How to backtest a trading strategy

Backtesting means testing a strategy's rules on past prices to see how it would have done. Here is how to do it properly, by hand or automatically, with real numbers from 694 backtests we ran ourselves.

What backtesting is

Backtesting is testing a trading strategy on past prices. Here is one real day of it.

Swipe the chart to see the rest of the day

1. 1 The rule marks the range High and low from 09:30 to 09:45
2. 2 A close above it is the entry 09:50, stop under the range, target 2R
3. 3 Out at 16:00 for +0.85R Neither the stop nor the target was hit

A backtest does this for every day in the test: 1,490 trades from 2020 to 2025, 46.4% won, +157.5% after costs.

Manual and automated backtesting

Two ways to run the same test. Many traders use both.

Manual backtesting

How it works

You replay the chart bar by bar with the future hidden, and take each trade by hand

What you need

Written rules and a replay tool. No code

Good for

Price action and discretionary setups, practicing execution, prop firm practice

Weak spot

Slow, so samples stay small, and it is easy to peek ahead without noticing

Automated backtesting

How it works

A program applies the rules to years of data in seconds or minutes

What you need

Code (Python, Pine Script, MQL or C#), or an AI that writes it from your rules

Good for

Rules that can be written exactly, large samples, many markets at once

Weak spot

Easy to over-fit, and it tests what the code says, not how you would trade it

How to backtest a trading strategy, step by step

One real example runs through every step: the 15-minute opening range breakout on MNQ, tested on 5-minute charts from 2020 to 2025.

1. Write the rules down

   Entry, exit, stop-loss, take-profit, position size, and when not to trade. If two people can apply your rules to the same chart and get different trades, they are not rules yet.

   Example rules 15-minute opening range breakout

   Range

   High and low from 09:30 to 09:45 New York time

   Entry

   A candle closes above the high (buy) or below the low (sell)

   Stop

   The other side of the range

   Target

   2 times the risk

   Exit

   16:00 at the latest

2. Pick the market, the timeframe and enough history

   Use several years that include trends, ranges, quiet months and wild ones. Aim for at least 100 trades: with fewer, luck can explain almost any result. We do not judge a test with fewer than 100 trades.

   What we tested it on

   Market

   MNQ (Micro Nasdaq 100)

   Chart

   5-minute

   Years

   2020 to 2025

   Costs

   Charged on every trade

   100 trades: the minimum 1,490 trades

3. Keep the last years unseen

   Build and adjust the strategy on one period, then run it once, unchanged, on later years you never looked at. A strategy that only works in the years it was built on was fitted to them.

   Build, then check on years it never saw Build 2020 to 2023 Unseen 2024 to 2025 +65.4% profit factor 1.13 · 994 trades +55.7% profit factor 1.20 · 496 trades

   Held up: it made money in the two years it was never tuned on.

4. Run the test

   By hand: start the replay before your first date so the future stays hidden, and log every trade. Automated: code the rules or describe them to an AI, then check a handful of its trades on the chart to be sure it does what you meant.

   A manual replay

   MNQ, 5-minute, February 4, 2025: price broke above the opening range at 09:50. Now decide, and take every trade the rules give you, including the ones you do not like.

5. Charge the costs

   Spread, commission and slippage come off every trade. On short timeframes they decide whether a strategy makes money at all.

   What costs took Before costs $89,372 After costs $78,771

   Costs: $10,601, 12% of the profit, over 1,490 trades.

6. Read the results, not just the profit

   Look at the profit after costs, the profit factor (gross profit divided by gross loss: under 1 loses money), the win rate, the maximum drawdown and the number of trades, then compare it with simply holding the market. Our example won fewer than half its trades and still more than doubled the account, yet holding MNQ made more.

   The result, after costs

   Return

   +157.5%

   Win rate

   46.4%

   Profit factor

   1.16

   Max drawdown

   15.6%

   Buy and hold

   +190%

7. Stress it, then trade it small

   Shuffle the order of the trades thousands of times (a Monte Carlo test) to see how deep the drawdown could get, try the rules on other markets, then trade them on a demo or a small size before real money.

   How deep it could fall In the backtest 15.6% Monte Carlo, worst 5% 30.4%

   Shuffling the same trades almost doubled the worst drawdown. Size your risk for that, not for the backtest.

What 694 of our own backtests show

We test popular strategies on 1-minute history, after costs, and publish every result. Most of them fail.

• Lost money 460 66%
• Too few trades to judge 48 7%
• Made money overall, lost in one period 93 13%
• Made money in both periods 64 9%
• Passed every check 29 4%

Famous is not the same as profitable

The stochastic crossover (14, 3, 3) lost money in 33 of its 36 tests.

A profit over all the years is not enough

93 tests made money overall but lost in the build years or in the unseen ones. One test over every year would have called them winners.

The win rate says little on its own

The opening range breakout won fewer than half its trades and made money in both periods.

See all 31 tested strategies

Common backtesting mistakes

Looking ahead

Seeing the next candles before you decide. A replay that hides the future is the fix.

Over-fitting

Tuning settings until the past looks perfect. Keep the last years unseen and test them once.

Too few trades

Ten great trades prove nothing. Wait for at least 100.

No costs

A strategy that wins small and often can lose everything to the spread.

Cherry-picking

Reporting the one market or timeframe that worked. Keep every test, losers included.

Skipping trades

Taking only the setups you like in a manual test. The rules decide, not the mood.

How to backtest in BacktestingMax

Both ways in one app, in your browser.

Replay the chart with the future hidden

Pick a market and a start date. The chart opens on the bars before it, and nothing after it shows until you press Play. Trade the replay with stops, targets and costs, and every trade lands in your stats.

• Play, pause or step one bar at a time
• Down to 1-second candles on Pro
• Prop firm rules if you are practicing a challenge

Or describe the rules to the AI

Write the strategy in plain English. The AI turns it into code, runs it on years of 1-minute history and returns the profit factor, drawdown, equity curve and every trade.

• No code to write
• Change one rule and run it again
• Monte Carlo and optimization built in

Backtest by market

What you can test on each market, and how far back the data goes.

Practicing for a prop firm? Try the prop firm simulator, free

Backtesting questions

Choosing a tool? Compare 10 backtesting tools

Backtest your first strategy for free

Pick a market and a date and trade it bar by bar, or describe the rules and let the AI test them. No credit card, nothing to install.

Every number comes from our own published backtests on 1-minute data, after costs, with 1% of the account risked per trade. Past results do not guarantee future results.