TradeHouse Fidelity Lab · Guides
Monte Carlo backtests
A Monte Carlo backtest takes the trades or returns from one backtest and generates many alternative histories from them. Instead of one equity curve you get a distribution, which shows how much of the result depends on the particular order in which things happened.
Common variants
| Variant | How it works | What it reveals |
|---|---|---|
| Trade reshuffle | Same trades, random order | Range of maximum drawdowns. Total profit does not change. |
| Bootstrap with replacement | Draw trades at random, allowing repeats, to build new sequences | Spread of total return and drawdown under the same trade distribution |
| Block bootstrap | Draw consecutive blocks of returns instead of single trades | Keeps short-range autocorrelation, so clustered losses are not broken up |
| Parameter or cost perturbation | Re-run the backtest with randomised settings | Sensitivity of the edge to settings; see robustness tests |
What to read from the output
- The 95th percentile drawdown is a better sizing input than the single drawdown your backtest happened to show.
- The share of simulated paths that end below zero shows how fragile the total return is.
- A very narrow distribution can mean the sample is dominated by a few large trades.
The assumption that breaks it
Resampling assumes the trades you have are a fair sample of the trades you will get. If the backtest is overfit or contaminated, every simulated path inherits the same fault. Resampling a stale-quote fill a thousand times produces a thousand profitable paths. That is why a Monte Carlo study belongs after data checks and an out-of-sample test, not instead of them.
How Fidelity Lab relates
Fidelity Lab does not run trade-order Monte Carlo. It checks the inputs that Monte Carlo takes on trust: contamination in the legs, real fees, whether a frozen rule survives on later data, and whether the number of independent days is large enough. Clustered trades are handled through a day-clustered effective sample size with lag-1 to lag-5 autocorrelation inflation.
Try it on your own trades. The free Fidelity Lab tier runs one report a month on up to 5,000 legs: all 11 contamination detectors, cost re-netting and the walk-forward verdict. Run a free report or see pricing.
Frequently asked questions
Does reshuffling trades change total profit?
No. Reshuffling only changes the path, so it affects drawdown and streaks but not the final sum.
How many Monte Carlo runs are enough?
Enough that the percentiles you care about stop moving when you add more runs. A few thousand is common for drawdown percentiles.
Related guides
- Backtest overfitting: a practical guide
- Walk-forward validation vs cross-validation
- Backtest validation tool: what it should check
- Probability of backtest overfitting (PBO)
- Trading strategy audit service
- Overfitting in backtrader
- Trading strategy robustness tests
- Look-ahead bias in backtests
- Survivorship bias in backtests
- Data snooping in trading
- Out-of-sample testing in trading
- What is backtest overfitting?
- Deflated Sharpe ratio calculator
- Fidelity Lab
Educational content, not investment advice. Updated 2026-10-28.