The Monte Carlo fallacy — also called the gambler's fallacy — is the belief that past independent outcomes change the probability of future ones. A coin that landed heads five times is not "due" for tails; the next flip is still 50/50. A roulette wheel that hit red ten times does not make black more likely on the next spin. Each trial is independent. The fallacy is named after a 1913 incident at the Monte Carlo casino where black came up 26 times in a row on a roulette wheel; bettors lost heavily believing red was "due." The wheel had no memory. The odds did not shift.
The intuition that "things even out" is strong. We see short runs and expect reversion. In truly random processes, reversion applies to the distribution over many trials, not to the next single trial. The law of large numbers says that as the number of trials grows, the sample proportion converges to the true probability. It does not say that a single outcome compensates for prior outcomes. After 26 blacks, the probability of the next spin being black is still roughly 1/2 (or 18/37 on a European wheel). The run was rare; the next spin is not.
Where the fallacy bites: interpreting streaks as predictive, doubling down after losses because "I'm due," or avoiding a bet because "it just hit." In survival and protection contexts, it appears when people assume a threat that has not materialised recently is "less likely" now, or when they treat past luck as a buffer. In deciding and judging, it distorts allocation (e.g. shifting capital or attention based on recent wins or losses) and evaluation (e.g. firing a strategy after a short bad run that is within normal variance). The corrective: treat each decision under uncertainty as a fresh draw from the same distribution unless you have evidence the process has changed.
Section 2
How to See It
The fallacy shows up when people cite recent outcomes to justify the next bet or when they expect "reversion" on the very next trial. Look for language like "due for," "overdue," "bound to change," or "law of averages" applied to a single or short-horizon outcome. The diagnostic: is the process independent? If yes, past outcomes do not alter the next one.
Business
You're seeing Monte Carlo Fallacy when a sales leader reassigns a territory because "they've had a bad quarter and someone else deserves a turn." If territory outcomes are largely independent (or driven by market and execution, not by "turn"), the reassignment is based on the fallacy. The next quarter's probability is not improved by the prior run. The same applies to rotating which product gets the push after a run of losses — unless the rotation is for incentive or information reasons, not because "luck will balance out."
Investing
You're seeing Monte Carlo Fallacy when an investor avoids a stock because "it's up 30% this year and must pull back" or buys another because "it's down 40% and is due to bounce." Short-term price moves in efficient markets do not make the next move more or less likely in a predictable way. The fallacy also appears when doubling position size after a loss ("I'm due for a win") or cutting after a win ("I'll give some back").
Risk
You're seeing Monte Carlo Fallacy when a team assumes that because they have had no safety incidents for two years, risk has decreased. The underlying process (equipment, behaviour, environment) may be unchanged; the run of good outcomes does not reduce the probability of the next incident. Complacency after a long "lucky" streak is the same error in reverse.
Operations
You're seeing Monte Carlo Fallacy when support staff are rotated off a queue because "they've had a run of hard tickets." If ticket difficulty is random or exogenous, the next ticket's difficulty is not affected. Rotation may still be rational for fatigue or fairness — but not because "the next one will be easier."
Section 3
How to Use It
Decision filter
"Before inferring that the next outcome is more or less likely because of recent outcomes, ask: are the trials independent? If yes, treat the next outcome as a fresh draw. Don't 'balance' or 'even out' single decisions; use base rates and process evidence, not run length."
As a founder
Don't rotate people, products, or bets simply because of a short run of wins or losses. If the process is stable and outcomes are independent, the run does not change the odds. Use process changes (better execution, different strategy) and base-rate evidence (long-run conversion, win rate) instead. Avoid "we're due" or "we've had our share of bad luck" as a decision rule. When protecting the company, don't assume that because you haven't been hacked or had a crisis lately, the probability has dropped — maintain defences and margins regardless of recent outcomes.
As an investor
Size positions and evaluate strategies using expected value and base rates, not recent P&L. A strategy with positive edge can have long losing streaks; a strategy with negative edge can have long winning streaks. Don't cut a good process after a bad run or double down on a bad one after a good run because of "reversion" on the next trade. Treat each allocation as a new draw.
As a decision-maker
When someone argues that "X is due" or "Y has had enough luck," check whether the trials are independent. If they are, reject the argument. Use explicit probabilities and base rates. In risk settings, do not reduce safeguards because "we've been incident-free" — the next incident's probability is not lowered by the past run.
Common misapplication: Confusing the Monte Carlo fallacy with regression to the mean. Regression to the mean is real: extreme outcomes (very high or very low) often follow with outcomes closer to the mean when there is measurement error or partial skill. That applies to distributions and repeated measures, not to the claim that "the next single outcome will reverse." The fallacy is insisting the next flip is due; regression is about the distribution of many subsequent outcomes.
Second misapplication: Applying the correction where trials are not independent. In card games without replacement, past draws do change the composition of the deck and thus future probabilities. In markets, if a stock's run-up reflects new information, the process may have changed. The rule is: independence first; then past outcomes do not affect the next one.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
Ed ThorpMathematician, author of Beat the Dealer; hedge fund founder
Thorp's work on blackjack and probability explicitly separated independent trials from the fallacy. He showed that in blackjack, cards are not independent — the deck composition changes — so counting cards and adjusting bets based on remaining composition is valid. In pure independent trials (dice, roulette), he emphasised that no system can beat the house. His discipline: know whether the process is independent; if it is, do not expect the next outcome to "correct" the past.
