The “hot hand” is the belief that a person or entity that has experienced a run of recent successes (a “hot streak”) is more likely than usual to continue succeeding on the next attempt. The same intuition in reverse is called a “cold hand.” People apply this idea to sports (a shooter “on fire”), gambling (a streak of correct coin guesses), and investing (a fund manager “on a roll”).
Key takeaways
– The hot-hand belief is common and intuitive, but treating independent events as dependent creates errors.
– Classic research treated the hot hand as a fallacy — a cognitive bias driven by heuristics and small-sample misperception.
– More recent statistical work (notably Miller & Sanjurjo, 2018) shows that some earlier analyses underestimated how streaks behave in small samples and that modest hot‑hand effects can exist in specific sporting contexts.
– For investors and bettors, the practical test is whether there is replicable, out-of-sample evidence of persistence (skill) rather than luck or selection effects.
– Best practice: assume independence unless you can demonstrate a mechanism and statistically significant persistence after correcting for sample and selection biases.
How the hot hand works (why it feels real)
– Representative heuristic: people expect short sequences to “look” like the long-term process (e.g., heads and tails should alternate), so runs stand out and feel unlikely. That makes streaks salient and memorable.
– Small-sample illusions: short-run randomness produces apparent streaks; human pattern-seeking exaggerates them.
– Behavioral consequences: a perceived hot hand can produce overconfidence, excessive risk-taking, confirmation bias (seeking confirming evidence), illusion of control, recency bias, and hindsight bias — all of which can damage decision-making in investing and gambling.
Evidence for and against the hot hand
– Against: Early academic work concluded that streaks are explained by chance — people misinterpret randomness and attribute skill where none exists. That led to the “hot‑hand fallacy” label: treating independent trials as dependent.
– For (qualified): Miller & Sanjurjo (2018) re-examined earlier analyses and identified a subtle small-sample bias that tends to understate streak effects. After correcting for that bias, they found evidence consistent with a modest hot‑hand effect in some sporting datasets. In short: outright denial of any hot hand was too strong; modest, context-dependent persistence can exist.
– Practical corollary: whether a hot hand is real depends on context (sport-by-sport, player-by-player), sample size, and the presence of plausible mechanisms (e.g., changes in confidence, defensive adjustments, physical state).
Implications for investing and gambling
– Investing: Investors often chase recent winners (fund managers, stocks) because of perceived “hotness.” Research shows this is a flawed heuristic: past short-term outperformance often reflects luck, not persistent skill (investor behavior studies show fund-following is driven by track record but overweights recent performance). Chasing managers without evidence of skill can lead to underperformance after fees and turnover.
– Sports betting: Legalization and mainstreaming of sports betting increases incentives to build strategies that exploit any real persistence. But bettors must distinguish true edges from noise and account for market pricing.
– Decision risk: Acting on perceived streaks without rigorous evidence increases exposure to overtrading, higher costs, and poor risk-adjusted returns.
Practical steps — how to evaluate and act (for investors, bettors, analysts)
1) Start with a default assumption of independence
– Treat trials/events as independent unless you can demonstrate otherwise with data and a plausible mechanism. That prevents overreacting to short streaks.
2) Look for a mechanism
– Ask: Is there a plausible reason performance would truly change (skill, fatigue, matchup, strategy change)? If not, favor a random-noise interpretation.
3) Use rigorous, out-of-sample testing
– Collect sufficiently large datasets. Test whether success probability after a streak differs from baseline in new data (not the sample that suggested the effect). Use cross-validation or holdout samples to avoid overfitting.
4) Correct for small-sample and selection biases
– Be aware of small-sample biases (Miller & Sanjurjo, 2018) that can distort estimates. Use permutation tests, bootstrapping, or analytics that adjust for the specific bias structure in streak detection.
5) Define streaks and baselines clearly
– Specify what constitutes a “hot” streak (length, context). Compare conditional success rates (after a streak) to an appropriate baseline that controls for opportunity, difficulty, and selection effects.
