Risk of Ruin & Bankroll Maths
Risk of ruin is the probability that a run of bad variance empties your bankroll before you choose to stop, and in any negative-edge game, given enough play, that probability is one. This lesson works through the gambler's-ruin maths to show why neither a bigger bankroll, more time, a staking system nor a stop-loss can convert a house edge into a win; they only change how quickly the loss arrives.
- Define risk of ruin and use the gambler's-ruin result to explain why, against an effectively unlimited operator bankroll, eventual ruin in a negative-edge game has probability one.
- Separate variance from expectation, and explain why a lucky session is dispersion around a negative mean rather than evidence of opportunity or skill.
- Show why a larger bankroll or more time played raises total expected loss and only delays ruin, using E[loss] = edge x total stake.
- Explain why no staking system or stop-loss changes per-bet expected value, and why Kelly sizing applies only to positive-edge bets.
Variance is dispersion, not opportunity
The hope that fuels most long-run losses is that a good run proves the game is beatable. It does not. Variance is the spread of outcomes around the mean, and in every house-edge game that mean is negative. The upswing you remember and the downswing you would rather forget are the same mechanism sampled twice; average enough of them and they collapse onto the expected loss. Volatility changes how far and how often a session strays from that mean, and nothing more. It never adds expectation, because standard deviation and expected value are independent quantities: you can raise or lower the bumpiness of a 96% game without touching its 4% edge. Treating a lucky tail as evidence of skill or opportunity is the first and most expensive error.
Gambler's ruin: the classical result
Risk of ruin has an exact form. Model even-money play as a random walk: each bet moves your bankroll up one unit with probability p and down one with probability q = 1 - p. Starting with i units and quitting only at a target of N, the probability of reaching N before hitting zero is [1 - (q/p)^i] / [1 - (q/p)^N] when p is not equal to q. Two conclusions follow. In a fair game (p = q) the ruin probability is 1 - i/N, which climbs toward 1 as your target grows. In a house-edge game (p < q, so q/p > 1) the term (q/p)^N explodes, and against an effectively unlimited operator bankroll the probability of eventual ruin converges to 1. The house does not need luck; it needs you to keep playing.
Why a bigger bankroll only buys time
If ruin is ultimately certain, what does a larger bankroll change? Only its timing. Expected loss is linear and pitiless: it equals the house edge multiplied by everything you stake, so E[loss] = edge x bets x stake. A deeper bankroll lets you afford more bets before busting, which means more total turnover through the edge and therefore a larger, not smaller, expected loss. It buys survival time, not a positive expectation; the drift still points down, only more slowly relative to your stack. This is exactly what the Bankroll Survival tool shows: hold the edge and bet fixed, extend the play, and the share of players still solvent falls steadily toward zero. Time in the game is the operator's asset, not yours.
Stop-losses and staking systems: rearranging the same sum
Martingale, D'Alembert, Fibonacci, stop-losses and win goals all promise to tame the variance, and all leave the edge untouched. Expected value is additive: the sum of negative-EV bets is negative regardless of the order you place them or the size you choose. A progression like the Martingale trades a high chance of a small win for a small chance of a catastrophic one, a reshaped distribution with an unchanged mean. A stop-loss is subtler but no exception: it can cap a session's damage, yet it does so only by ending play, which reduces the total you wager. That lowers expected loss for the same reason simply betting less does, less turnover, never by creating an edge. No sequencing rule changes any single bet's expectation, and expectation is the only thing that compounds.
Kelly sizing is a positive-edge tool
Kelly staking is often invoked as disciplined bankroll management, but its scope is the exact opposite of a casino game. Kelly sizes each bet as a fraction of bankroll proportional to your edge; for an even-money bet that fraction is f = p - q, the edge itself. The formula is defined only when that edge is positive. Feed it the negative edge of a house game and it returns a negative or zero fraction, which is the mathematics telling you the growth-optimal stake is nothing at all. There is no fractional-Kelly, unit-sizing or 'risk management' refinement that rescues a negative-EV bet, because every one of them presupposes an edge to size. Kelly belongs to advantage play, such as card counting or a genuinely +EV promotion, not to games engineered to hold an edge over you.
Reading bankroll maths honestly
So what is bankroll maths actually for? It is a tolerance and survival instrument, not a profit one. Choosing a stake and a volatility your bankroll can absorb changes how long the money lasts and how violent the swings feel; matching a high-variance slot to too small a bankroll simply guarantees a fast bust. None of it alters the destination, which the law of large numbers fixes at a loss equal to the edge times everything wagered. Risk of ruin in a house game is not a dial you tune toward zero by playing well; given enough rounds it is one. Variance is the noise, the edge is the signal, and the signal points down. This is entertainment carrying a real, built-in cost, never a route to income.
Key terms
Check yourself
Fact-checks (5)
- Gambler's ruin: a player making even-money bets against an opponent with effectively unlimited funds is ruined with probability approaching 1, and for win/loss probabilities p and q = 1 - p (p != q), starting with i units and quitting at N, the probability of reaching N before 0 is [1 - (q/p)^i] / [1 - (q/p)^N]; in a house game q > p, so (q/p)^N grows and ruin probability rises toward 1 as N increases. Source: W. Feller, An Introduction to Probability Theory and Its Applications, Vol. 1 (gambler's ruin / random walk); Wizard of Odds - Risk of ruin.
- House edges used as examples are current: European roulette 2.70%, American roulette 5.26%, blackjack about 0.5% under optimal strategy, and baccarat banker 1.06%. Source: Wizard of Odds - House edge of casino games (accessed August 2026).
- Betting systems (Martingale, D'Alembert, Fibonacci and similar) change the timing and shape of wins and losses but not the per-bet expected value; the sum of negative-EV bets remains negative. Source: Wizard of Odds - Betting systems.
- The Kelly Criterion (J. L. Kelly Jr., 'A New Interpretation of Information Rate', Bell System Technical Journal 35, 1956) maximises long-run bankroll growth by staking in proportion to a positive edge; with no edge the growth-optimal fraction is zero, and overestimating the edge causes overbetting, which is why practitioners use fractional Kelly. Source: Kelly 1956; Wizard of Odds - Kelly criterion.
- UK measures that cut turnover reduce expected loss but do not change per-bet EV: an online-slot spin-speed floor of 2.5 seconds and an autoplay ban took effect 31 October 2021, and per-spin stake caps of GBP 5 (age 25+) from 9 April 2025 and GBP 2 (age 18-24) from 21 May 2025 applied thereafter. Source: UK Gambling Commission Remote Gambling and Software Technical Standards (RTS) and associated guidance.