What the term means
Risk of ruin is the probability that a fixed pool of funds falls to zero (or below some level at which play must stop) before a stated goal is reached, or before a stated amount of time has passed. It is a probability, so it sits between 0 and 1, or between 0% and 100%. The word “ruin” is technical and simply means running out of the pool.
The idea was studied for centuries under the name gambler’s ruin, a classic problem in probability. It is used in insurance, finance and game analysis, anywhere repeated uncertain gains and losses meet a limited reserve.
Every figure below is an illustration with simple, made-up odds. Real games differ, and the numbers are meant to show how the math behaves.
The simplest model
The standard model makes these assumptions:
- Each play is a bet of 1 unit at even money: win 1 unit with probability p, lose 1 unit with probability q = 1 − p.
- Plays are independent, so results do not influence one another.
- Play continues until the balance hits 0 (ruin) or reaches a target N.
- The starting pool is a units.
Under these assumptions the balance follows a path of steps up and down, which the math calls random walks, and the question is how likely the path is to touch zero first.
Case 1: a fair game
If p = q = 0.5, the chance of ruin before reaching the target is 1 − a/N.
- Start with 10 units, target 20: ruin probability is 1 − 10/20 = 50%.
- Start with 10 units, target 100: ruin probability is 1 − 10/100 = 90%.
Even in a perfectly fair game, aiming for a target far above the starting pool makes ruin the likely outcome, because the pool is small relative to the swings needed to reach the goal. The chance of success is simply the starting pool divided by the target.
Case 2: an unfavourable game
Let r = q/p. When p is not 0.5, the chance of ruin before reaching the target is (r^a − r^N) / (1 − r^N).
Take p = 0.45, so q = 0.55 and r is about 1.222, with a = 10 and N = 20:
- r^10 is about 7.44 and r^20 is about 55.3.
- Ruin probability is (7.44 − 55.3) / (1 − 55.3), about 0.88, or 88%.
Compare that with the 50% found for the fair game with the same start and target. A modest tilt against the player, a 10% loss on average per unit staked in this example, moves the chance of ruin a long way. Reaching the target is now roughly a 12% outcome.
If play has no target and no stopping point, the position is starker. When p is 0.5 or below, ruin eventually happens with probability 1 given unlimited play, no matter how large the starting pool. A larger pool only delays it. This is the mathematical core of why long play in a game with a house edge tends toward depletion.
Case 3: a favourable game
A game where the player has the advantage is largely hypothetical in ordinary casino play, but it shows the effect of pool size. If p > 0.5 and play continues indefinitely, the chance of eventually being ruined is r^a, where r = q/p is below 1.
Take p = 0.55, so q = 0.45 and r is about 0.818:
- With a = 10 units, r^10 is about 0.134, so roughly 13%.
- With a = 20 units, r^20 is about 0.018, so roughly 1.8%.
Doubling the pool cuts the risk by far more than half, because the figure shrinks with each added unit. Even with an advantage, ruin has a real chance of occurring when the pool is small relative to the stake. That is why variance matters: it connects to the standard deviation of each play, and a larger stake relative to the pool raises the chance of an early wipe-out.
What the formulas assume
These results come from a tidy model. Real situations break the assumptions in several ways:
- Unknown odds. The formulas need p. If p is estimated from limited data, small errors change the answer a great deal, and confidence intervals around an estimate of p are often wide.
- Varying stakes. If bet sizes change, the simple formulas no longer apply.
- Payouts that are not even money. Different payout structures require different calculations, though the principle is the same.
- Dependence between plays. Games where outcomes interact need other methods.
- Fees, limits and rule changes. Table limits, fees and changes in conditions are not captured.
How ruin is estimated in practice
When the formula is not enough, analysts use simulation. A Monte Carlo simulation plays out the game many thousands of times on a computer, using the same rules, and counts the share of runs that reach zero. If 1,000 simulated runs hit zero in 880 of them, the estimate is 88%, with some sampling error that shrinks as the number of runs grows.
Related questions are how long a pool tends to last, known as expected duration, and how the answer depends on the pool-to-stake ratio. In negative-expectation games, the typical time to depletion falls as stakes rise relative to the pool.
How to read a risk-of-ruin figure
A few cautions help:
- A low risk of ruin in a game with a positive average does not make the game safe in a practical sense, because a loss can still be large or sit alongside other pressures.
- A high figure in a losing game is a statement about the game’s structure. Lowering stakes can lower the number only by delaying the same drift.
- No figure here predicts what will happen on a particular night.
- The calculation does not offer a method for winning, and a game with a built-in disadvantage cannot be turned into a profitable one by choosing stakes.
Gambling can lead to real harm. Money that is needed for housing, food, debts or family should never be part of the pool, and anyone finding it hard to stop can contact a local support service.