Monte Carlo Simulation for Trading Strategies, in Plain English

A backtest gives you one story. Your strategy took these trades, in this order, and ended here. It feels like a fact.

But that order was partly luck. If the same trades had come in a different order, the ride could have felt very different. A losing streak that showed up in month three might have shown up on day one, right after you put money in.

A Monte Carlo simulation is a simple way to ask, "What else could have happened?" This post explains how it works, how to read the results, and where it stops being useful.

The idea in one paragraph

You take the list of trades from your backtest. Then you let a computer mix them up and replay them, again and again, maybe 1,000 or 10,000 times. Each replay is a new version of history that could have happened with the same trades. Instead of one result, you get a spread of results. That spread tells you more about risk than the single backtest ever could.

Three common ways to mix up the trades

1. Shuffle the order. Keep every trade, but put them in a random order. The final total stays the same, because it's the same trades. What changes is the path: how deep the worst drop gets and how long the worst losing streak lasts. This is the best way to answer, "How bad could the drawdown have been?"

2. Resample with replacement. Pick trades at random from the list, allowing repeats, until you have as many as the original. Some runs get extra winners, some get extra losers. Now the final total changes too. This shows how much of your result could just be a lucky draw.

3. Add real-world friction. Randomly skip some trades, as if the order never filled. Make fees and slippage a bit worse on some trades. This checks if the idea still holds up when things go a little wrong. We cover where that mess comes from in paper trading vs live trading.

A simple example (the numbers are made up)

Say a backtest has 120 trades. In the original order, the worst drop from a peak was 11%. The longest losing streak was 5 trades in a row. Looks fine.

Now you shuffle the order 1,000 times. Here's what you might see:

• In the middle run, the worst drop is about 13%.
• In the worst 5% of runs, the drop is 24% or deeper.
• The longest losing streak in the worst 5% of runs is 9 trades.

Then you resample with replacement 1,000 times and find that 820 of 1,000 runs net positive (simulated).

Again, these numbers are made up to show the idea. But the lesson is real. The backtest said 11%. The simulation says you should be ready for something closer to 24%. If a 24% drop would make you quit, or would break your risk limits, you've learned that before it cost you anything.

How to read the results

Look at the bad end, not the middle. The middle run is roughly what the backtest already told you. The useful part is the worst 5% or 10% of runs. Plan for those.

Check the losing streaks. Most people don't quit a strategy because of the average. They quit after nine losses in a row. If you know that streak is normal for this strategy, you're less likely to panic and turn it off at the worst time.

Watch how many runs end below zero. If a lot of resampled runs end net negative, the original result may have leaned on a few lucky trades. That's a warning sign. It often goes with overfitting.

Use it to size positions. If the bad-case drawdown is twice the backtest's, you may want smaller positions.

What Monte Carlo can't do

This part matters. A Monte Carlo simulation only mixes up the trades you give it. If those trades are wrong, it gives you a very detailed picture of something false.

It can't fix bad data. If the backtest used information the strategy couldn't have known at the time, every shuffle carries that error along. That's why we care so much about point-in-time truth.

It can't fix overfitting. If you tuned the rules until they fit the past, the trades themselves are too good. Shuffling them doesn't make them honest.

It assumes each trade is independent. Real markets have moods. Losses often bunch up in certain conditions. A plain shuffle breaks those clusters apart, so it can make things look smoother than they are.

It can't see the future. A new kind of market can bring trades unlike anything in your list.

So treat Monte Carlo as a stress test on a backtest, not a stamp of approval. It helps you understand risk. It doesn't tell you the idea works.

Where the real test happens

The only way to know if an idea survives is to watch it trade on data that didn't exist when you built it. That's forward testing. We explain why in backtest vs forward observation.

At Pickled Lizard, a strategy is first backtested on the full data set. If it passes, it moves to Paper, where it trades forward on paper with real fees and real timestamps. That's the real test. To be marked Verified, it has to keep up with its own backtest: at least 80% of the backtest's trade rate and 80% of its win rate. Until it has taken enough trades to judge, it shows as still collecting.

We keep every result, including the ones that failed. If a strategy can't hold up on Paper, the record says so. We're not claiming any strategy is ready for real money. We're building a system that makes ideas prove themselves first.

A quick checklist

• Did you shuffle the trade order to see the bad-case drawdown?
• Did you resample to see how much of the result could be luck?
• Did you add skipped trades and worse fees?
• Do you know the longest losing streak you should expect?
• Could you live with the worst 5% of runs?
• Is the next step a forward test on paper?

For more checks like these, see can you trust an AI trading bot's backtest? and walk-forward analysis, explained.

Try it free

The beta is free. Bring a strategy idea from your own AI, ChatGPT, Claude, Grok, or Cursor, and see how it does on paper. You can also see simulated paper results on our paper trading leaderboard.

Test it free at pickledlizardz.com/join

More on testing ideas: see all posts on Learn. Why I'm building this: read the story on our About page.

Paper trading only. Results are simulated. Not financial advice.