Risk of Ruin: The Number That Decides Position Size

9 min readQuantParadox research

A positive expectancy says you make money eventually. Risk of ruin asks whether you are still there when eventually arrives, and the two questions have very different answers.

The short answer

Risk of ruin is the probability that a strategy loses a defined portion of the account before it compounds, and it depends far more strongly on position size than on the size of the edge, which is why a profitable strategy can still ruin an account that bets too much.

What is risk of ruin?

Risk of ruin is the probability that an account falls to some defined level — zero, or more usefully a level at which you would stop — before the strategy's edge has time to accumulate.

It exists because expectancy is an average over the long run and an account is a finite thing traded in the short run. A strategy that earns +0.3R per trade over a thousand trades can still lose fifteen in a row somewhere in there, and whether that matters depends entirely on how much each one costs.

The defined stopping level matters more than zero does. Almost nobody trades to a literal zero balance; they stop at some drawdown that destroys their confidence or breaches a prop firm's rule. Computing ruin against that realistic threshold gives a number that describes an actual risk.

It is a probability rather than a prediction, and it is one of the few places in trading where the arithmetic is unambiguous.

What drives risk of ruin?

Three inputs drive it, and they are nowhere near equal in weight: position size dominates, edge matters, and the shape of the return distribution matters more than most people expect.

Position size is the lever with the most leverage, in every sense. Halving risk per trade does not halve risk of ruin — it collapses it, because ruin requires a specific run of losses and each halving roughly doubles the number of consecutive losses needed to get there.

Edge matters in the obvious direction and with surprisingly little force at the sizes retail traders use. Going from +0.2R to +0.4R expectancy improves the odds meaningfully; going from 2% risk per trade to 1% improves them far more.

The distribution shape is the input most often ignored. A strategy whose losses cluster — and most do, because market conditions persist — has a much higher risk of ruin than an independence assumption suggests, which is why block resampling matters when estimating it.

How do you actually calculate it?

The practical method is simulation rather than a closed-form formula: resample your own trade sequence thousands of times and count what fraction of the simulated paths breach your stopping level.

Closed-form risk-of-ruin equations exist and assume fixed win rates, fixed payoffs and independent trades — three things that are never quite true of a real strategy. They are useful for intuition and misleading in the third decimal place.

The simulation version is straightforward. Take the historical R-multiples, draw sequences from them with replacement, apply your position sizing rule to each path, and record how many paths hit the floor. Ten thousand paths takes seconds and gives a distribution rather than a point estimate.

Use block resampling if the strategy clusters its losses. Drawing individual trades independently breaks the runs, and the runs are precisely the mechanism by which ruin happens.

What risk per trade does that imply?

For most retail strategies, risking somewhere between 0.5% and 2% of the account per trade keeps the simulated risk of ruin small, and the upper end of that range is already aggressive for a strategy with a modest edge.

The number to work backwards from is your own tolerance. Decide the drawdown at which you would stop, run the simulation at several risk levels, and pick the one where the probability of reaching that level is acceptably low over the horizon you intend to trade.

This is also where fixed-fractional sizing earns its reputation. Risking a percentage of the CURRENT balance rather than a fixed amount means position size falls automatically during a drawdown, which lengthens the losing run required to reach the floor.

The corresponding cost is honest: smaller size means slower compounding. The trade is between growth and survival, and survival has to win, because the account that is not there does not get the good year.

It is worth running the same simulation with the edge reduced, too. Assume the expectancy that survives out of sample is half what the backtest showed — which is a common outcome rather than a pessimistic one — and see what position size keeps ruin acceptable under that weaker assumption. Sizing for the edge you hope you have is how a survivable strategy becomes an unsurvivable one.

Questions people actually ask

Can a profitable strategy still blow up an account?

Yes, and this is the entire reason risk of ruin is worth computing. Positive expectancy describes an average over many trades; a finite account experiences a specific sequence, and a bad enough sequence at a large enough position size ends the account before the average has time to assert itself.

Is 2% risk per trade safe?

The commonly cited 2% figure is reasonable for a strategy with a solid edge and independent-ish trades, and it is aggressive for a marginal edge or a strategy whose losses cluster. Rather than adopting a number, run the resampling at 0.5%, 1% and 2% against your own trade history and look at how the probability of hitting your personal stopping level changes.

Does risk of ruin apply to prop firm accounts?

It applies with a much closer floor, which changes the answer substantially. A prop account with a 5% daily and 10% overall drawdown limit has a stopping level far nearer than a personal account's, so the position size that keeps ruin acceptable there is considerably smaller — and simulating against the firm's specific rules is the only way to size it honestly.

The only backtest that settles it is yours.

Build a strategy from a sentence, paste your own Python, or import your live trade history and have it graded. Five full backtests free, no card, and we'll tell you plainly when the result is indistinguishable from luck.

We publish research and tooling, not trading advice, and we make no claim about future returns. Everything above describes how to test an idea — not a reason to trade one.