Ancient Gambling Myths and What Probability Does to Them

From astragali to the «hot» slot machine: the old superstitions rest on the same few mistakes about randomness, and each one has a cost that can be written down.

Published: December 31, 2025 • By Harriet Lowell

The oldest randomising devices were not designed as randomisers at all. Astragali — the small ankle bones of sheep and goats — land on four distinguishable faces, and people were throwing them long before anyone wrote down what a throw meant. Cubical dice came later and became common enough in Rome to leave traces in its literature and its laws; card games appear in China well before Europe. What runs through all of it is not the equipment but the commentary: a throw has almost always been read as a message about luck or fate rather than as one sample from a distribution.

Most of the superstitions built on that reading are still in circulation, in modern clothes. They are worth taking apart not because believing them is embarrassing, but because each has a cost that can be written down.

The gambler's fallacy

A European roulette wheel has 37 pockets: eighteen red, eighteen black and a single zero. Six blacks land in a row, the table goes quiet, and someone leans in to back red, because red is «due».

That intuition is not wrong, it is applied at the wrong moment. Before any of the six spins happened, the chance of six blacks in a row was about 1.3 %, roughly one run in seventy-five. But once five have landed, the only question left is the sixth spin, and the sixth spin is a wheel with 37 pockets, eighteen of them black: 18/37, or 48.6 %, exactly what it was before the run started. The wheel has no memory and no way of balancing a sequence it cannot perceive. The cost is not the bet itself — red and black carry the same 2.70 % edge whenever they are placed — but that the fallacy manufactures reasons to stake more than intended, and the conviction grows with the length of the run.

Hot machines, cold machines and jackpots that are «due»

The same idea reappears at slots as the belief that a machine which has not paid for hours is ready, or that one which has just paid is emptied out. Every round is generated independently and the software keeps no ledger of what it owes: a game that paid its top prize a minute ago is as likely to pay in the next round as one that has paid nothing all day.

One narrow exception feeds the myth. Some progressive jackpots are advertised as dropping before a stated amount or time; where that appears in the game's own rules it is a rule and can be relied on. Repeated in a forum about an ordinary slot with no such published mechanism, it is the myth again.

Lucky numbers and rituals

On a single-zero wheel every straight-up number pays 35 to 1 and owns one pocket out of 37. Stake a unit on a birthday: 1/37 of the time it returns 35, and 36/37 of the time nothing, for an expected result of −1/37, or −2.70 %. Stake a unit on a number chosen by a random generator and the arithmetic is identical. A lucky number is not a worse bet than any other — it is exactly the same bet, which is the point. The ritual adds nothing to the outcome and a good deal to the confidence, and confidence is what changes stake sizes. Blowing on dice, tapping a card twice, waiting for a particular dealer: all free in themselves, and expensive only when treated as evidence, because evidence justifies raising the stake.

The illusion of control

Ask people to throw dice for a high number and many throw harder; ask for a low number and the throw softens. Slot players hold a stop button to «land» a symbol, although the outcome was fixed when the round was requested and the button only shortens the animation.

This is the most durable of the family, because the physical act is real — you did throw the dice — and the mind draws a line from act to result. It transfers neatly to live tables, where seat choice, bet timing or the identity of a dealer feels like an input. The rules set the edge there and nothing done with a seat changes it, which is worth keeping in view when reading about the tactics players bring to live dealer games.

Belief in the streak

Random sequences contain far more structure than people expect. Flip a fair coin a hundred times and a run of six or seven identical results will usually appear somewhere — not because the coin is behaving oddly, but because a hundred flips contain many chances for a run to start. Shown that sequence, most observers pick out the run and treat it as the interesting part. Gambling supplies a stream of such sequences, and a player needs only to have sat through one good run to have a story with a beginning, a middle and an end. The story is real; what is false is the claim that the run was a state the player was in and could re-enter.

Martingale meets the table limit

The most persistent myth is not about luck but about structure: that a doubling progression turns a losing game into a winning one. Bet one unit on black; if it loses, bet two, then four, eight, and so on. Any win recovers everything staked and leaves one unit of profit. The logic is sound as far as it goes, and it goes exactly as far as the table maximum.

After seven consecutive losses the player is 127 units down and the system instructs them to stake 128 to win back a single unit. A table with a minimum of one unit and a maximum of one hundred refuses that bet: the progression is stopped not by bad luck but by a published rule, and the loss is locked in at 127 units in pursuit of a profit of one.

How often does that happen? On a single-zero wheel an even-money bet loses 19 times in 37, or 51.35 % of the time, so seven straight losses come up about 0.94 % of the time — roughly once in 106 cycles. Put the halves together: 99.06 % of cycles win one unit, 0.94 % lose 127, and the average result is 0.9906 − 1.1959 = −0.205 units per cycle.

That figure repays a second look, because it is not a coincidence. Across a full cycle the expected amount wagered is about 7.59 units, and 2.70 % of 7.59 is 0.205. The progression has not touched the house edge; it has rearranged when the same 2.70 % is charged, trading many small wins for one rare, very large loss. No staking pattern does anything else, and a bigger bankroll does not escape it: surviving ten straight losses requires 1 023 units and a table willing to accept a stake of 512.

Survivorship bias in the stories

The last myth is built out of what gets repeated. The player who won on a birthday number tells everyone; the thousands who backed it that week say nothing. A screenshot of a 400× hit travels; quiet sessions are not photographed. Forums, feeds and marketing filter outcomes by the outcome itself, so the visible sample is guaranteed to be unrepresentative — and it is the sample most players judge a game by. The same filter applies to systems: a progression is described by those whose cycles closed, not by the one in 106 who met the table limit. That scepticism extends to anything an operator says about itself, the newer claims included — the sustainability and energy arguments now appearing in casino marketing deserve the same treatment, as the look at eco-conscious gambling claims sets out.

What survives

Probability removes a lot from this subject and leaves a short list intact. Game selection is real: European roulette costs 2.70 % of every unit staked and the American double-zero wheel 5.26 %, so choosing the wheel is worth more than any betting pattern ever devised. Blackjack by basic strategy runs at roughly half a per cent to one per cent, but a table paying 6:5 on a blackjack instead of 3:2 adds about 1.4 percentage points and quietly multiplies that cost. Limits on bankroll and session length are real, because they are decided in advance rather than during a run, and so is reading terms before accepting them. None of it promises a winning session; it sets the price, which is all that can be set — the practical version is in the guide to playing with a plan.

The old superstitions have lasted because they answer a question probability refuses to answer: why this throw, why me, why now. Probability's answer is that there is no why, only a price per unit staked. It is the duller answer, and the only one that has ever been of use at a table.