Gambler's Fallacy: Does the Balloon Remember?

The gambler's fallacy is the belief that a random event becomes more or less likely because of what just happened — that tails is due after a run of heads, or that a balloon that has popped early three times running is bound to go the distance next. Most explanations stop at a coin. This one uses a real game with a published risk per step, and a simulator you can run to see whether the game has any memory at all.

Monte Carlo, 1913: twenty-six blacks

On 18 August 1913, the roulette wheel at the Monte Carlo Casino landed on black again and again. As the streak grew, gamblers piled onto red, sure that the wheel had to even itself out. It landed on black twenty-six times in a row, and the players betting against it lost millions of francs.

A run of twenty-six blacks on a single-zero wheel is genuinely rare — about one in 137 million. But the rarity of the whole streak says nothing about the next spin. After twenty-five blacks, the twenty-sixth spin was black with the same probability as the first: 18 in 37. That is why the gambler's fallacy is also called the Monte Carlo fallacy.

Why the balloon is a better example than a coin

A coin makes the point, but each flip is over in an instant and nothing builds from one flip to the next. A ladder game is where the fallacy gets tempting, because every step asks you to predict the next one while the amount at stake keeps growing.

In Balloon Pump, each pump survives with a flat 90% chance, up to twenty pumps. The pump the balloon bursts on is fixed before you start: the game draws one random number, turns it into a burst point, and commits to it with a published hash. So the game itself gives us two precise questions:

• Inside a round: once the balloon has survived ten pumps, is the eleventh riskier, safer, or the same?
• Between rounds: after three balloons in a row pop on the very first pump, is the next one any safer?

Inside a round: the balloon is not "due"

The burst point follows what statisticians call a geometric distribution, and its defining property is that it is memoryless. A balloon that has survived fourteen pumps faces its fifteenth with exactly the odds a fresh balloon faces its first: 90%.

What changes as you climb is not the risk of the next pump but the compounded odds of having got this far. More than half of all rounds burst by pump seven, and only 12.2% reach twenty. Those two facts are easy to confuse, and confusing them is the fallacy: "most balloons don't get this big, so this one must be about to pop." The balloon that is already this big has already beaten those odds. They are behind it, not in front of it.

Between rounds: bad luck does not bank

The second version is the one that drains budgets. A balloon pops on the first pump — a 10% event. Then another. Two in a row is a 1% event; three is 0.1%. At that point it feels as though the game owes you a good round.

It does not. Each round uses a fresh random number, drawn independently of the last. The next round survives its first pump 90% of the time and reaches ten pumps 34.9% of the time, exactly as every other round does. Three early bursts were improbable; they are also finished. Nothing carries over.

The simulator draws each round the way the game does. Look at the right-hand column: whatever happened before — a fresh balloon or one that has survived nineteen pumps, a previous full run or three first-pump bursts in a row — the next pump survives about 90% of the time. The rows with fewer rounds wobble a little more on each run, which is what small samples do. Press the button a few times and watch them wobble in both directions.

The mirror image: the hot hand fallacy

The gambler's fallacy expects a streak to reverse. The hot hand fallacy expects it to continue: "I've had three full runs, I'm on a roll, pump further." In a game driven by a random number, the two are the same mistake pointed in opposite directions. The balloon is neither cold nor hot.

The name comes from basketball. In 1985, Gilovich, Vallone and Tversky analysed shooting records and concluded that the hot hand players and fans believed in was an illusion. In 2018, Miller and Sanjurjo showed that the original analysis contained a subtle statistical bias, and that once it was corrected the data did show a modest hot hand in shooters. That is a fair result for people, whose confidence, fatigue and focus really do change from minute to minute. It does not transfer to a random number generator, which has no confidence to gain.

Why smart people fall for it

In 1971, Amos Tversky and Daniel Kahneman described the root cause as a belief in the law of small numbers: we expect a short sequence to look like the long-run average. Over thousands of rounds, 10% of balloons really do pop on the first pump. Over five rounds, three first-pump bursts are unusual — and the mind tries to "correct" the sample by predicting a run of good rounds to restore the balance.

But the long-run average is not restored by compensation. It is restored by dilution: the three early bursts are still there, simply outweighed by the thousands of ordinary rounds that come after them.

Where it actually costs money

Within a round, the fallacy is mostly harmless on a fairly built ladder. If you cash out because the balloon "feels due", you have not lost anything in expectation, because every stopping point on Balloon Pump returns the same 95%. We explain why in when to cash out.

Between rounds is different. The expensive version is raising the stake because a win is due. A bigger bet does not buy better odds; it buys the same odds at a bigger size. The house's share scales with what you stake, so a player who doubles up after a bad run pays more for exactly the same chance. That is the Monte Carlo mistake — not believing in the streak, but betting on its reversal.

How to tell whether a game has a memory

Everything above is true of Balloon Pump because of how it is built. A system that is designed to adjust its odds according to your history is a different thing, and from the outside the two can look identical. So don't take a game's word for it. Look for:

1. A published chance per step. Balloon Pump states 90% per pump and publishes the full twenty-rung ladder beside it.
2. An outcome committed before you act. The burst point is locked behind a hash before your first pump, so it cannot react to what you do.
3. A way to check it afterwards. Each round can be verified from its revealed seed. Our provably fair verification guide walks through it step by step.

If a game gives you all three, the round in front of you owes nothing to the rounds behind it. If it gives you none, you cannot know — and that is worth knowing too.

The short version

• A streak is information about the past, not a force acting on the future.
• On Balloon Pump, the next pump survives 90% of the time, whatever came before — inside the round or across rounds.
• "Due to pop" and "on a roll" are the same error in opposite directions.
• The costly form is raising your stake to chase a reversal. The odds do not move; only the amount at risk does.