How to check an expert advisor's backtest before you buy
A backtest can be honest and still mislead you. These seven questions help you read one, whoever the vendor is. For each one we also say where to find the answer for our own program.
1. Are commissions and spreads included?
The Strategy Tester runs without commission unless someone sets it. A program that trades often with a small average profit per trade can look good without costs and lose money with them. Ask what cost per trade was used, and where that number comes from.
Our answer: $2.00 per round turn on MNQ, measured on real account statements, plus a scenario with slippage. See how the backtest was made.
2. Which tick model was used?
MetaTrader 5 can test on generated ticks or on real ticks. For a program with tight stops and targets, generated ticks can change which level is hit first. "Every tick based on real ticks" is the closest to reality the tester offers, and the report header shows which model was used.
Our answer: real ticks on every contract.
3. Was the program developed on the same data it is tested on?
If the rules were tuned on the test period, the backtest is in-sample: it shows how well the rules fit the past, not how they will do next. Ask whether any period was kept aside, and whether there is a forward test on data the program has never seen.
Our answer: our backtest is in-sample, and we say so. The forward test started on 16 October 2026.
4. How deep was the drawdown, and how long did it last?
Net profit means little without the maximum drawdown next to it. Check whether the drawdown is measured on closed trades or on open equity (open equity is deeper), and how many months it took to recover. Then ask yourself whether you would have kept the program running through it.
Our answer: on the results page, with the drawdown chart under the profit curve.
5. How long were the losing streaks?
A program that wins 40% of its trades will have long losing streaks; that is arithmetic, not a defect. A program that wins 90% of its trades may hide rare, very large losses. Look for the longest losing streak, the largest daily loss and the largest single loss, not only the win rate.
6. Does a small part of the profit carry the result?
Remove the best day or the best trade and look at what remains. If one event makes a large share of the profit, the result depends on that event repeating. Also check every year separately: a program that made all its money in one year may have found a market phase, not an edge.
Our answer: one day produced 17% of our backtest profit, and we show the results per year and per strategy.
7. Are simulated and real results kept apart?
A chart that joins a backtest to a live account makes the live part look longer and smoother than it is. Ask which part is simulated, which part is a demo account, and which part is a real account. A live record should be monitored by a third party, such as an MQL5 signal, not shown as screenshots.
Our answer: simulated results are marked in orange, the forward test in green, and they are never on the same chart.
One more step: test it yourself
On the MQL5 Market every expert advisor has a free demo that runs in your own Strategy Tester. Run it on your broker's data, with your broker's commission, on a period the vendor does not show. If the result is very different, ask the vendor why.
This guide is general information, not investment advice. Trading leveraged products carries a high risk of loss. See our risk disclosure.