What a betting system promises
A betting system is a rule for choosing how much to bet next, based on what has already happened. The most famous is the Martingale: bet 1 unit, and after each loss double the bet, until a win recovers all losses plus one unit. Others raise bets after wins (Paroli), step them up and down by one unit (D’Alembert), or follow a number sequence (Fibonacci). All of them aim to turn a game with a built-in margin into a steady source of small wins.
They cannot, and the reason is not subtle. This article explains the math so that the claim can be checked rather than taken on trust.
The core argument
Take a game with a negative expected value on each bet, such as an even-money bet on European roulette, where the expected value is -1/37, about -2.70%, of the stake. Each spin is independent of the previous ones, so the odds of the next bet do not depend on earlier results.
Now consider any betting system. Over a session, the total result is the sum of the results of each individual bet, and each bet’s result is its stake multiplied by a per-unit outcome. The expected value of a sum is the sum of the expected values. So the expected total is:
expected total = -2.70% x (expected total amount wagered)
Whatever the system does, the average loss is 2.70% of the average money bet. The system can change how much is wagered and when, but every unit wagered still carries the same negative expectation. There is no sequence of negative-expectation bets whose total is positive on average. This holds for stopping rules too: deciding to quit after a win, or after a certain number of bets, changes the pattern of results, not the average.
A worked Martingale example
Suppose a player bets on red on a European wheel, starting at 1 unit, doubling after each loss, and giving up after 7 consecutive losses (a table limit or bankroll limit forces the stop somewhere). The probability of losing a single bet on red is 19/37, about 0.5135.
- Stakes in a losing run: 1, 2, 4, 8, 16, 32, 64. Total at risk across all seven: 127 units.
- Probability of losing all seven: 0.5135 to the 7th power, about 0.0094, or just under 1%.
- Probability of winning at some point within the seven bets: about 1 - 0.0094 = 0.9906.
If the player wins at any step, the doubling means the loss so far is covered and the net result is +1 unit. (For example, losing 1, 2 and 4 then winning 8 gives -7 + 8 = +1.) If all seven lose, the result is -127 units.
Expected value of one round:
(0.9906 x 1) + (0.0094 x -127) = 0.9906 - 1.1956 = about -0.205 units.
The system looks like it wins nearly 99 rounds in 100, and it does, but the roughly 1% of rounds that lose are large enough to pull the average below zero. The result matches the earlier argument: the average amount wagered per round is about 7.6 units, and 2.70% of that is about 0.205. The math agrees from both directions.
Why it feels like it works
Progressions produce a distinctive pattern: many small wins, then an occasional large loss. In short samples, the small wins dominate, and a player may see a long run of profits. This is a matter of short-run results rather than evidence of an edge. The shape of the distribution is skewed, so the average is pulled down by rare events the sample might not contain.
It is also an average outcome trap. Reporting “wins 99% of rounds” is accurate and still misleading, because it ignores size. A system with a 99% win rate and a rare large loss can have the same negative average as one that wins 48% of the time.
The practical limits
Real tables add two more problems for doubling progressions:
- Table limits: maximum bets cap how many times a stake can double. Once the cap is reached, the player cannot recover.
- Finite bankroll: doubling grows exponentially. After 10 straight losses from a 1 unit start, the next stake is 1,024 units and the total lost so far is 1,023. The chance of such a run is about 0.5135 to the 10th power, a little over 0.1%, which is small but not remotely impossible across many sessions. This is the idea described in the article on risk of ruin.
The progression does not remove risk. It reshapes it: frequent small gains are bought with a small chance of a very large loss.
Positive progressions and pattern betting
Raising bets after wins (so-called positive progressions) reverses the shape: many small losses and an occasional large win. The expected value argument is unchanged. The same applies to betting on patterns, such as “bet black after four reds”. The wheel has no memory, so there is no pattern to find, a point made in the article on the gamblers fallacy. A bet’s expectation depends on its odds and payout alone.
Some people claim that tracking a game’s results allows a better bet to be spotted. That only helps if the odds differ between bets, which on a fair wheel they do not, and any such claim needs far more data than most people collect.
What changes expectation
Only two things change the expected value per unit: the odds of the bet and the payout. Choosing a different bet with a lower margin (the craps odds bet is the usual example, with zero margin), or a different game with a lower edge, changes the average cost. Choosing different bet sizes does not. In games like blackjack, the player’s decisions can change the odds, which is a different kind of choice.
Gambling outcomes are uncertain and no system guarantees a profit. Anyone being sold a system with a promise of steady winnings should be sceptical, and anyone who is losing more than they planned can find support in the article on responsible play.