Hot-Hand vs Gambler’s Fallacy: Why Streaks Fool Us in Games of Chance

Streaks feel meaningful. After a run of good luck, we grow bolder; after a streak of losses, we tell ourselves a win must be “due.” Two competing intuitions emerge from that feeling: the hot‑hand belief (success breeds success) and the gambler’s fallacy (long losses make a win more likely). They are logically opposed — if past wins truly make future wins more likely, then the gambler’s fallacy claim that losses make future wins more likely must be false, and vice versa — yet people ironically often endorse both statements. In sports, the existence of a genuine hot hand remains debated; in pure chance settings like many casino games, any momentum is an illusion. In this post, I use two simple agent‑based models to make that illusion concrete. The first model tests whether streaks change next‑trial probabilities, and the second shows how streak‑driven betting rules affect wealth and survival when the game has a small house edge. At the end, you can try both models yourself and reproduce the patterns I describe.

Several psychological forces make both beliefs feel compelling. Humans are pattern seekers who overweight salient sequences and underweight the dull mass of ordinary outcomes, so dramatic streaks stick in memory and feel diagnostic; small‑sample randomness then looks like a real signal. Cognitive heuristics such as representativeness lead people to expect short samples to mirror long‑run probabilities, so a run of wins invites a “hot” explanation while a run of losses prompts a balancing expectation. Emotional and motivational factors amplify this: confidence after wins encourages belief in momentum, while loss aversion and regret drive revenge bets after losses. Social stories, anecdotes of big wins, and selective reporting (survivorship and confirmation biases) further reinforce both intuitions despite their logical inconsistency. Together, these biases create a feedback loop where vivid experiences and social reinforcement maintain both hot‑hand and gambler’s fallacy beliefs even when statistical evidence contradicts them.

Testing the intuition: do streaks change probabilities?

Imagine you flip a biased coin many times and ask what happens immediately after a short run of heads or tails. Intuitively, people often expect deviation from the coin’s base rate: a player on a roll must be “hot,” or after several losses, a win must be “due.” The first model simulates many short sequences of independent Bernoulli trials and records the empirical frequency of wins that follow observed streaks. What we see is simple but powerful: early in sampling, the continuation rates wobble and sometimes look dramatic; as the number of sampled streaks grows, the empirical continuation rates settle and converge to the true win probability you set. That convergence happens regardless of streak length or whether you examine “wins following wins” or “wins following losses.” The practical message is that short runs produce noisy impressions; only large samples reveal the underlying independence.

This result explains the paradoxical psychology: people overweight memorable short sequences and underappreciate sampling variability. When an early sample produces a few striking continuations, it feels like evidence for a hot-hand or for balancing tendencies, but those apparent effects typically wash out as data accumulates.

Screenshots from the first model showing various streaks; early results can fluctuate but with enough samples the empirical win rate converges to the true win‑rate. Click the image to view the original large version.

Betting on streaks: survivorship, wealth, and risk

This second model explores how simple streak‑based betting heuristics affect gamblers’ survival and wealth over time when the underlying game favors the house by a small margin. The setup compares four gambler types—hot‑hand believers, gambler‑fallacy believers, gamblers who apply both heuristics, and rational bettors—and plots how many of each type remain solvent and how their average chips evolve across ticks.

Every bet is an independent trial with a fixed win probability slightly below 50% (to mimic a casino edge). If a gambler wins, they gain the current bet amount; if they lose, they lose that same amount. Each agent’s bet is never larger than their remaining chips. Hot‑hand believers increase their bet after a run of consecutive wins (simulating boosted confidence); gambler‑fallacy believers increase bets after a run of losses (simulating a belief that results will “even out” and an urge to recover losses). The “both” group follows both triggers; rational bettors keep bet size unchanged except to respect the chips cap.

