Trap/Cognitive Bias/No. 0476
Hot Hand Fallacy
The hot hand fallacy, or hot hand bias, is the belief that a run of success makes further success more likely without evidence that the process supports it. Studied in basketball, it differs from real hot hands, where a player’s performance can improve.
Also called Hot Hand Bias
- Evidence
- Mixed evidence
- Read
- 6 min
- Links
- 9 connections
01You've seen this when…
- in life
You win four online roulette bets in a row. You increase the next bet because you feel lucky, although nothing about the game has changed.
- at work
After three accepted proposals, a salesperson treats the next deal as almost certain. Nobody checks whether the earlier buyers were unusually easy to win.
- out in the world
A TV commentator calls a basketball player unstoppable after three baskets. The next shot is tightly defended, but the prediction ignores that.
02The idea
A few successes arrive together, and they start to feel like evidence that something has changed. Winning right now feels significant. The next attempt seems more promising because of the last few.
The hot hand fallacy is making that inference when the process doesn’t warrant it. With an unbiased roulette wheel and independent spins, previous wins don’t improve the odds of your next bet. The wheel doesn’t respond to your streak.
This differs from the gambler’s fallacy. The hot hand prediction is that success will continue. The gambler’s fallacy predicts a reversal because the opposite result is supposedly due. For independent outcomes with known probabilities, neither prediction follows from the streak.
Independence needs to be established. Human performance can change, and a streak can reveal new information. A basketball player’s performance can vary while an unbiased roulette wheel’s probabilities stay fixed. The error is treating a streak as sufficient proof of improved prospects, rather than checking what produced it.
03Why it happens
- A cluster looks like a change. We expect randomness to look evenly mixed. Random sequences contain runs, so ordinary clusters can look like exceptional performance. This connects to the law of small numbers bias: expecting a short sample to behave like a much larger one.
- A story arrives before a test. We explain a streak by invoking luck or by describing how confidence or momentum has shifted. Some explanations could be true. A plausible story still needs evidence to establish that it fits the streak. Illusory pattern perception fills the gap.
- Success attracts attention. A winning run gets replayed and discussed. Scattered wins and unsuccessful runs rarely get equal billing, leaving a distorted picture of how often streaks happen.
- The feeling spreads to the forecast. Recent success can make you feel capable and in control. That feeling may become overconfidence about the next attempt, even when the relevant conditions are different.
04A worked example
Imagine you’re playing European roulette on an unbiased wheel whose spins are independent. It has one zero and 36 numbered pockets. After winning four $10 bets on red in a row, you decide to bet $50 on red next because the evening seems to be going your way.
What it looks like Taking advantage of a favorable moment. The four wins seem to distinguish this session from an ordinary one.
What’s actually going on The next spin still has 18 red pockets out of 37, giving your red bet about a 48.6% chance of winning. Your previous wins leave that probability unchanged, so the larger stake exposes more money to the same unfavorable odds.
What would have helped Deciding your spending limit and maximum stake before playing, then sticking to those limits when a streak tempts you to raise them. A run of wins changes how much money you have while leaving the next spin’s odds unchanged.
05How to spot it
06What to do instead
- Identify what could carry over. Practice, physical condition, attention or a shared customer need might link successive outcomes. For an independent random device, there may be nothing that carries over. Name the mechanism before betting on it.
- Write down a rival explanation. Three unusually receptive buyers might explain three sales that seem to signal a salesperson’s temporary improvement. Use the consider-the-opposite strategy to make the competing account concrete.
- Compare comparable attempts. For a shooter, compare shots from the same location while accounting for defense and fatigue. For a salesperson, consider lead quality and offer terms. Compare after-streak performance with an appropriate baseline grounded in data about normal performance.
- Test the streak rule fairly. If you analyze records, choose the streak length in advance. Compare against random sequences of the same length, applying the same selection rule. Searching many streak definitions until one looks impressive makes the test unfair.
- Record forecasts before outcomes. Write down your predicted probability for the next attempt and check it over many attempts. That builds calibration. Even with improved odds, failure remains possible, and a much larger stake needs justification beyond that improvement.
07When it isn’t a fallacy
Hot hands can occur. A person’s performance may temporarily improve, and successive attempts may share favorable conditions. Empirical research must establish a hot hand’s occurrence and duration to assess whether it helps predict the next result.
A streak can also teach you about someone’s ordinary ability. If you’ve never seen a player before, several successful shots reasonably increase your estimate of their skill. That’s Bayesian updating: revising your estimate of ordinary ability. Concluding that they’re better now than they usually are requires evidence of a temporary improvement.
Even an improvement may be too small to support the popular story. Slightly improved shooting still calls for considering defense before choosing a shot. Improved sales performance can still leave some prospects unlikely to buy.
The useful rule is to examine streaks by separating evidence of a changed state from an appealing interpretation of a short sequence.
08Roots
In 1985, Thomas Gilovich collaborated with Robert Vallone and Amos Tversky to examine a familiar basketball ritual: after a player makes several shots, fans want the ball returned to that player. They surveyed people and examined basketball performance using records and a controlled shooting experiment to investigate whether the belief matched performance. Their paper became a classic account of people finding meaningful patterns in random sequences and helped establish the hot hand fallacy as a textbook example.
Three decades later, Joshua Miller and Adam Sanjurjo found a problem in an influential way of testing the claim. The test measured the success rate of shots selected because they immediately followed streaks of makes. In short sequences, that selection rule introduces bias: averaging the selected rates can produce a number below the true success probability even when shots are independent. That makes improvement harder to detect.
Their 2018 paper explained the bias and reported evidence of hot-hand shooting in the original controlled experiment after correcting for it. Data long cited against the hot hand now supported it. The meaning of any particular winning streak still depended on evidence. It changed the burden of analysis: testing a streak fairly requires distinguishing human performance from independent chance while keeping both belief and skepticism tied to the evidence for each case.
09How solid is this?
Unsupported streak-based judgments are documented, and hot hands can occur. Miller and Sanjurjo identified selection bias in influential early analyses and found evidence of hot-hand shooting in the original controlled experiment after correction. The existence and predictive value of a hot hand depend on the task and conditions.
10Connections
- Often confused with Gambler’s Fallacy
- Countered by Calibration, Consider-the-Opposite Strategy, Bayesian Updating
- Can lead to Overconfidence Effect
- Part ofIllusory Pattern Perception
- See also Independence, Law of Small Numbers Bias, Selection Bias
11Origin and sources
Thomas Gilovich, Robert Vallone and Amos Tversky’s 1985 basketball study established the classic account. Joshua Miller and Adam Sanjurjo’s 2018 paper identified a selection bias in influential tests and revised the interpretation of the original experiment.
- [1]Gilovich, T., Vallone, R., & Tversky, A. (1985). The hot hand in basketball: On the misperception of random sequences. Cognitive Psychology, 17(3), 295–314.
- [2]Miller, J. B., & Sanjurjo, A. (2018). Surprised by the Hot Hand Fallacy? A Truth in the Law of Small Numbers. Econometrica, 86(6), 2019–2047.
Suggest an edit· Updated 2026-10-02