Statistics & Data
Why Your Best Day at Work Is Probably Followed by a Worse One (and What That Reveals About Everything)
The Israeli Air Force once concluded that praising pilots made them fly worse — and they were completely wrong, in a way that changes how you should interpret almost any pattern you notice.
The Idea
Regression to the mean is one of the most quietly powerful forces in statistics, and one of the most reliably misunderstood facts about the world. Here is the core of it: extreme outcomes — very good or very bad — tend to be followed by less extreme ones, not because anything has changed, but because extreme outcomes are partly caused by luck, and luck doesn't hold. When luck regresses toward average, so does the result. The trap is that we are wired to see cause and effect. So when something extreme happens and then moderates, we reach for a story. A student bombs an exam, gets extra tutoring, and improves — was it the tutoring, or would the improvement have come anyway? A company has its worst quarter, hires a new CEO, and rebounds — was it the leadership, or regression? A patient tries a new remedy during a flare-up of chronic pain and feels better — was it the remedy, or the almost inevitable return toward their baseline? The psychologist Daniel Kahneman, who helped bring this idea into mainstream thinking, describes it as one of the most important statistical concepts a person can absorb. It doesn't mean that tutoring, leadership, and medicine don't matter. It means that without a careful comparison — a control group, a counterfactual — you cannot tell what worked and what was just the universe levelling out. We chronically over-credit interventions that happen to coincide with natural rebounds, and we punish apparent regressions that would have happened anyway.
In the World
In the 1960s, Kahneman was teaching psychology to Israeli Air Force flight instructors, and one of them made a point that seemed completely reasonable at the time. He said he had noticed that when he praised a trainee for an exceptionally smooth landing, the next landing was almost always worse. But when he reprimanded a pilot for a rough landing, the next one was usually better. His conclusion: praise degrades performance; criticism improves it. Kahneman recognised immediately what was actually happening. The exceptionally smooth landing was partly skill and partly a lucky run of conditions and small instinctive corrections. The next landing was almost guaranteed to be less exceptional — not because of the praise, but because perfection is rare and doesn't repeat. The rough landing, similarly, involved an unlucky cluster of factors. The next landing would almost certainly be closer to average. The instructor's feedback was just noise layered on top of an inevitable statistical pattern. What makes this story sting slightly is that the instructor's belief was perfectly rational given what he could observe. He had real data. He had decades of experience. He just had no way of seeing the counterfactual — what would have happened without any response at all. This exact structure plays out in sports commentary (the 'curse' of being featured on a magazine cover), in financial journalism (managers who beat the market one year routinely underperform the next), and in medicine (people seek treatment when symptoms peak, then credit the treatment for the natural recovery).
Why It Matters
Once you see regression to the mean, you cannot unsee it — and that is genuinely useful. The next time you change a habit after a terrible week and have a better one, pause before concluding the change caused the improvement. The next time a team performs brilliantly and you restructure it, consider whether you might be about to punish them for their own success. This is not an argument for fatalism or inaction. Interventions do work. Leadership does matter. Medicine does heal. The point is subtler: it is an argument for humility about attribution. Our minds are pattern-matching machines that evolved to find causes quickly, not accurately. Regression to the mean is one of the places where the quick answer and the correct answer diverge most reliably. The practical upshot is to be more demanding about evidence before crediting or blaming anything — including yourself. That terrible run you're having may not mean something is fundamentally broken. And that extraordinary streak may not mean you've cracked the code. Both are partly luck, and the universe is already pulling them back toward the middle.
A Question to Ponder
Think of a time you changed something — a routine, a relationship dynamic, a way of working — right after a low point, and it seemed to help: how would you know whether the change actually made a difference, or whether things were simply due to improve?
Get a new one of these every morning.
Start learning with Thinkable