A backtest with 200 trades feels like certainty: one curve, one final balance, one verdict. But that curve is just one of many orders in which the same winners and losers could have arrived. Reshuffle the order and the same edge produces a different account — sometimes a clearly better one, sometimes one you would have closed in frustration halfway through.
That is exactly what a Monte Carlo simulation makes visible. It takes three sober numbers — win rate, average win in R, risk per trade — and plays out not one future but a thousand. The result is not a line but a fan. And that fan is the most honest picture you can get of a trading system before real money sits inside it.
Language first: R, not dollars
The simulator thinks in R multiples, like everything else in this house: a loss is 1 R by definition — the amount you put at risk per trade. A win is a multiple of that, the payoff. Win rate and payoff together give the expectancy per trade:
Expectancy = win rate × payoff − loss rate × 1 R.
At a 45 percent win rate and a 2 R payoff that is +0.35 R per trade. It sounds unspectacular — and it should. Because whether +0.35 R per trade becomes a growing account is decided not by the expectancy alone but by its dispersion. You can see it below.
Monte Carlo simulator
A backtest is ONE path. The same edge, rolled 1,000 times, is a distribution. Set your numbers — ideally the ones from your journal, not the ones from your wishes.
1,000 runs, deterministic seed — same inputs, same picture.
Expectancy per trade
+0.35 R
Workable edge: what matters now is whether you can sit through the losing streaks in the lower left without switching systems.
Median ending
+97 %
P5 ending (weak branch)
+42 %
P95 ending (strong branch)
+174 %
Max drawdown (median / P95)
−10 % / −15 %
Longest losing streak (median / P95)
8 / 12 losers in a row
The kill-switch comparison
Both rows come from the same rolls — the only difference is the rule to stop trading at −25%.
| Runs touching −25%: 0 % | Median ending | P5 ending (weak branch) |
|---|---|---|
| Without stop | +97 % | +42 % |
| With stop | +97 % | +42 % |
In this setup hardly any run touches the level — the kill switch costs nothing and still stands guard.
Lesson one: the median is not your backtest
Look first at the three ending numbers under the fan. The median is the middle outcome — half of all runs end below it. The P95 ending is the strong branch: the run you would put in the brochure if you were a guru. The P5 ending is the weak branch: same edge, same discipline, different dice.
The distance between those branches is not a weakness of the system. It is the system. Whoever has seen only one curve mistakes the strong branch for skill and the weak one for a malfunction. Whoever has seen the fan knows: both were in play from the start.
Lesson two: losing streaks are the norm, not an accident
The number that actually kills accounts sits in the "longest losing streak" row. At a 45 percent win rate across 200 trades, the longest run of losers lands at eight to nine in a row in the median — not as bad luck, but as a mathematical baseline. The P95 run adds a few more.
You should know that number before it happens. The most expensive reaction to a normal losing streak is to switch systems in the middle of it — abandoning the edge right before it pays. Loss aversion reliably makes eight losers in a row feel like twenty.
Lesson three: what a kill switch actually does
The comparison table at the end of the simulator is the core of the tool. Both rows come from the same rolls; the only difference is one rule: if the account falls below the protection level, trading stops and the remaining capital is frozen.
The result surprises most people the first time: the median barely moves. In the normal runs a kill switch costs almost nothing — it never triggers. But it caps the weak branch, precisely where a drawdown turns into the debt trap of percentages: −50 percent needs +100 percent to get back.
So a kill switch does not improve your edge. It makes sure you are still around on the day the edge plays out. That is why, in the WVPO method, it is not a recommendation but part of the contract — a mechanical safeguard that sits next to every trade in the WVPO trading journal, not in a good intention.
Where the numbers must come from
A Monte Carlo simulator is exactly as honest as its inputs. Feed it a wish-list 65 percent win rate with a 3 R payoff and it will produce the same beautiful fans as any other tool — garbage in, garbage out, just prettier.
The reliable source for win rate and payoff is a maintained journal: documented trades of the same setup, under the same rules, across enough cases. That is what the edge database exists for — every decision is recorded, and the numbers you dial in here come from that record. Only then does the fan show your future rather than your imagination. How to choose the risk per trade that the simulator uses as its lever is covered in the position sizing article.
Note: Teaching model with simplified assumptions (independent trades, constant win rate, fixed payoff, no costs and no slippage). Not a return promise, not investment advice, not a trading signal. The risk disclosure applies.
Reality is less convenient than any model: win rates drift with the regime, payoffs scatter, and the bad stretches like to arrive in clusters. That is why the method starts not with the simulation but with the filter — which market phase gets traded at all. The simulation then answers the second question: whether you survive what remains, financially and psychologically.
Subscribe to the engine room
New posts by email — double opt-in, unsubscribe anytime, no spam.