The gambler’s fallacy: why your brain errs after 7 heads

On 18 August 1913, in the Monte-Carlo casino, the roulette ball stopped on black. Then on black again. Then on black once more. Twenty-six times in a row. As the streak stretched out, gamblers crowded around the table, betting more and more on red — utterly convinced that the opposite colour was ‘due’, that it ‘had to’ come up, that statistically it could not stay away any longer. Fortunes were lost that night. The probability of stringing together 26 blacks on a fair European roulette wheel is roughly one in 137 million; yet on every spin, red kept exactly the same chance it had at the very first turn.

That episode gave the phenomenon its nickname: it is sometimes called the Monte-Carlo fallacy, more commonly the gambler’s fallacy. It is one of the most universal and stubborn cognitive biases — and it doesn’t just hit casino regulars. It slips into our everyday decisions, the moment we picture a streak as something that ‘has to’ reverse. Understanding why the brain falls for it is a concrete step toward an honest reading of randomness.

What the gambler’s fallacy actually is

The gambler’s fallacy is the belief that an independent random event becomes more likely because it is ‘missing’ from a recent streak — or, conversely, that it becomes less likely because it has just happened several times. Toss seven heads in a row on Heads or Tails: your brain whispers that on the eighth flip, tails is ‘more likely’. Roll a die four times without ever landing a 6: you would happily bet it has to come up next.

That is wrong. On a fair coin, every flip is rigorously independent of the previous ones: the probability of heads stays at 1/2, and the probability of tails stays at 1/2, on every single toss, with no exception. On a six-sided die, every face keeps its 1/6 probability, no matter what came before.

The concept was first formalised scientifically in the early 1970s by psychologists Amos Tversky and Daniel Kahneman, in their landmark paper Belief in the Law of Small Numbers, published in 1971. Their thesis: our intuitions about randomness are deeply misleading. We expect a sample — even a tiny one — to already look like the statistical behaviour of a large one. We project onto ten flips what only becomes true across ten thousand. They called this tendency the representativeness heuristic: we judge a random sequence as ‘plausible’ only if it looks random — in other words, only if it alternates and visually mixes itself up. A run of seven heads in a row strikes us as suspicious, even though it is statistically mundane.

Tversky and Kahneman popularised an experiment that has become canonical. If you ask someone which of the two sequences HTTHTH or HHHHTT looks more like the result of a fair coin toss, the spontaneous answer is almost always the first: it ‘looks random’. Yet the two sequences have rigorously the same probability of appearing — one chance in sixty-four. Our brain files the first as ‘normal’ and the second as ‘strange’, when randomness has no preference whatsoever between them.

What the mechanics of randomness really say

A coin has no memory

That is the sentence to remember. When you flip the coin on Heads or Tails, it does not ‘know’ you have just landed seven heads. It carries no trace of the previous tosses. The calculation performed by your browser — detailed in our article on how our draws work — produces every result independently of the previous one. That is exactly the mathematical definition of an independent event: its probability does not depend at all on the history that came before.

In concrete terms: after seven heads, the probability of tails on the eighth flip is 50%. Not 60%, not 75%. Fifty. Just like the first toss. And if you get heads again on the eighth, the probability of tails on the ninth will still be 50%. What shocks the intuition is not the absurdity of the result — it is the absurdity of our intuition.

The massive confusion with the law of large numbers

Here is where some of the fog clears. Many people — including seasoned gamblers — confuse the gambler’s fallacy with the law of large numbers, a perfectly valid mathematical theorem. That law says that over a very large number of draws, the observed frequency of an event converges toward its theoretical probability. Across 10,000 coin flips you will get very close to 5,000 heads; across a million, you will be more precise still.

But — and this is the whole point — that law says nothing about what should happen in the short term. It does not claim that nature ‘rebalances’ local imbalances. It simply says that imbalances become negligible in proportion as the sample grows. If you land seven heads in a row, the sequence is not going to ‘correct itself’ with seven straight tails; it will dilute into the tens of thousands of tosses that follow, where opposite streaks will also appear — without plan, without intent, without compensation.

