The Gambler's Fallacy: Why Due Numbers Don't Exist

The gambler's fallacy is the belief that a random outcome becomes more likely because it has not happened lately. It feels like common sense and it is mathematically wrong. On a fair European roulette wheel, red after ten blacks is exactly as likely as red after ten reds: 18 in 37, every single spin.

The night black came up 26 times

The most famous example happened in 1913 at the casino in Monte Carlo, where a roulette wheel landed on black 26 times in a row. As the story is usually told, players piled onto red as the streak grew, convinced it had to end. Each new black made them more certain, and they lost heavily.

How unlikely was it? On a European wheel, P(black) = 18/37 ≈ 48.65%. The chance that a particular run of 26 spins is all black is (18/37)^26, about 1 in 137 million. The chance that 26 spins are all the same color, red or black, is twice that, about 1 in 68 million.

That is rare for one sequence, but roulette wheels around the world have spun an enormous number of times. Rare runs are bound to turn up somewhere. And on spin 27, red was still 48.65%.

Why the wheel has no memory

Two events are independent when one tells you nothing about the other. For independent events, the probability of both happening is the product of their probabilities, and the probability of the next one is unchanged by the last.

A coin makes the logic easy to see. Six heads in a row has probability (1/2)^6 = 1/64. But the sequence HHHHHT also has probability 1/64. Once five heads have already happened, the only question left is the sixth flip, and the two endings are equally likely.

The fallacy comes from mixing up two different questions. "How likely is a streak of six?" is about the future before it starts. "How likely is heads now?" is about one flip. The streak's rarity was already used up by the flips that happened.

The middle column grows dramatically. The right column never moves. That is the whole fallacy in one table.

Streaks are more common than your gut thinks

Part of the problem is that people expect randomness to look tidy, with lots of switching and no long runs. Real randomness is clumpy.

Working out the exact probabilities for 100 fair coin flips:

• A run of five or more identical results appears about 97% of the time.
• A run of six or more appears about 81% of the time.
• A run of seven or more appears about 54% of the time.
• The longest run averages about seven.

Try writing 100 random-looking flips by hand, then compare them with the real thing in the free coin flip lab. If your hand-made list never runs longer than four, it is less streaky than genuine chance.

Sleeping numbers are normal too

The roulette version of the fallacy often targets a number that has not hit for a while, sometimes called a sleeper. Sleepers are guaranteed by the math.

In 37 spins of a European wheel, you might expect each number to appear about once. In fact the chance that a given number does not appear at all is (36/37)^37 ≈ 36%. On average, about 13 of the 37 numbers are missing after 37 spins, and only about 24 different numbers show up.

Even after 100 spins, any particular number has a (36/37)^100 ≈ 6.5% chance of not having appeared, and on average about 2.4 numbers are still missing. None of them is any more likely on spin 101.

Psychologists describe the underlying habit as expecting a small sample to look like the whole population. A fair wheel only looks even across a very large number of spins, and even then it gets there by dilution, as the law of large numbers explains, never by catching up.

The mirror-image mistake: the hot hand

The gambler's fallacy says a streak should break. Its twin, the hot-hand fallacy, says a streak should continue: red has hit five times, so ride red.

For independent games, both are the same error in opposite directions. Many roulette tables display the last dozen or so results, and players read patterns into them either way. The display is accurate history and useless prediction.

When the past really does matter

The fallacy applies to independent outcomes. Some outcomes are not independent, and it is worth knowing the difference.

• Blackjack. Cards dealt from a shoe are not replaced until the shuffle, so what has been played changes what remains. A shoe that has shed many small cards really is richer in tens and aces. That is the basis of card counting.
• Within one lottery draw. Balls are not replaced during a draw, so each ball changes the odds for the next. Across draws, the machine resets and every draw is independent again.
• A physically faulty wheel. If a wheel were biased, a long record would reveal it, but the bias would favor the numbers that keep hitting, the exact opposite of what the fallacy predicts.
• Slots and digital games. Each result comes from a random number generator with no memory of past spins. A machine is never due. See how RNGs work.

What the fallacy costs you

The fallacy is not just a quirk of thinking. It drives betting decisions, and those decisions have prices.

A $10 bet on red after five blacks has an EV of (18/37 × $10) − (19/37 × $10) ≈ −$0.27, exactly the same as a $10 bet on red at any other moment. But if the belief that red is due persuades you to bet $50 instead of $10, your expected loss on that spin rises fivefold to about $1.35.

Systems built on the fallacy, like doubling after every loss, do this repeatedly. The Martingale math shows how that ends.

The bottom line

Independent games do not keep score. A wheel, a pair of dice and a random number generator start fresh on every result, so no number, color or machine is ever due.

Streaks are a normal feature of randomness, not a signal. The past only matters when outcomes are physically linked, as in a blackjack shoe. Everywhere else, the next result has the same probability it always had, and the house edge on it is the same too.