The Analytics → Risk segment runs a Monte Carlo simulation on your own closed trades. It answers one question: if you kept trading exactly the way you have been, how much of your account could you reasonably expect to be down at the worst point along the way?
What it actually does
It does not predict prices, and it does not model the market at all. It reshuffles your results:
- Every closed trade is converted to a percentage of the account it was taken on — so a $500 win on a $50,000 account becomes +1%.
- One simulated run draws that many results at random, with replacement, from your own history — the same wins and losses you actually had, in a different order, and some appearing more than once.
- That is repeated 5,000 times. The chart shows a sample of the resulting equity paths; the figures underneath summarise all of them.
Two controls change the question. Number of trades is how far ahead you are asking about — 100 trades might be a quarter for one trader and two years for another. Ruin threshold is what you would call a blown account: the percentage of the starting balance you never want to fall below.
Reading the four figures
- Median final balance — the middle outcome. Half the runs finished above it, half below. It is deliberately the median and not the average: a handful of very good runs drag an average upwards and make the picture look safer than it is.
- Max drawdown — the typical worst peak-to-trough fall within a run. This is usually the most useful number on the page, because it is the one that decides whether you would still be trading the system when it happens.
- Probability of ruin — the share of the 5,000 runs that touched your ruin threshold at any point. It counts a run as ruined even if it later recovered, which is correct: in real life you would have stopped.
- Probability of profit — the share of runs that finished above the starting balance.
Why the same settings give slightly different numbers
The draw is random, so re-running shifts the figures a little. Treat them as a range, not a reading — “roughly a one-in-twenty chance” is the level of precision on offer here, not “4.8%”. Editing or adding a trade changes the input set and will move them properly.
What it cannot tell you — read this part
- It assumes your future looks like your past. If your edge decays, your size changes, or the market regime turns over, the simulation has no way to know.
- It assumes trades are independent. Real losing streaks cluster — a bad regime hits several positions at once, and revenge trading is correlated by definition. Reshuffling breaks that clustering, so real drawdowns tend to be worse than the simulated ones, not better.
- A short history gives a confident-looking answer to a question it cannot answer. Below about 30 closed trades you are mostly reshuffling noise. The card needs 5 trades to draw anything at all; that is a floor, not a recommendation.
- It uses P&L as a share of the account size on each trade, so a book where position sizing changed a lot over time mixes two different systems together.
None of this is a forecast or a recommendation — it is a description of your own recorded results, rearranged. See also the Metrics glossary and MFE / MAE, and the risk-of-ruin and Kelly calculators on the Tools page.