The gambler's fallacy: nothing is ever "due"
After ten blacks in a row, red feels inevitable. It isn't. Here is why independent events don't balance out on schedule, and where this bias quietly shows up outside casinos.
WrongBlack has hit ten times in a row, so red is due.
RightEach spin is independent; red is exactly as likely as it was before the streak.
What people get wrong#
The gambler's fallacy is the belief that a random process "owes" you an outcome after a streak. Ten heads in a row, so tails must be coming. Five losing spins, so a win is overdue. The Medical Journal of Australia describes it as "the erroneous belief that the outcomes of recent random events have some bearing on future random events."
The fallacy's other name comes from a famous night in 1913 at a Monte Carlo casino, as the same article retells it. Black came up on the roulette wheel spin after spin. Convinced red was overdue, gamblers bet harder against the streak with every black. The run reached 26 consecutive blacks before it ended, and the losses along the way gave the error its nickname: the Monte Carlo fallacy.
Why it's wrong#
For a fair coin or a fair wheel, each trial is independent. The coin does not know what it did last time. As the MJA piece puts it, "a coin has no memory." The chance of tails on the eleventh flip after ten heads is still one in two.
What confuses people is mixing up two different questions. The odds of flipping ten heads in a row before you start are small, one in 1,024. But, as the article points out, that is also the probability of any other specific sequence of ten flips. Once the ten heads have happened, they are history. The only question left is the next flip, and it is a coin flip.
The "law of averages" people invoke is a garbled version of something real. Over a very large number of trials, the proportion of heads tends toward half. But that happens because the early streak gets diluted by thousands of later flips, not because the coin compensates by producing extra tails.
Why it happens#
We expect short random sequences to look "random," meaning well mixed. A streak looks like a pattern, and patterns feel like they must correct. So we treat a run of one outcome as a debt the universe has to pay back.
This is not only a casino problem. Economists Daniel Chen, Tobias Moskowitz and Kelly Shue studied three high-stakes settings: refugee asylum courts, loan application reviews and Major League Baseball umpires' pitch calls. They found "consistent evidence of negative autocorrelation in decision-making that is unrelated to the merits of the cases," meaning decision-makers leaned against repeating their previous call (NBER Working Paper 22026).
The NBER Digest summary gives a concrete figure: the odds that a judge rejects an asylum seeker were 3.3 percentage points higher if the judge had approved the previous case, all else being equal. The effect was stronger after longer streaks. In other words, a case's outcome partly depended on the unrelated case that came before it.
The fix#
My lottery numbers haven't come up in two years, so they're overdue.
My lottery numbers have the same odds this week as every other week.
Past independent draws don't change the next draw's odds.
Ask one question before acting on a streak: is there a mechanism that links one outcome to the next?
- If no (coin flips, roulette spins, lottery draws, a fair die), the past tells you nothing. Bet, or don't, as if the streak never happened.
- If yes, the outcomes may genuinely be linked. Cards dealt without reshuffling change what's left in the deck. A machine that keeps failing may be wearing out. Those are real dependencies, not fallacies.
- If you decide things in sequence (grading, hiring, reviewing), judge each case against a fixed standard, not against the one you just saw.
Sources
Every factual claim on this page links to one of these. Checked November 16, 2026. See how we source.