows

📖 The Stories We Tell

✈️ Kahneman's Flight Instructors

The Scenario

Israeli Air Force instructors believed: "Praise a pilot after a great landing, they do WORSE next time. Criticize after a bad landing, they do BETTER." They concluded punishment works better than praise!

But Kahneman realized: after an extreme performance (good OR bad), the next one will likely be closer to average—REGARDLESS of feedback!

Try It Yourself

Each pilot's TRUE skill is 70/100. Today's scores include random variation. Click a pilot, give feedback, and see their next performance:

🏆 The Sports Illustrated Jinx

Athletes who appear on the SI cover often perform poorly afterward. Is it pressure? Distraction? A literal curse?

No—it's regression to the mean! Athletes make the cover after exceptional performances. Exceptional = unlikely to repeat. Their next performance will probably be closer to their average.

  • Madden Curse: NFL players on the video game cover often have bad seasons afterward.
  • Sophomore Slump: Rookies of the Year often disappoint in their second year. Their first year was unusually good!
  • Oscar Curse: Best Actress winners often divorce soon after. They won during a career peak; regression affects everything.

💊 Medicine's Greatest Illusion

When do you seek medical treatment? Usually when symptoms are at their WORST. What happens next? They improve toward your normal baseline.

This creates the illusion that treatments work—even when they don't! Patients and doctors attribute the improvement to the intervention, not to regression.

Why Placebos "Work": You take a sugar pill at your worst. You improve (you would have anyway). You credit the pill. This is why we need randomized controlled trials—to separate regression from genuine treatment effects!

Alternative medicine thrives on regression. Homeopathy, acupuncture, crystals—all "work" because people seek them at symptom peaks.

🧠 Why We Miss It

Causal Thinking: Our brains are wired to find causes. "She praised him, then he got worse—praise caused it!" We don't see that the outcome was predetermined by the extreme starting point.

Invisible Counterfactual: We can't see what would have happened without our intervention. We only see the regression.

Confirmation Bias: Instructors remember when punishment "worked" (after bad → better). They forget when it "failed" (after average → average).

Galton's Discovery (1886): Tall parents have shorter children (on average). Short parents have taller children. This doesn't mean height is equalizing—it's regression to the population mean! Extreme parents are extreme partly due to luck; their children get different luck.