Home › FX Trading School › Advanced

Monte Carlo Analysis of Your Equity Curve

Advanced Updated 14 July 2026 · 9 min read · PipTax education

Monte Carlo analysis takes the single equity curve your backtest or live track record produced and asks a harder question: what else could plausibly have happened with the same trades, in a different order or a different sample? This is Module 17 of the PipTax FX Trading School, and it assumes you've already covered backtest validation (Module 15) and walk-forward testing (Module 16) — Monte Carlo doesn't replace either, it adds a layer on top once you already trust your data isn't curve-fitted.

Why One Equity Curve Isn't Enough

Every backtest report shows you exactly one path: the specific sequence of wins and losses that happened to occur in that historical window. Traders routinely treat that single line as "the" performance of a strategy — but it's one draw from a much wider range of outcomes the strategy's edge (or lack of it) could have produced.

Consider a system with 55% win rate and a 1:1.5 reward-to-risk ratio, tested over 200 trades:

• The historical order might have front-loaded the losses early and the wins late, producing a deep initial drawdown that looks alarming.
• A different order of the exact same 200 trades might show a smooth curve with barely any drawdown at all.
• Both are equally "real" outcomes of the same edge — the order is partly down to chance.

If you only ever look at the one curve you got, you're anchoring your risk expectations to a single roll of the dice. Monte Carlo analysis fixes this by generating thousands of alternative rolls from the same underlying trade data, so you can see the spread of what's plausible rather than the one thing that happened.

This matters most for position sizing and drawdown tolerance — decisions you make once, in advance, that then have to survive whichever path actually unfolds live.

The Two Core Monte Carlo Methods

There are two standard ways to generate alternative equity curves from a completed set of trade results. They ask subtly different questions, so it's worth knowing both.

1. Trade-order shuffling (permutation) - Takes your exact list of trade outcomes (in £, %, or R-multiples) and randomly reorders them. - Every simulated curve uses the same trades — same win rate, same average win/loss — just a different sequence. - Answers: "Given this exact edge and sample, how much does sequencing risk alone affect my drawdown?"

2. Trade resampling (bootstrapping with replacement) - Randomly draws trades from your historical set, allowing repeats, until you have a new "path" of the same length. - Some trades appear multiple times, others not at all — this varies win rate and average trade slightly between runs, not just order. - Answers: "Given the uncertainty in my sample itself, what range of outcomes is consistent with this edge?"

Shuffling is simpler and a good starting point. Bootstrapping is more rigorous for small samples because it also captures sampling uncertainty, not just ordering. Serious validation work — the kind PipTax's own methodology leans on — tends to run both and compare.

Setting Up the Simulation

You don't need proprietary software. A spreadsheet with a random-shuffle function, or a short Python/R script, will do the job. The workflow is the same either way:

1. Export your trade log — closed trades only, with the £ or % result of each, net of realistic costs. 2. Choose your method — shuffle for a quick check, bootstrap for a fuller picture. 3. Set the run count — 1,000 to 10,000 simulated equity curves is standard. 4. Apply your real starting balance and position sizing rules — the simulation should mirror how you'd actually size trades, not just sum raw R-multiples. 5. Record three outputs per run: final equity, maximum drawdown, and whether the account breached a stop-out or "system-off" threshold.

Once you have thousands of runs, you build a distribution — not a single number — for each of those three outputs. That distribution is the actual product of a Monte Carlo analysis, and it's what you review next.

Reading the Output: Drawdown Distribution and Ruin Risk

The headline numbers to pull out of your simulation results are:

| Metric | What it tells you | |---|---| | Median max drawdown | The "typical" worst dip you should expect, better than any single backtest figure | | 95th percentile max drawdown | A realistic worst-case you should be prepared to sit through | | Probability of ruin | % of simulated paths that hit your defined stop-out level (e.g. -20% or margin call) | | Spread of final equity | How wide the range of plausible outcomes is — a tight spread is reassuring, a wide one is a warning |

A strategy that looks fine on its one historical curve can still show a 95th-percentile drawdown of 35–40% under simulation — a level that would force most traders to abandon it emotionally, or breach a prop firm's daily loss limit, long before the "average" recovery arrives. This is exactly the gap Monte Carlo analysis is built to expose: it turns a vague sense of "this feels risky" into a specific, checkable number you can size against.

What Monte Carlo Cannot Do

It's worth being blunt about the limits here, because Monte Carlo analysis gets oversold online.

• It cannot create edge that isn't there. Shuffling a losing system's trades still produces, on average, a losing distribution — just reshuffled.
• It inherits your sample's biases. If your backtest was curve-fitted, over-optimised, or drawn from one unusual regime, every simulated path carries that same flaw forward.
• It doesn't model regime change. A strategy tested only in trending conditions will look robust in simulation right up until real markets turn choppy — the simulation only knows the trades you fed it.
• It's not a replacement for walk-forward testing. Monte Carlo tells you about sequencing and sampling risk within a fixed dataset; walk-forward testing tells you whether the strategy holds up on genuinely unseen data. You need both.

Treat it as one validation layer among several, not a stamp of approval on its own.

Building Monte Carlo Into Your Validation Workflow

For this to be useful in practice, it needs to sit in a sequence, not run in isolation:

1. Validate the strategy logic and avoid look-ahead bias (Module 15). 2. Confirm it holds up out-of-sample with walk-forward testing (Module 16). 3. Run Monte Carlo analysis on the resulting trade set to map drawdown and ruin risk (this module). 4. Feed realistic broker costs into every stage — spread, commission, and swap all shrink your edge and widen your drawdowns. Check what Pepperstone or IG would actually charge on your instrument and account type using the cost tool at /audit.html before you finalise the trade log you simulate. 5. Set position sizing and a maximum-loss circuit breaker based on the 95th-percentile drawdown, not the median. 6. Re-run the analysis periodically as new trades accumulate — a strategy's risk profile isn't static.

Done properly, Monte Carlo analysis won't make your strategy look better. Its job is to make your risk expectations more honest before real money is on the line, which is a far more useful outcome than a smoother-looking backtest.

Keep going: Audit Methodology Index Index