One backtest is one roll of the dice
When you backtest a strategy you get a single equity curve — one specific ordering of every win and loss. But that order was an accident of history. Take the exact same trades and shuffle them, and the curve transforms: in one shuffle the winners cluster early and you ride a fat cushion through the rough patch; in another, a savage losing streak hits in the first thirty trades and the account is margin-called before a single big winner lands. Same edge, same win rate, same reward:risk — completely different fate. Judging a strategy by one backtest is like judging a coin as "lucky" because it came up heads. Monte Carlo deals the hand thousands of times so you see the whole range, not the one history you happened to get.
What the percentile table is telling you
Each simulated run compounds your balance trade-by-trade: a win adds risk × reward:risk percent, a loss subtracts risk percent, sequenced randomly at your win rate. The table sorts thousands of those runs and shows you the spread. The median (50th percentile) is the outcome you should actually expect. The 5th percentile is your realistic bad-luck scenario — if that column shows a blown account, your strategy can ruin you even though the median looks great, and the fix is almost always a smaller risk-per-trade, not a better entry. The 95th percentile is the lucky run that, if it happened to be your live one, would fool you into over-sizing. Plan around the median and survive the 5th.
Probability of ruin is the number that keeps you alive
The headline most traders chase is the median return. The number that actually decides whether you're trading in a year is the probability of ruin — the share of runs that breached your drawdown threshold at any point. A strategy with a glorious median and a 15% ruin rate is a slot machine: profitable on average, but with a one-in-seven chance of wiping you out before the average arrives. You only get one live run, so a tail that blows up even 5% of the time is a tail you cannot accept. Push the risk-per-trade down until that column reads near zero, then read the risk of ruin calculator and the losing streak simulator for the same truth from two other angles.
How to use it
1. Enter your strategy's real, fee-adjusted win rate and its average reward:risk.
2. Set the risk per trade you size with, and how many trades a run represents.
3. Choose the drawdown that counts as "blown" for you, then read the median, the profit odds, and — most importantly — how often the 5th-percentile run ruins the account.
FAQ
Does this assume independent trades? Yes — each trade is an independent draw at your win rate with a fixed reward:risk. Real markets add streaks of correlation and changing volatility, which make the unlucky tail somewhat worse, so treat the ruin probability as an optimistic floor.
Why does my median return differ from win-rate × reward? Because returns compound. A loss shrinks the balance every subsequent win is calculated on, so the geometric (compounded) outcome is always below the naive arithmetic expectancy — which is exactly why sizing matters so much.
What inputs should I trust? Use a win rate and reward:risk from a meaningful sample of real trades, not your best month. If you don't have one, run a range (e.g. 45/55/65% win rate) and watch how fast the ruin tail grows as the edge thins.