Gambler's fallacy
Also known as: Monte carlo fallacy
Believing that after a run of one outcome, the opposite has become 'due.'
What it means
The gambler's fallacy is the mistaken belief that in a sequence of independent random events, a streak of one outcome makes the opposite outcome more likely on the next trial — that chance somehow 'balances out' in the short run. The error stems from a faulty intuition about randomness: people expect even small samples to look representative of the underlying probabilities, so a run of reds 'should' soon be corrected by black, when in fact each spin is statistically independent and the wheel has no memory. It is the flip side of the hot-hand fallacy, where a streak is expected to continue rather than reverse; both spring from misreading the lumpy texture of true randomness. A crucial boundary condition is independence: the reasoning is only fallacious when trials are genuinely independent — with sampling without replacement (drawing cards from a finite deck), the 'due' intuition can actually be correct. It matters because the same flawed logic drives costly betting strategies, misjudged lottery-number picks, and even real-world decisions like a loan officer rejecting an application simply because the previous several were approved.
Why it happens
The leading account traces the error to the representativeness heuristic, the tendency to judge how likely something is by how closely it resembles a mental prototype. A short run of one outcome does not look like the person's image of a random sequence, which is expected to alternate frequently and stay balanced, so the mind treats a corrective outcome as overdue. Amos Tversky and Daniel Kahneman called the underlying assumption the belief in the law of small numbers: people wrongly expect even tiny samples to mirror the properties of the population that generates them, so they read local balance into a process that only balances out over the very long run. On this view the fallacy is a byproduct of an otherwise useful shortcut for detecting pattern in noise. A related strand emphasizes that people are poor generators and poor judges of randomness in general, systematically producing and preferring sequences with too many alternations and too few long streaks, which makes genuine runs feel anomalous and in need of correction.
The original demonstration
The most cited version of the demonstration appears in Tversky and Kahneman's 1974 synthesis in Science, building on the belief in the law of small numbers they had set out in 1971. Asked to judge sequences of coin tosses, people showed a robust preference: a mixed order such as heads-tails-heads-tails-tails-heads was rated more probable than heads-heads-heads-heads-tails-heads, a sequence with too many heads to represent a fair coin, and more probable than heads-heads-heads-tails-tails-tails, which has the same three-to-three split but a clustered, non-random-looking arrangement, even though every specific sequence of six tosses is equally likely. The same intuition, they argued, leads people to expect that after a run of heads a tail becomes more likely, because a tail would make the local sample look more representative of a fair coin. Crucially, they framed this not as a quirk of gamblers but as a general feature of intuitive statistical reasoning shared by naive subjects and trained scientists alike, the latter shown by their tendency to over-trust results from small samples. Matthew Rabin later gave the intuition a formal model, showing how an agent who believes small samples are representative will misread independent draws and, in some settings, over-infer about the process generating them.
What the evidence shows
The fallacy has been documented in the field as well as the laboratory, though the field record is more textured than the headline suggests. Rachel Croson and James Sundali analyzed videotaped roulette play in a casino and found that bets on a color or number rose after that outcome had failed to appear for several spins, exactly the pattern the fallacy predicts, but the effect was concentrated in a minority of players and most bets showed no such tendency. Charles Clotfelter and Philip Cook studied a state numbers game and found that the amount wagered on a specific number fell sharply right after it won and recovered only gradually over the following months, consistent with players believing a recent winner is unlikely to repeat. Dek Terrell's work on pari-mutuel lottery-style games reached a similar conclusion while noting that the effect shrank as the stakes and information available to bettors grew, implying that experience and incentives can dampen it. The most cited field evidence comes from Daniel Chen, Tobias Moskowitz and Kelly Shue, who examined sequential decisions by asylum judges, loan officers reviewing applications, and baseball umpires calling pitches. Across all three they found that a decision-maker who had just made one call was more likely to make the opposite call next, even though the cases arrived in essentially random order, an effect they attributed to the gambler's fallacy. The estimated magnitudes were modest, on the order of a percentage point or two of decisions, but they were consistent across very different professional settings and survived controls for the true merits of consecutive cases, which is part of why the study is influential. Taken together the literature supports the phenomenon as real and consequential while cautioning that it is neither universal nor large: many people show little of it, and expertise, feedback and clear incentives tend to attenuate it.
