Monte Carlo Backtesting: Simulate Trading Risk & Drawdowns

A traditional backtest tells you how a trading system performed along one single historical path. A Monte Carlo backtest tells you how that same system would perform across thousands of alternative realities where the trades occurred in different sequences.

If you test a strategy over 100 historical trades and see a maximum drawdown of 8%, that 8% figure is path-dependent. If market noise had clustered four of your losing trades together during week two instead of spreading them across six months, your drawdown might have hit 22%.

Monte Carlo simulation eliminates sequence bias. It shows whether your survival depends on true statistical edge or just on a lucky sequence of trades.

The Flaw of Single-Path Backtesting: Sequence Risk

Consider two traders who both take the exact same 100 trades over a year. Both finish with:

55 winning trades averaging +$300 each (+$16,500)
45 losing trades averaging -$200 each (-$9,000)
Total net profit: +$7,500 on a $10,000 starting account (75% return)

Trader A (Lucky Chronological Order)

In the historical market, winners and losers alternated smoothly: W, L, W, W, L, W, L...

Maximum consecutive losses: 3
Maximum historical drawdown: $750 (7.5%)
Emotional result: Calm, confident execution.

Trader B (Clustered Sequence Risk)

In an alternate market sequence, 8 losing trades clumped together during month three: L, L, L, L, L, L, L, L, W, L...

Maximum consecutive losses: 8
Maximum drawdown: $2,100 (21.0%)
Emotional result: Self-doubt, abandoned strategy, or margin reduction right before the winning streak began.

Both equity curves contain the exact same trades. But Trader B faced three times the drawdown because of sequence risk. Monte Carlo backtesting models this clustering before you trade real capital.

How Monte Carlo Simulation Works: Permutation vs Bootstrap

Quantitative trading platforms use two primary simulation algorithms:

Historical Trade List: [+$300, -$200, +$150, -$180, +$420, -$210, ...]
<div class="overflow-x-auto">

                               |

</div>

               ---------------------------------
<div class="overflow-x-auto">

               |                               |

</div>

    Method 1: Permutation             Method 2: Bootstrap
  (Without Replacement)              (With Replacement)
Exact same 100 trades shuffled     Random draws; duplicate trades allowed
Total Net Profit is constant       Total Net Profit varies per run
Isolates pure sequence risk        Tests sample & sequence variance

Method 1: Shuffling Without Replacement (Permutation Test)

The algorithm keeps all original trade outcomes and randomly reorders them.

Run 1: Trade #42, Trade #12, Trade #99, Trade #3...
Run 2: Trade #17, Trade #84, Trade #02, Trade #61...

Because every trade is used exactly once per simulation run:

Final Net Profit remains identical across all runs ($7,500 in our example).
Maximum Drawdown and Drawdown Duration vary wildly. This method isolates pure sequence risk without changing trade expectancy.

Method 2: Resampling With Replacement (Bootstrap)

For each simulated trade, the algorithm selects a random trade from your history, records it, and puts it back into the pool. A single great trade can be picked three times; a bad loss can be picked four times or skipped entirely.

Final net profit varies across runs.
Generates a wider distribution of potential outcomes.
Reveals what happens if future win rates drop below historical averages.

Worked Simulation: Reading a 1,000-Run Monte Carlo Distribution

When you run 1,000 Monte Carlo iterations on a 100-trade sample with a $10,000 starting balance, the engine outputs percentiles rather than single numbers:

Metric	Historical Backtest	Monte Carlo 50th % (Median)	Monte Carlo 95th % (Worst 5%)	Monte Carlo 99th % (Worst 1%)
Ending Equity	$17,500	$17,500	$14,200 (Bootstrap)	$11,800 (Bootstrap)
Max Drawdown ($)	-$840	-$1,350	-$2,480	-$3,150
Max Drawdown (%)	8.4%	12.8%	23.5%	29.8%
Longest Losing Streak	4 trades	6 trades	9 trades	12 trades
Longest Drawdown (Days)	18 days	34 days	68 days	94 days
Probability of 20% Drawdown	0.0% (Observed)	—	14.2% Chance	—
Probability of Ruin ($5k floor)	0.0% (Observed)	—	0.8% Chance	—