Munger has repeatedly warned against the gambler's fallacy in business and investing: "The same principles that ruin the gambler ruin the businessman." He stresses using base rates and expected value, not recent runs, and avoiding the impulse to "get even" or to assume reversion on the next decision. The mental model is to treat each decision on its merits under the true probability structure, not on the basis of how the last few turned out.
Section 6
Visual Explanation
Monte Carlo Fallacy — Independent trials: past outcomes do not change the probability of the next one. The fallacy is expecting the next outcome to 'correct' or 'balance' the run.
Section 7
Connected Models
The Monte Carlo fallacy sits in probability and judgment. The models below clarify when past outcomes do or do not affect the next one, and how to think in terms of base rates and distributions.
Reinforces
Gambler's Fallacy
Gambler's fallacy is the same error under another name: expecting the next independent trial to "correct" for prior outcomes. The Monte Carlo label ties the idea to the 1913 casino incident. Both say: independence means ignore run length for the next outcome.
Reinforces
Law of Large Numbers
The law of large numbers describes long-run convergence of the sample proportion to the true probability. It does not imply that the next trial "balances" the run. Distinguishing LLN (many trials) from the fallacy (next trial) is the key to correct application.
Tension
Regression to the Mean
Regression to the mean is real: extreme first measurements tend to be followed by values closer to the mean when there is noise or partial skill. That is about the distribution of subsequent outcomes, not a guarantee that the next single outcome reverses. Confusing "regression" with "the next one is due" is a common slip.
Leads-to
Probability Theory
Probability theory defines independence: P(A and B) = P(A)P(B), and P(A|B) = P(A) when A and B are independent. The fallacy is a violation of that definition. Correct application leads back to explicit probability: state the process, then the next-outcome probability.
Section 8
One Key Quote
"The probability of a future event is the same as it would be if the past had never existed."
— Pierre-Simon Laplace, A Philosophical Essay on Probabilities
For independent trials, the past does not change the future. Laplace's formulation is exact: the probability of the next outcome is what it would be with no history. The fallacy is acting as if the past does exist in a way that shifts the odds — "we're due," "it's overdue." The discipline is to ask: is this process independent? If yes, treat the next draw as if the past had never existed.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
The Monte Carlo fallacy is one of the highest-leverage corrections in decision-making. People constantly use recent runs to justify the next bet, the next hire, or the next risk. In independent or near-independent settings, that is wrong. The next outcome is a fresh draw. Train yourself to spot "due," "overdue," and "law of averages" applied to single or short-horizon decisions — and to reject them unless the process is clearly not independent.
In protecting and surviving, the fallacy is doubly dangerous. Complacency after a long run of no incidents ("we're safe") and panic after a run of bad outcomes ("we're cursed") both misattribute probability to the run. Maintain defences and margins based on process risk and base rates, not on how long it has been since the last failure or win.
In investing and strategy, run length is not information for the next trial. A strategy with positive edge can underperform for years; one with negative edge can outperform. Evaluate using expected value and base rates. Do not fire a good process after a bad run or scale a bad one after a good run because of a false sense of reversion on the next trade.
Independence is the gate. When trials are not independent (e.g. cards, information arrival), past outcomes do inform the next one. So the first question is always: is this process independent? If you cannot answer, default to not assuming that the past run changes the next outcome. The cost of wrongly assuming independence is usually smaller than the cost of the fallacy.
Use base rates explicitly. When someone says "X is due," ask: what is the base-rate probability of X on the next trial? If the answer is unchanged by the run, the argument is fallacious. Making the base rate explicit is the fastest way to disarm the fallacy in yourself and in others.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A roulette player has seen black 15 times in a row. He bets heavily on red because 'red is due.'
Scenario 2
A blackjack player increases her bet because she has seen few high cards in the last two decks and the remaining deck is rich in high cards.
Scenario 3
A CEO reassigns a regional manager because 'they've had three bad quarters; someone else deserves a chance.'
Scenario 4
After five years with no workplace accidents, a plant manager relaxes safety audits.
Section 11
Summary & Further Reading
Summary: The Monte Carlo fallacy is believing that past independent outcomes change the probability of the next one — e.g. "red is due" after a run of black. They don't. Each trial is a fresh draw; the law of large numbers applies to long-run distributions, not to the next single outcome. Use it to protect and decide: don't rotate bets or relax risk controls because of recent runs. Treat each decision under uncertainty as a new draw unless the process has changed. Rely on base rates and process evidence, not run length.
Kahneman discusses the gambler's fallacy and the narrative impulse to see patterns in random sequences. Accessible treatment of how we misread independence and run length.
Laplace's formulation that the probability of a future event can be the same as if the past had never existed — the mathematical basis for ignoring run length when trials are independent.
Thorp distinguishes independent trials (no system) from dependent trials (card counting). Clear application of when past outcomes do and do not inform the next one.
Leads-to
Neglect of Probability
Avoiding the fallacy means using base rates and process probabilities for the next outcome. Neglect of probability is the broader error: ignoring or underweighting explicit probabilities in favour of narrative or intuition. Both corrections point to the same discipline: anchor on the right probability.
Tension
Confirmation Bias
We notice and remember runs that fit a story ("I was due," "streaks continue"). Confirmation bias reinforces the fallacy by making us overweight run-based narratives. The fix is to insist on independence and base rates before drawing conclusions.