Related but distinct
The gambler's fallacy is often paired with the hot-hand fallacy, the expectation that a streak will continue rather than reverse, and the two can look contradictory since one predicts reversal and the other persistence. Peter Ayton and Ilan Fischer proposed that both flow from the same misreading of randomness but are triggered by different cues: people expect reversal when they believe the outcomes come from an inanimate, memoryless mechanism such as a wheel or a coin, and expect continuation when they believe a skilled human agent is involved, because streaks then seem to signal competence or momentum. This distinction matters for prediction, because it means the same person can commit opposite errors depending only on whether they attribute the sequence to chance or to skill. It also connects the fallacy to broader work on how attributions of agency shape probability judgments.
Limits and caveats
The reasoning is only mistaken when trials are genuinely independent, and that boundary condition is easy to miss. When outcomes are drawn without replacement from a finite pool, as when cards are dealt from a single deck, a run of one type really does make the other more likely on the next draw, so the sense that an outcome is due can be correct rather than fallacious. The intuition also becomes reasonable when the fairness of the mechanism is itself uncertain: a long unbroken run is weak evidence that a coin or a wheel may be biased, and updating toward the streak actually reflects sound inference, the opposite of the fallacy. Empirically, the effect is heterogeneous, showing up strongly in some individuals and settings and barely at all in others, and it tends to weaken with experience, feedback and stakes. Care is therefore needed before labeling any given anticipation of reversal an error, since the same behavior can be rational or fallacious depending entirely on the structure of the process that produces the outcomes.
Examples
After red comes up five times in a row at roulette, players pile their chips onto black, convinced it is overdue — though black's odds are unchanged at just under half.
A roulette player watches red come up six times in a row and moves a larger stake onto black, reasoning that black is overdue, even though each spin is independent and the odds are unchanged.
In a daily numbers game, wagers on a particular three-digit combination drop sharply the day after it wins and only slowly return over the following weeks, as players assume a recent winner is unlikely to repeat.
A loan officer who has just approved several applications in a row becomes measurably more likely to reject the next one, as if approvals and rejections had to balance out, despite the cases arriving in no meaningful order.
A commuter who has caught green lights at four consecutive intersections braces for a red at the fifth, as though a long good run must soon be paid back, reading a balancing rule into a sequence that carries none.
First described in Named for a 1913 Monte Carlo roulette run.
Key references
- Tversky, A., & Kahneman, D. (1971). Belief in the law of small numbers. Psychological Bulletin, 76(2), 105-110. doi.org/10.1037/h0031322
- Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124-1131. doi.org/10.1126/science.185.4157.1124
- Clotfelter, C. T., & Cook, P. J. (1993). The "gambler's fallacy" in lottery play. Management Science, 39(12), 1521-1525. doi.org/10.1287/mnsc.39.12.1521
- Terrell, D. (1994). A test of the gambler's fallacy: Evidence from pari-mutuel games. Journal of Risk and Uncertainty, 8(3), 309-317. doi.org/10.1007/BF01064047
- Rabin, M. (2002). Inference by believers in the law of small numbers. The Quarterly Journal of Economics, 117(3), 775-816. doi.org/10.1162/003355302760193896
- Ayton, P., & Fischer, I. (2004). The hot hand fallacy and the gambler's fallacy: Two faces of subjective randomness? Memory & Cognition, 32(8), 1369-1378. doi.org/10.3758/BF03206327
- Croson, R., & Sundali, J. (2005). The gambler's fallacy and the hot hand: Empirical data from casinos. Journal of Risk and Uncertainty, 30(3), 195-209. doi.org/10.1007/s11166-005-1153-2
- Chen, D. L., Moskowitz, T. J., & Shue, K. (2016). Decision making under the gambler's fallacy: Evidence from asylum judges, loan officers, and baseball umpires. The Quarterly Journal of Economics, 131(3), 1181-1242. doi.org/10.1093/qje/qjw017