Risk of Ruin: The Math That Decides Whether You Blow Up Before Your Edge Pays Off

A trading system with a genuine, positive edge can still go to zero. Not because the edge stopped working — because a losing streak that was always statistically going to happen showed up before the edge had enough trades to prove itself. Risk of ruin is the name for that probability, and almost nobody sizes their trades with it in mind.

What risk of ruin actually measures

Expectancy tells you what a system does on average, over a large number of trades. Risk of ruin tells you something different: the probability that you don't survive long enough for "on average" to mean anything. Averages are a long-run property. A losing streak is a short-run event. A system can have positive expectancy and still carry a real chance of wiping out the account in the short run, because the short run doesn't know about the long run yet.

This is the gap retail traders miss. They ask "is my edge real?" and stop there, as if a real edge guarantees survival. It doesn't. Survival depends on a second variable entirely — how much you risk on each trade — and that variable can turn a profitable system into a blown account with no change to the edge at all.

The formula

Here's the simplified version, the one that isolates the effect this post is about. Assume every winning trade and every losing trade risks the same unit of size (a 1:1 payoff), so the only thing varying is how big that unit is relative to your account. Let p be your win rate, q = 1 − p your loss rate, and A = p − q your edge. Divide your capital into risk units, where N is how many of those units fit in your account — which is just 1 divided by the percentage you risk per trade. Risk 1% per trade and N is 100. Risk 10% and N is 10.

Risk of ruin ≈ ((1 − A) / (1 + A)) ^ N

This is the classic gambler's-ruin equation, and it has one property worth sitting with: N is an exponent, not a multiplier. Halving your risk per trade doesn't halve your risk of ruin. It doubles N, and doubling an exponent on a number less than one collapses the result far faster than linear intuition expects.

Take a system with a 55% win rate and a 1:1 payoff — a real edge, since it makes 10 cents of expectancy per dollar risked on average. Risk 10% per trade and N is 10; the formula gives a risk of ruin around 13%. Risk 1% per trade on the exact same system and N is 100; risk of ruin drops to a number so small it rounds to zero. Same edge. Same win rate. Same market. The only thing that moved was sizing, and it moved the outcome from "roughly one blown account in eight" to "essentially never."

The table: same edge, only sizing moves

Hold the system fixed — 55% win rate, 1:1 payoff, edge of 10% — and change nothing but risk per trade.

• Risk 1% per trade (100 units) → risk of ruin ≈ 0.0000002%
• Risk 2% per trade (50 units) → risk of ruin ≈ 0.004%
• Risk 5% per trade (20 units) → risk of ruin ≈ 1.8%
• Risk 10% per trade (10 units) → risk of ruin ≈ 13.4%
• Risk 20% per trade (5 units) → risk of ruin ≈ 36.7%
• Risk 33% per trade (3 units) → risk of ruin ≈ 54.8%
• Risk 50% per trade (2 units) → risk of ruin ≈ 66.9%

Look at where the curve turns. Between 1% and 2% risk, ruin probability is a rounding error either way. By 10% it's a coin flip's cousin. By 20% it's worse odds than a fair coin flip on whether the account survives at all — on a system that, per trade, is profitable. The edge never left the building. The sizing just gave the losing streak enough room to end the game before the edge could compound.

Why the curve bends this way

N is also, roughly, how many consecutive losses it takes to wipe out the account. At 1% risk, that's on the order of 100 straight losers — something a 55%-win-rate system essentially never produces in a realistic sample. At 20% risk, it's 5. A system with a 45% loss rate throwing five losers in a row isn't a freak event; it's a Tuesday. The math isn't punishing the trader for being wrong about their edge. It's punishing them for handing a normal losing streak enough leverage to be fatal.

This is also why "I'll just size up while I'm confident" is exactly backwards. Confidence doesn't change q. The losing streak that ends the account doesn't announce itself as a losing streak in advance — it looks like every other trade, right up until it's the fifth one in a row and the size behind it was set by how good the setup felt, not by how many units of capital were left standing after it.

Survival is a sizing decision, not a forecast

None of this requires being wrong about the market. A trader can have a real, back-tested, statistically valid edge and still ruin the account, because risk of ruin isn't a statement about whether you're right — it's a statement about whether you gave a normal losing streak enough room to end things before your edge got to compound. Two traders with the identical system and identical win rate can have wildly different outcomes, and the only variable that explains it is how much of the account each one put behind a single trade.

On the institutional desk, this wasn't left to the trader's judgment in the moment. I remember being down a set amount once and still trying to trade back to even, sizing up as I went, when a risk manager walked over and closed my session — the limit was never mine to override. The point wasn't that my read on the market was wrong. It was that the size I was running had already moved me into the range of this curve where a normal bad stretch stops being survivable, and nobody was going to let that get decided by how confident I felt at the time.

What to do with this tomorrow

Before your next session, work out the actual number: take your honest win rate and payoff ratio, plug them into the formula, and find the risk-per-trade where ruin probability drops below roughly 1–2% — for most retail systems with a real edge, that lands somewhere between 0.5% and 2% of the account per trade, not the 5–10% that feels more exciting to run. Then size to that number on every trade, including the ones that feel like a sure thing, because the losing streak that tests this math has never once announced itself as a losing streak in advance.

I built Fourdesk's Risk Manager desk around exactly this idea — it enforces a sizing limit on every trade so a normal losing streak can't quietly move you into the range where the math stops forgiving.