Every gambler meets a system eventually. Double after a loss; follow a sequence; raise after a win and drop after a loss. They are presented as ways to turn a losing game into a winning one, and they have been circulating for centuries.

They do something real. It is just not the thing on the label.

The one-line proof

The argument against every betting system is short enough to fit in a paragraph, and it does not depend on which system you mean.

The expected value of a series of bets is the sum of the expected values of the individual bets. If each bet has a negative expectation, then any sum of them has a negative expectation. Changing the size of the bets scales the numbers; it cannot change the sign.

A system is a rule for choosing stake sizes. There is no rule for choosing stake sizes that turns a sum of negative numbers positive. That is not a claim about roulette or dice — it is arithmetic.

What Martingale actually does

Double after every loss, and the first win recovers everything plus one unit. It works, in the sense that the arithmetic of the recovery is correct.

What it does is redistribute your outcomes. You now win small amounts very often, and lose an enormous amount rarely. Multiply those together and you arrive at the same expected loss you started with — you have simply reshaped it into a form that feels like winning right up until it does not.

Starting at one unit, an eight-loss run needs a 256-unit bet to continue, and you are 255 units down at that point. Eight losses in a row is roughly a 1-in-200 event on a near-even bet. Play a few hundred rounds and you should expect to meet it.

Two things then stop you: the table maximum, and your bankroll. Both exist. Neither cares about your sequence.

The others are the same shape

Fibonacci escalates more slowly, so the ruinous bet takes longer to arrive and is smaller when it does — in exchange for recovering less. D’Alembert escalates gently and mostly just adds noise. Labouchère tracks a list of targets and gives the pleasant illusion of a plan being executed.

They differ in how the risk is distributed and in how it feels. None differs in expectation, because none of them can.

Where the RNG comes in

Systems usually smuggle in an assumption: that something about the sequence carries information. After five reds, black is somehow more likely, or the game is due.

Against a properly implemented random number generator this is not merely wrong, it is wrong in a stronger way than at a physical table. A real roulette wheel could at least in principle develop a bias — a worn pocket, an unlevel base — and people have historically found and exploited exactly that.

An RNG has no physical state to wear. Each result is drawn independently. The generator does not know what it produced last round and has no mechanism by which it could care. Five reds is not a pattern; it is five draws that happened to agree, and the sixth draw is unmoved by all of it.

This is why the physics-based attacks that once worked on real wheels are meaningless online. There is no wheel to measure.

The one thing systems are genuinely good for

A flat stake is a system, and a good one. Deciding in advance that every bet is the same size does not change your expectation either — but it makes your losses predictable, keeps sessions inside a budget, and removes the moment where you talk yourself into a bet four times larger than you meant to make.

That is the honest version of what a staking plan can do: control variance and protect a bankroll. Not generate an edge.

How to recognise the sales pitch

Any system claiming to beat a negative-expectation game will have one of these at its centre: a belief that outcomes are due, a stake-escalation scheme presented as risk-free, results over a run too short to mean anything, or a bankroll assumed to be unlimited.

If someone is selling one, the reliable way to make money from a winning gambling system is to use it. That they are selling it instead is the whole review.

The gambler’s fallacy in detail · What the house edge really is