Mean-Reversion Reality Check
Explain post-peak declines before inventing causes or superstitions
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 96%
The Mean-Reversion Reality Check begins by asking whether a case is being examined precisely because its recent result was extreme. Extreme performance usually combines underlying ability with temporary influences, so the temporary component is unlikely to repeat at the same magnitude. The next observation therefore tends to move closer to the normal baseline even when nothing causal has changed. This creates misleading stories when an award, intervention, criticism, or ritual occurs at the peak. The framework separates statistical expectation from causal explanation: identify the selection event, estimate the baseline, predict the likely reversion, and only then test competing causes against an appropriate comparison group. It is especially useful wherever leaders evaluate people after exceptional results or investors interpret performance immediately following unusually strong returns.
Origin
Gilovich uses the alleged Sports Illustrated cover jinx to explain how regression toward the mean creates superstitious causal beliefs. Extracted from Masters in Business.
Core principles
- 01Extreme outcomes tend to be followed by less extreme outcomes
- 02Selection at a peak creates misleading before-and-after comparisons
- 03Timing does not establish causation
- 04Memorable reversals distort intuitive frequency estimates
How to run it
- 1
Flag the extreme observation
Determine whether the case was selected because performance was unusually high or low.
Pro tip Compare it with the subject's longer-run average rather than the immediately preceding result.
Watch out A peak or trough is an unreliable baseline.
- 2
Identify the selection mechanism
Ask what made this case visible, newsworthy, rewarded, criticized, or eligible for intervention.
Pro tip Look for rules that select only exceptional performers.
Watch out Selection on extreme results mechanically biases the next comparison.
- 3
Estimate the baseline
Use a longer history or comparable population to estimate ordinary performance.
Pro tip Use multiple periods when performance is noisy.
Watch out Do not treat the most recent streak as the new normal without evidence.
- 4
Predict reversion first
Before constructing a story, anticipate that the next result will probably be less extreme.
Pro tip Write down this prediction before observing the next outcome.
Watch out A correctly predicted decline does not identify its cause.
- 5
Test the proposed cause
Compare affected cases with similar cases that did not receive the award, intervention, or alleged trigger.
Pro tip Use a control group or counterfactual whenever possible.
Watch out Temporal sequence alone does not prove causation.
In the wild
Athletes reach the magazine cover after extraordinary success. Their subsequent performance is often still good but less exceptional, encouraging the belief that appearing on the cover caused bad luck.
→ Regression toward ordinary performance explains the apparent jinx without invoking a causal effect.
An investor buys a fund solely because it has just posted an exceptional annual return. Before blaming a later slowdown on publicity or management complacency, the investor compares the result with the fund's long-term baseline and similarly extreme funds.
→ The investor distinguishes expected reversion from evidence of genuine deterioration.
Common mistakes
Using the peak as the baseline
An exceptional observation contains temporary influences and should not anchor expectations for the next period.
Inventing a causal jinx
A visible event occurring between an extreme result and a less extreme one may be incidental.
Calling normal performance failure
A result can regress from extraordinary to merely good without representing collapse.
Is it for you?
Best for
It is best for evaluating performance after unusually strong or weak results in markets, sports, business, or personal metrics.
Not ideal for
It is not ideal when a verified structural change has permanently shifted the underlying baseline.
From the transcript
“There are a lot of superstitions that are essentially a failure to recognize the operation of regression, the Sports Illustrated jinks being one of them.”
“extraordinary success at time one is going to be followed not by abject failure but by somewhat less extreme a run of success afterwards.”
From the episode
Thomas D. Gilovich Talks About Human Behavior
Thomas D. Gilovich