Monte Carlo, Variance and Risk of Ruin
Your backtest is one path. Monte Carlo shows the range of paths your edge could realistically produce.
Your backtest is one path through history. Your actual trading will be a different path, drawn from a distribution your backtest only samples once.
Monte Carlo methods take the trades your strategy produced and ask a different question: given this edge, what range of outcomes was actually possible?
Reshuffling your own trades
The simplest useful method takes your list of trade results and randomly reorders them thousands of times. Each shuffle produces a different equity curve with the same trades in a different sequence.
The point is that sequence matters enormously for drawdown. Ten losses followed by ten wins and ten wins followed by ten losses have identical totals and completely different psychological and practical consequences. Reshuffling reveals the worst realistic ordering.
Monte Carlo Simulation
Simulation Parameters
Method: one synthetic list of 100 trades (45 wins, 55 losses) is reshuffled 1000 times. Every path holds the same trades — only the order changes. Stats use all 1000 shuffles; the chart draws 20. The trade list is synthetic and the two inputs above may be illustrative, so treat every number here as a worked example, not a forecast.
Projected Outcomes
Risk of ruin
Risk of ruin is the probability that your account falls to a level from which you cannot continue. It is not the probability of a losing trade — it is the probability of the sequence ending your participation.
It rises sharply with risk per trade. Risking 1% per trade makes ruin very unlikely under most assumptions. Risking 5% makes it a genuine possibility within a few hundred trades, even with a positive expectancy.
Losing streaks are normal
For a strategy with a 40% win rate, a run of ten consecutive losses has a meaningful probability of occurring within a few hundred trades. It is not evidence the strategy broke. It is arithmetic.
Knowing the likely worst streak in advance is what stops you quitting or interfering at exactly the wrong moment. If you have not calculated it, you will interpret a normal streak as a failure and change something that was working.
Size for the bad path
The practical output of this analysis is a position size. Look at the fifth percentile outcome across thousands of simulations — the bad but plausible path — and choose a size you could follow through that path without abandoning the method.
Most traders size for the median outcome and are then surprised and derailed by the lower tail. Sizing for the tail is what actually keeps people in the game long enough for expectancy to work.
Illustrative worst-case consecutive losses by win rate over 500 trades
| Win rate | Typical longest losing run |
|---|---|
| 60% | 7–8 losing trades |
| 50% | 9–10 losing trades |
| 40% | 11–12 losing trades |
| 30% | 15+ losing trades |
Key takeaways
- Your backtest is one path; Monte Carlo shows the range of paths your edge allows
- Risk of ruin rises very sharply with the percentage risked per trade
- A ten-trade losing streak is statistically normal for a 40% win rate
- Size for the fifth percentile outcome, not the median one
- Reshuffling trade order reveals worst-case drawdowns
- Risk of ruin rises sharply with risk taken per trade
- A losing streak of ten is statistically normal, not a broken strategy
- Size for the bad path rather than the average one
Sizing for the backtest's best run instead of its worst plausible one.
Reading is not verification
Take the concept you just read and turn it into explicit rules, then test it. That is the only way to know whether it actually works.
Build a strategy