Why this is simulated rather than solved
Closed-form risk-of-ruin formulas exist and they assume things trading does not provide: fixed bet sizes, even payoffs, or an infinite horizon. Each version answers a slightly different question, which is why published formulas disagree with each other.
Simulation avoids the problem by asking directly. Two thousand sequences are generated at your win rate and your sizing, each trade compounding on the balance that remains, and the answer is simply how many of them hit your ruin level.
That also makes the assumptions inspectable rather than hidden inside a formula:
- Fixed-fractional sizing — each trade risks the same percentage of the current balance.
- Independent outcomes — each trade is drawn separately.
- Constant edge — the win rate does not change.
Sizing is the whole answer, and it is not linear
Take a solid edge — a 45% win rate with wins at 2R and losses at 1R, an expectancy of +0.35R per trade — and hold everything constant except the size risked. Over 200 trades, against a 50% ruin level:
| Risk per trade | Chance of a 50% drawdown |
|---|---|
| 1% | ~0% |
| 2% | ~0% |
| 5% | 1.5% |
| 8% | 8.7% |
| 10% | 16% |
| 15% | 36% |
Same edge in every row. Only the position size changed.
Two things are worth taking from that. At 1–2% per trade, ruin is genuinely off the table for a strategy with a real edge — which is the honest answer and the reason those figures are the convention. And the relationship is not proportional: five times the risk is a hundred times the ruin probability, because losses compound against a shrinking balance.
Now weaken the edge slightly, to a 40% win rate at 1.8R, and the same three sizes give 0.5%, 21% and 60%. A modest deterioration in the edge does not shift the curve a little; it moves the whole thing. Which is why an edge estimated from a short sample — with the wide error bar the expectancy calculator puts on it — is dangerous to size against aggressively.
An edge decides where you end up; position size decides what you go through on the way. Only the second is under your control.
Ruin is not zero
The ruin level is a field because zero is the wrong threshold.
Nobody trades an account to zero. They stop — at 30%, at 50%, at whatever point the method stops feeling like a method. That is the real ruin level, and it is far closer than the arithmetic one.
Set it to the drawdown at which you would genuinely stop, not to the one that empties the account. The losing streak calculator shows why the two are so far apart: at 2% per trade it takes 35 consecutive losses to halve an account, and almost nobody is still following the plan by then.
The assumption that flatters the result
Independence. Real losses cluster — a strategy that suits one regime fails through all of it, and correlated positions lose together.
Clustered losses are strictly worse than independent ones for this calculation, so the true probability is higher than the figure shown. Treat the output as an optimistic bound rather than an estimate, and size for something worse than it says.