Hot Hand Fallacy
Hot Hand Fallacy: Believing that recent success makes future success more likely in independent random events. A basketball player who's made their last 4 shots doesn't have a higher probability on the next shot due to a "hot hand" β if shots are independent, the streak provides no predictive information. But recent research suggests small genuine streaks may exist in skilled performance, making this more nuanced than originally thought.
What Is the Hot Hand Fallacy?β
The Hot Hand Fallacy was first formally studied by Thomas Gilovich, Robert Vallone, and Amos Tversky in 1985, applying statistical analysis to NBA shooting data. They found no evidence of streak shooting β shots were essentially independent, and players described as "hot" were not shooting at higher percentages. The conclusion: the hot hand was an illusion created by the Representativeness Heuristic (expecting small samples to look like population patterns).
However, subsequent research by Miller and Sanjurjo (2018) corrected a subtle statistical error in the original analysis and found evidence of small genuine hot hand effects in controlled basketball free throw data. The current consensus: pure random events have no hot hand; skilled human performance may show small, real streaks due to physiological and psychological factors (confidence, focus, rhythm).
How It Worksβ
The original fallacy:
β After X successes in a row, people assume the next will also succeed
β For truly independent events, this is wrong
β Each shot/flip/trade is independent of the last
The nuanced reality for skilled performance:
β Genuine streaks may exist due to: confidence effects, physical
rhythm/calibration, opponent adjustment lags
β But effects are small (3-7% increase in hit probability)
β Much smaller than the perceived "hot hand" effect
β Betting heavily on a "hot" performer overestimates the effect
The Gambler's Fallacy vs. Hot Hand:
β Gambler's Fallacy: streak β opposite outcome expected (correction)
β Hot Hand Fallacy: streak β same outcome expected (continuation)
β Both wrong for pure random events
β Hot Hand may be slightly right for skilled performance
Three Real-World Examplesβ
NBA Free Throwsβ
Gilovich et al.'s original analysis of free throw data found no significant hot hand β players who made their first free throw were no more likely to make their second than their baseline percentage suggested. Fan belief in hot hands was persistent and strong. Players themselves believed in the hot hand. But the statistical analysis showed it was an illusion in the data examined.
Casino Roulette Bettorsβ
Roulette bettors who observe a colour appearing multiple consecutive times consistently increase bets on that colour continuing (hot hand) or bet against it (gambler's fallacy) depending on how they frame the streak. Since roulette outcomes are truly independent, both strategies are equally wrong. Casino data shows that long streaks of any outcome temporarily increase betting on that outcome β classic hot hand fallacy.
Investment Manager Selectionβ
Investors systematically select investment managers based on recent performance streaks, attributing them to skill. Studies of mutual fund performance find that funds with top-quartile performance in one period revert to average or below-average in the next period at rates consistent with randomness rather than skill persistence. Investors are pattern-matching to streaks that are partly or fully random, then attributing them to skill (hot hand) and investing accordingly.
When to Recognise Itβ
π¨ Hot Hand Fallacy is likely operating when:
- You're betting more on someone because they've been winning lately
- You believe a "momentum" player will continue their recent form
- You're hiring or firing based primarily on recent performance streaks
- You're selecting investments based on recent exceptional performance
β Countermeasures:
- Determine whether the performance is genuinely skill-based or random
- For skill-based domains: acknowledge small genuine streaks but don't overweight them
- For random domains: treat each event as independent with base probability
- Apply base rates: what's the historical average, regardless of recent run?
| Pairs well with | Why |
|---|---|
| Gambler's Fallacy | Mirror image: expecting reversal vs. continuation after a streak |
| Recency Bias | Recency bias overweights recent events, feeding hot hand reasoning |
| Representativeness Heuristic | Both arise from expecting patterns in small samples |
Common Misuses and Limitationsβ
Claiming the hot hand is always a fallacy. The original 1985 paper overclaimed β Miller and Sanjurjo's 2018 correction found small genuine effects in controlled data. For skilled performance, a modest hot hand effect may be real. The fallacy is in overstating the streak's predictive power, not necessarily in believing in it entirely.
Applying it to all domains equally. Pure random events (roulette, coin flips, lottery) definitively have no hot hand. Skilled human performance (basketball, chess, sales) may show small genuine streaks. The analysis requires domain-specific assessment.
Related Modelsβ
| Model | Relationship |
|---|---|
| Gambler's Fallacy | Mirror image β expecting reversal vs. continuation |
| Recency Bias | Recency bias feeds hot hand reasoning |
| Representativeness Heuristic | Both arise from expecting small samples to show population patterns |
Frequently Asked Questionsβ
Does the hot hand exist in any domain?
Probably yes, to a small degree, in skilled performance domains. Free throw data analysis by Miller and Sanjurjo (2018) found genuine hot hand effects of approximately 3β7 percentage points. Research on successful tennis serves, competitive chess, and sales performance also suggests small genuine streak effects. The effect is much smaller than people believe and much less predictive than "go to the hot hand" strategy implies. For pure random events, there is definitively no hot hand.
If small hot hands are real, does that mean betting on hot performers is rational?
Only if the bet is priced appropriately for the small actual effect. In sports betting, odds typically already incorporate "hot hand" intuitions, often overpricing them. The 3β7% genuine effect doesn't justify the degree to which bettors, coaches, and sports commentators weight hot streaks. The practical advice: acknowledge a small hot hand may exist but bet much more modestly on it than intuition suggests.
How does this affect investment decisions?
Investment manager selection based on recent performance streaks is largely hot hand fallacy. Research consistently shows that past fund performance predicts future performance only weakly (if at all) beyond what chance explains. The exceptions β genuine persistent skill β are hard to identify in advance and rare. The practical implication: select investment managers based on strategy, process, and cost (factors more persistent than recent returns), not primarily on recent performance streaks.
Further Readingβ
- Gilovich, T., Vallone, R. & Tversky, A. (1985). "The Hot Hand in Basketball." Cognitive Psychology β the original paper
- Miller, J. & Sanjurjo, A. (2018). "Surprised by the Hot Hand Fallacy?" Econometrica β the correction
- Kahneman, D. (2011). Thinking, Fast and Slow β streaks, randomness, and pattern recognition
Apply with AIβ
π Evaluate streak-based decisions with MindMax β
This page is part of the MindMax Mental Models Knowledge Base.