6) Control for confounding factors
– In sports: control for shot quality, defender strength, game context, fatigue, and role changes. In investing: control for exposures (style, sector), market conditions, and survivorship bias.
7) Use simple statistical checks
– Compute the conditional probability of success after a k-length streak vs overall probability. Use hypothesis testing (with small-sample corrections) and report confidence intervals. If you find an effect, test robustness across time periods, opponents/market regimes, and out-of-sample.
8) Manage risk if you act on streaks
– Position sizing: limit exposure to any single “hot” signal. Consider Kelly or fractional Kelly sizing if you can estimate edge and variance.
– Pre-commit rules: set entry/exit rules and limits before chasing streaks to reduce emotional bias.
– Costs and slippage: include transaction costs, taxes, market impact, and betting vig — many apparent edges vanish after costs.
9) For mutual-fund or manager selection
– Don’t chase short-term outperformance. Evaluate longer horizons and look for persistent skill measures (risk-adjusted returns net of fees, style-adjusted alpha). Be skeptical of short track records. Research shows investor decisions are overly influenced by managers’ recent performance (Goetzmann & Peles, 1997; Capon et al., 1996).
10) For bettors
– Bet only when you have a measurable, positive expected value after fees/vig. Track your bets statistically and avoid narrative reasoning (e.g., “player is hot”) unless you can quantify an edge that the market hasn’t priced.
How to test for a hot hand — a practical checklist
– Define event and streak length (e.g., k consecutive successes).
– Assemble a large, well-documented dataset and an out-of-sample holdout.
– Compute P(success | prior k successes) and compare to P(success | prior k failures) and to unconditional P(success).
– Run permutation tests: randomly shuffle outcomes within subjects to generate the null distribution of streak effects.
– Adjust for selection bias (players selected to take shots, survivorship, etc.).
– Control for contextual covariates (difficulty, opponent, venue) with regression or fixed effects.
– Validate any detected effect in out-of-sample periods and different samples.
Practical examples
– Finance: Before reallocating to a manager after a hot quarter, examine multi-year, risk-adjusted performance and persistence measures; require evidence of skill beyond luck.
– Sports betting: Rather than betting because “the shooter is hot,” quantify whether the shooter’s shot quality or opponent defenses changed, and whether market odds underreact to that information.
– Trading: Avoid increasing position sizes solely because recent trades were winners. Base sizing on objective edge and drawdown tolerance.
Conclusion
The hot hand is a powerful, intuitive idea that can be both misleading and, in limited contexts, partially true. Classic studies emphasized that people over-interpret random streaks; more recent work shows that small-sample biases can mask modest persistence in some sporting settings. For practical decision-making, assume independence by default, require transparent mechanisms and rigorous, out-of-sample statistical evidence before acting on streaks, and always manage risk and costs.
References
– Miller, Joshua B., and Adam Sanjurjo. “Surprised by the hot hand fallacy? A truth in the law of small numbers.” Econometrica, vol. 86, no. 6, 2018, pp. 2019–2047.
– Goetzmann, William N., and Nadav Peles. “Cognitive dissonance and mutual fund investors.” Journal of Financial Research, vol. 20, no. 2, 1997, pp. 145–158.
– Capon, Noel; Gavan J. Fitzsimons; Russ Alan Prince. “An individual level analysis of the mutual fund investment decision.” Journal of Financial Services Research, vol. 10, no. 1, 1996, pp. 59–82.
– American Gaming Association. “97% of Expected $10 Billion Wagered on March Madness to be Bet Illegally.” (accessed Jan. 25, 2022).
– Supreme Court of the United States. Murphy, Governor of New Jersey, et al. v. National Collegiate Athletic Association, et al., May 2018 (decision easing federal prohibition on commercial sports betting).
– Walk through a concrete example (sports or trading) and run the simple conditional-probability tests described above on sample data.
– Provide a short checklist you can print and use before increasing exposure based on a perceived hot hand.