How to Interpret These Results

Your true drawdown expectation is the 95th percentile, not the historical backtest: Your chronological chart showed an 8.4% drawdown. But across 1,000 iterations, 5% of simulated traders experienced drawdowns of 23.5% or worse. If your trading account cannot survive a 24% drawdown without a margin call, your system is undercapitalized.
Prepare for losing streaks twice as long as your backtest: Your historical record showed at most 4 consecutive losses. Monte Carlo reveals that a 9-trade losing streak has a realistic 5% probability of occurring. If you risk 3% per trade, 9 consecutive losses equal a 27% haircut.
Probability of Ruin: If you define "ruin" as losing 50% of your account ($5,000 balance floor), the simulation calculates how many out of 1,000 curves touched that barrier at any point. In this test, 8 out of 1,000 runs breached the ruin threshold (0.8% probability).

6-Step Checklist for Running a Valid Monte Carlo Test

Follow these standards to ensure your Monte Carlo simulation produces actionable risk numbers:

Gather a minimum of 50 closed trades: Simulations run on 15 or 20 trades produce excessive standard errors. Aim for 50 to 200 trades across diverse market regimes.
Deduct realistic commissions and slippage first: Raw gross trade returns distort the simulation. Factor in exchange fees, broker commissions, and 1 to 2 ticks of slippage per fill.
Run at least 1,000 iterations: 100 runs do not populate extreme tail events. 1,000 to 5,000 iterations stabilize the 95th and 99th percentiles.
Define your hard ruin threshold: Set your ruin boundary based on your broker's intraday liquidation rule (e.g., account equity dropping below 50% or below exchange maintenance margin).
Evaluate both fixed-fractional and fixed-dollar sizing: Run simulations using fixed contract quantities (1 micro contract) and fixed risk percentages (1% of equity) to see how compounding affects drawdown depth.
Stress-test with 10% lower win rate: Run a bootstrap pass with winning trades reduced by 10% to simulate changing market regimes or edge decay.

How to Test Trade Sequences in ChartMini Replay

Before feeding trades into a Monte Carlo model, you need a sample of trades executed without look-ahead bias:

Open ChartMini Free Simulator: Select a futures or crypto instrument (such as MES or BTC/USD) and choose a randomized start date with future bars concealed.
Execute 50 consecutive setups: Apply your trading rules strictly bar-by-bar. Record your entry, stop loss, exit price, and dollar P&L for each trade.
Export your P&L vector: Take your 50 trade results (e.g., [+$120, -$50, +$80, -$60, ...]).
Run a permutation shuffle: Shuffle the array 1,000 times using a Python script or spreadsheet to plot your system's true drawdown distribution.
Adjust position size: If your 95th percentile drawdown exceeds your emotional tolerance, reduce your position size until the simulated risk fits your plan.

Frequently Asked Questions

What is a Monte Carlo backtest in trading?

A Monte Carlo backtest is a simulation technique that takes a set of historical or simulated trade results and reshuffles or resamples their order thousands of times. It tests how sensitive an equity curve is to trade sequence risk, revealing potential drawdowns and ruin probabilities that a single chronological backtest hides.

Why is a single backtest equity curve misleading?

A standard backtest represents only one specific historical path. If your 10 losing trades happened to be spread out evenly among 90 winners, your historical drawdown might look mild (e.g., 6%). But if those same 10 losses occurred consecutively due to random market clustering, your real-world drawdown could exceed 25%, triggering emotional panic or broker liquidation.

What is the difference between resampling with and without replacement?

Resampling without replacement (permutation/shuffling) takes your exact historical trade results and rearranges their sequence; total net profit remains identical, but the drawdown path changes. Resampling with replacement (bootstrapping) draws trades randomly from your distribution where individual trades can be picked multiple times or skipped, testing both sequence variance and sample variance.

What is the 95th percentile maximum drawdown?

In a 1,000-run Monte Carlo simulation, the 95th percentile maximum drawdown is the threshold that 95% of simulated equity curves stayed below (and only 5% exceeded). Quantitative traders use the 95th or 99th percentile drawdown rather than the historical maximum drawdown to calibrate emergency account reserves and margin limits.

How many trades do you need for a valid Monte Carlo simulation?

A statistically reliable Monte Carlo simulation requires at least 50 to 100 executed trades. Running simulations on fewer than 30 trades produces wide confidence intervals and high sample error, making risk metrics unreliable.

Monte Carlo Backtesting for Traders: How to Simulate Sequence Risk, Drawdown Distributions & Ruin Probability