Because the game favors the house, the long‑run arithmetic is unforgiving: with enough time all gamblers tend to lose their chips. What differs is timing and volatility. Large multipliers for streak escalation create pronounced “pulse” patterns—episodes of big gains followed by steep collapses—for non‑rational gamblers. Empirically the gambler‑fallacy and the combined strategy often blow up fastest: increasing stake right after a losing streak amplifies exposure when expected value remains negative, accelerating ruin. Hot‑hand escalators sometimes delay collapse because they raise stakes only after wins; a subsequent loss resets bet size, which occasionally tempers immediate ruin. Adding caps to bet escalation slows the pace of collapse but typically still leaves non‑rational gamblers more vulnerable than rational fractional staking. Lowering the escalation multiplier likewise slows blow‑up and reduces extreme variance.

What to look for and why it’s interesting

Watch the survival curves and average‑chips plots by gambler type. Notice pulse events in individual trajectories and how they map onto group averages. Compare runs with and without caps, with high and low multipliers, and with thresholds set to 1 versus higher values. These comparisons reveal how the same underlying negative expectation produces very different lived experiences depending on belief‑driven stake sizing.

Simulation of the second model comparing four gambler types (hot‑hand, gambler‑fallacy, both, rational): survival counts and average chips over time show pulse‑like gains and collapses for non‑rational bettors, faster ruin when escalation follows losses, and slower decline when caps or lower multipliers are used. Click the image to view the original large version.

The model produces two lessons that matter in practice. First, streak‑triggered betting does not alter expected value per bet in independent games; it only changes variance and ruin timing. Second, the timing of escalation matters: escalating after losses is usually the riskiest because it compounds exposure at precisely the moment wealth is already depleted. This mechanism mirrors “revenge trading” in financial markets—when traders try to recoup losses by increasing position size after a drawdown, they amplify downside risk, magnify variance, and accelerate account depletion. The agent‑based simulation shows exactly this: agents that upsize stakes following losses accumulate exposure when their bankrolls are weakest, producing faster and more frequent ruin than agents who escalate only after wins or who follow steady fractional staking.

Conclusion

Streaks seduce us, but in independent chance games, they don’t change the odds. Short samples and vivid memories create a persistent illusion of predictability that fuels both the hot‑hand belief and the gambler’s fallacy. Agent‑based models convert that statistical reality into visible individual stories—who survives, who explodes, and how quickly—and give a practical demonstration of why prudence beats intuition in negative‑expectation games.

Reference books

Thinking, fast and slow: Daniel Kahneman’s masterpiece work is an essential reference for anyone exploring judgement and decision‑making because it clearly explains how two mental systems—an intuitive, pattern‑seeking System 1 and a deliberative System 2—produce both useful shortcuts and systematic errors. This book is worth reading for its rigorous blend of experimental evidence and accessible explanations that show why people routinely misread randomness, cognitive biases, overinterpret streaks, and apply misleading heuristics. Kahneman provides the theoretical tools to understand why observers infer meaning in sequences (hot‑hand) and why players expect reversals after runs (gambler’s fallacy), as well as practical insights on when intuitive judgments are likely to go wrong and how deliberate reasoning can correct them.

Fooled by Randomness: The Hidden Role of Chance in Life and in the Markets (Incerto): Nassim Nicholas Taleb’s book examines how chance and luck shape outcomes in markets and life, and how humans systematically misinterpret those outcomes by seeing patterns and causal stories where randomness dominates. Taleb’s framing complements psychological accounts: it shows why observers and participants mistake noise for signal (supporting hot‑hand beliefs) and why misunderstandings about independent random events, offering practical caution about drawing strong conclusions from short sequences.

Standard Deviations: Flawed Assumptions, Tortured Data, and Other Ways to Lie with Statistics: Gary Smith’s book is a clear, engaging guide to how statistical reasoning is misused in everyday life, business, and research. The book walks readers through common errors—misleading graphs, cherry‑picked samples, overfitting, spurious correlations, and the confusion of noise with signal—using concrete examples and straightforward explanations rather than heavy math. A few chapters explicitly examine the hot‑hand and gambler’s fallacies, combining real‑world examples and simple statistical demonstrations to show where intuition goes wrong.

Try the models

When you run the models in a browser, they can be noticeably slower than on the desktop app — increase the tick speed to speed things up.