Calibration Certificate No. gambler-s-fallacy

Gambler's Fallacy

The mistaken belief that an independent random process is "due" to reverse after a run of one outcome, so the recently frequent outcome becomes less likely and the absent one more likely.

systematic

documented offset from the normative answer

Status — robustgrounding ± strong

The behavioral phenomenon — negative recency in betting and sequential professional decisions — replicates across laboratory studies and multiple independent field datasets (lottery, casino roulette, asylum courts, loan officers, umpires), so its existence is well-established. What is genuinely contested is the mechanism and scope: whether it is best explained by representativeness/law-of-small-numbers (Tversky & Kahneman) or by rational adaptation to limited experience (Hahn & Warren 2009), and how it relates to the often co-occurring hot-hand belief. Direction can even reverse with game structure (e.g., multi-prize lotteries elicit more hot-hand than gambler's-fallacy betting), so the effect is robust as a behavior but boundary-conditioned.

Calibration record

Clotfelter & Cook lottery field studySharp post-win drop in bets on the winning number, recovering over roughly three months; magnitude reported descriptively (e.g., bets on a drawn number fell and took months to recover). Precise standardized effect size not reported.The amount of money bet on a number fell sharply immediately after it won, then gradually recovered to its prior level over several months — a behavioral signature of the gambler's fallacy (players avoid recently-drawn numbers because they feel 'less likely' to recur). — Charles T. Clotfelter, Philip J. Cook, 1993 · Maryland numbers game wagering records (thousands of draws/numbers)

Sundali & Croson casino roulette field studyOf 139 identified players, a substantial share bet consistently with negative recency; 26 of 112 individual logistic models showed significant dependence on prior outcomes; a chi-square test rejected randomness at p < .0001. No single summary effect size.Players exhibited the gambler's fallacy (negative recency) — betting less on outcomes/colors that had recently appeared — alongside hot-hand betting in other players; the two biases were positively correlated across individuals, suggesting a shared 'subjective randomness' construct. — James Sundali, Rachel Croson, 2006 · 904 roulette spins; 139 identifiable players

Chen, Moskowitz & Shue field study of professionalsAsylum judges: 0.5 pp less likely to grant after a prior grant (baseline); after two consecutive grants, 5.5 pp less likely to grant — a ~19% reduction in approval probability. Loan officers: ~8 pp less likely to approve after a prior approval under flat incentives, with the per-prior-approval effect weaker and statistically insignificant under stronger incentives. Umpires: 1.5 pp less likely to call a strike after a prior called strike (effect roughly doubles for pitches near the strike-zone edge).Decision-makers were less likely to repeat a decision after a streak of the same decision, even though prior decisions were uninformative about the current case — consistent with the gambler's fallacy / law of small numbers. Effects were larger for moderate, less experienced decision-makers and weaker or statistically insignificant under strong accuracy incentives. — Daniel L. Chen, Tobias J. Moskowitz, Kelly Shue, 2016 · 150,357 asylum decisions (357 judges); just under 1 million consecutive called MLB pitches; loan-officer field experiment (thousands of loan reviews)

Source of systematic error

The dominant account (Tversky & Kahneman) is the representativeness heuristic operating via a 'belief in the law of small numbers': people expect even short subsequences of a random process to resemble the long-run distribution, so a streak of one outcome looks unrepresentative and a reversal feels 'due' to restore balance. This treats chance as a self-correcting force, which it is not for independent trials.

Hahn & Warren (2009, Psychological Review) challenge the pure-error view, arguing the fallacy can emerge from rational sensitivity to the statistics of limited experience: given finite memory/observation windows and the long waiting times for streaks, expecting alternation can be a reasonable response to actually-experienced short sequences rather than a misconception. A separate line (Ayton & Fischhoff; Burns & Corpus) frames the gambler's fallacy and the hot-hand belief as 'two faces of subjective randomness,' applied respectively to processes judged inanimate/mechanical versus human/skill-based. Recent work (Xiang, Dorst & Gershman, 2025) examines the robustness and provenance of the effect.

Recalibration procedure

Ask explicitly whether the process has memory. For genuinely independent trials (fair coins, roulette, lottery balls), the next probability is unchanged regardless of history — write it down. For sequential professional judgments, decouple each case from the last (randomize order, hide prior decisions, review in fresh batches) and tie incentives to accuracy; Chen, Moskowitz & Shue found the bias shrank for experienced decision-makers and under strong accuracy incentives.

In the asylum/loan/umpire study, negative autocorrelation was weaker among experienced decision-makers and under stronger pay-for-performance, indicating the bias is reducible through experience, incentives, and structural decoupling of decisions.

Cross-calibrated against

Hot-hand fallacysiblingOpposite prediction (a streak will continue) from the same underlying misperception of randomness; the two often co-occur in the same gamblers and are debated as 'two faces of subjective randomness.' Hot-hand applies more to human/skill processes, gambler's fallacy to mechanical ones.
Representativeness heuristicparentTversky & Kahneman attribute the gambler's fallacy to representativeness — expecting short sequences to look like the parent distribution.
Law of small numbersparentThe specific belief, named by Tversky & Kahneman (1971), that small samples must mirror population statistics; the gambler's fallacy is its most cited manifestation.
Clustering illusionsiblingBoth stem from faulty intuitions about what random sequences should look like; clustering illusion is seeing too much pattern, gambler's fallacy is expecting too much alternation.
Regression to the mean (neglect of)easily-confusedGenuine regression makes extreme outcomes likely to be followed by less extreme ones; the gambler's fallacy is the invalid extension of this 'balancing' intuition to truly independent trials where it does not apply.
Positive vs. negative recencymechanistically-linkedThe gambler's fallacy is the negative-recency case; hot-hand belief is positive recency. They are mutually exclusive for a single prediction but can both appear across a population.

Open the same case on your own draft.