Confusing the two is the heart of the gambler’s fallacy. It is applying to ten flips what is only true across ten thousand.

Why the brain insists

That resistance is not, deep down, irrational: our brain is a high-performance pattern detector. Spotting a regularity — rain after a grey cloud, the repeated behaviour of a predator — was a major survival edge. Faced with a sequence of draws, the same circuit fires up and looks frantically for a pattern. When it finds none, it manufactures one: the ‘law of compensation’, the idea that a streak ‘has to’ reverse.

Kahneman, who would later describe this mechanism in Thinking, Fast and Slow (the title of his landmark book on fast and slow thinking), talks about an intuitive, immediate response, produced before any statistical reasoning could even be summoned. It is as fast as a reflex — and just as hard to disable. Watching seven heads in a row triggers in you an almost physical sensation of imbalance. It is that sensation, not the maths, that guides the hand toward the opposite bet.

The most surprising thing, perhaps, is that knowing the theory does not protect you. Tversky and Kahneman showed in their 1971 paper that the bias also hits researchers trained in statistics — who, in concrete situations, reason as if small samples already had to faithfully reflect the entire population. The reflex precedes the calculation, even among those who master the calculation.

When the bias becomes a trap

Far beyond the casino

The gambler’s fallacy doesn’t just affect roulette enthusiasts. It shows up everywhere we picture an independent event as ‘due’: a parent expecting a fourth child who feels it ‘has to’ be a boy after three girls; an investor convinced that a stock ‘must rebound’ after several losing sessions; a marker who persuades themselves that after four good papers, the fifth ‘will inevitably’ be weaker; a driver changing route because the rain ‘has to stop’ since it has lasted too long. None of these intuitions has any statistical basis. All share the same mental machinery. In the lottery, this bias drives millions of players to believe that certain numbers are ‘due’ because they have not been drawn for several weeks — our article on the real odds of winning the lottery dismantles that mechanism with the numbers in hand.

A cognitive engine of gambling addiction

Where the stakes turn serious is in pathological gambling. Quebec psychologist Robert Ladouceur, professor at Université Laval and founder of the Quebec Centre of Excellence for Gambling Prevention and Treatment, devoted a major part of his work to showing that erroneous thoughts about randomness — first among them the gambler’s fallacy — sit at the core of what keeps problem gambling going. The struggling gambler does not just play because they enjoy playing: they keep playing because they sincerely believe their losing streak is ‘abnormal’ and that a win is now statistically imminent.

Ladouceur’s team showed as early as 2001 that a cognitive therapy targeted at correcting these beliefs allowed 86% of participants to no longer meet, by the end of treatment, the criteria for pathological gambling — a result that became a reference in the field. The clinical work consists of observing, in real time, the thoughts the gambler voices out loud while playing: ‘the 7 is due’, ‘it’s been ten spins without a big win, it has to drop’, ‘I can feel it’s the moment’. Once identified and named, those thoughts become attackable — they can be confronted with the actual mechanics of the draw and replaced with accurate formulations. Understanding the bias intellectually is not enough to neutralise it emotionally, but it is the first step: you cannot disable a mechanism you have not identified. If you recognise these ruminations in your own relationship with gambling, or in someone close to you, our article on the signs of problem gambling offers concrete markers.

Outsmarting it in yourself

A few simple reflexes help defeat the bias when it shows up. First, name what is happening: ‘I am thinking it’s due — that’s my brain inventing a rule’. That simple labelling slows down intuitive thinking and gives reasoning a chance. Then, bring the key sentence back: this coin, this die, this wheel has no memory. What has happened has no influence on what is going to happen. Finally, run the test: roll a die twenty times on Virtual Dice and write down the results. You will see streaks appear — three 4s in a row, four turns without a 6 — that look abnormal but are perfectly normal. Across 10,000 tosses, randomness even produces runs of 13 consecutive heads — that is the log₂(N) formula, and our article on the real probabilities of Heads or Tails illustrates it with the standard deviation and the figures from the Kerrich experiment (5,067 heads out of 10,000, a perfectly mundane result).