The Hot Hand Fallacy: Why Streaks Are Random and Your Brain Refuses to Accept It

Somewhere in the third quarter, a basketball player hits three shots in a row. The crowd feels it. The commentators feel it. The coach calls a play to get him the ball again. Everyone in the arena agrees: he's hot.

Photo by SHVETS production on Pexels.

He misses the next one.

This plays out millions of times a year across sports, finance, sales teams, and poker tables. The hot hand fallacy is the belief that a person who has experienced recent success has a higher probability of further success. It feels so obviously true that questioning it sounds absurd. Streaks happen. Momentum is real. Everyone has seen it.

Except the data keeps not cooperating.

Where the Research Started

In 1985, psychologists Thomas Gilovich, Robert Vallone, and Amos Tversky published a paper that scandalized the sports world. They analyzed shooting data from the Philadelphia 76ers and found that players' hit rates after a successful shot were statistically indistinguishable from their hit rates after a miss. The "hot hand" wasn't there. Players who had just hit three in a row were no more likely to hit the next shot than their baseline average predicted.

Sports fans were furious. Athletes pushed back. The paper was called naive, ignorant of the game. Bill James called it one of the most controversial findings in sports research. People simply could not accept it.

The problem is that human brains are pattern-recognition machines running in overdrive. Show someone a sequence like H-H-H-T-H-T-T-H-H-T, which is genuinely random, and they will find the clusters suspicious. Then show them T-H-T-H-T-H-T-H, which looks clean and alternating, and they'll call that one fake. Real randomness produces runs and clusters. It looks lumpy. Our intuitions about what random sequences should look like are systematically wrong.

The Lumpiness Problem

Here's what a fair coin actually produces over 20 flips: you'll get a run of four or five heads roughly half the time you try. That's not unusual. That's just how randomness distributes itself.

When we see those runs in a basketball game, we attribute them to something. A shooter finding his rhythm. Defenders losing focus. The ball feeling right. We construct a narrative around what is, statistically, an expected cluster in a random sequence.

graph TD
    A[Random sequence generated] --> B{Does a streak appear?}
    B -->|Yes - expected ~50% of the time| C[Brain attributes streak to skill/momentum]
    B -->|No streak| D[Brain notices nothing]
    C --> E[Hot hand belief reinforced]
    D --> F[No story to tell]
    E --> G[Behavior changes: more passes to hot player]

Notice the asymmetry. Streaks get explained. Non-streaks get ignored. The belief gets reinforced every time a cluster appears, and there's no feedback loop correcting it when clusters fail to predict future performance.

The 2016 Comeback

The story got complicated. In 2016, economists Joshua Miller and Adam Sanjurjo published a correction to the original Gilovich paper and found a subtle but real statistical bias in how the original study was analyzed. When you condition on previous hits in a finite sequence, you're sampling from a distribution that slightly underrepresents subsequent hits. The 1985 paper, they argued, used a biased estimator.

After correcting for this, they found evidence of a small but genuine hot hand effect in the original NBA data. Roughly three to four percentage points. Real, but modest.

So the full picture looks like this: a small hot hand effect probably exists in some contexts, particularly in sports where confidence and physical rhythm play a role. But the hot hand that fans, coaches, and commentators believe in? The dramatic momentum shift that justifies doubling down on the hot player? That version is wildly overstated.

We see a three-point streak and infer a 70% shooter. The data says he's still a 45% shooter having a normal cluster.

Why This Costs You Money

The hot hand fallacy doesn't stay in arenas. Investors chase funds that had a strong last quarter. Managers promote salespeople on hot streaks and pull budget from those in slumps without checking whether the variation is meaningful or just noise. Poker players go on tilt because they're convinced a run of bad cards "has" to turn.

Regression to the mean is coming regardless. The hot shooter cools off. The fund reverts. The streak ends because it was a streak, not a signal.

The mean is still lying to you, but this time it's the streaks around the mean doing the talking. Your brain hears them loud and clear, assigns them cause and consequence, and builds strategy on top of a statistical artifact.

Three in a row means exactly as much as the baseline says it means. Which is to say: less than you think, every single time.