How to read a gold EA backtest without being misled

A backtest runs a robot on historical prices to show how its rules would have traded. It is useful, but it is a simulation. A single headline number, such as a monthly return, says very little without the context behind it.

Return means nothing without drawdown

Always read the return next to the maximum drawdown: the largest drop from a peak before a new high. A 10% return with a 9% drawdown is a very different risk from the same return with a 40% drawdown. Ask whether the drawdown is measured on balance (closed trades only) or equity (including open trades). For grid robots, equity drawdown is the number that matters, because losses can sit in open positions for a long time.

Check the test period

One month is a small sample. Gold has calm months and violent ones. Ask for results across several years, including strong trend periods and major news events, and look at the worst month as well as the average.

Check the modelling and the costs

• Price data: tests based on real tick data are more realistic than tests built from one-minute bars.
• Spread: was a fixed, unrealistically low spread used, or a variable one similar to your broker?
• Commission and swap: were they included? Grid positions held overnight pay swap.
• Slippage and execution: live fills on fast gold moves are often worse than the test.
• Contract size: does the test use the same lot and account type you plan to use?

Watch for over-fitting

If settings were tuned until the past looked perfect, the robot may simply have memorised history. A healthier sign is when results stay reasonable on data the settings were not tuned on, and when small changes to the settings do not flip the result from great to terrible.

Compare with live behaviour

The best complement to a backtest is a live account you can observe. Read-only investor access lets you compare what the backtest promised with how the robot behaves with real spreads and real execution.

Educational content, not financial advice. Leveraged trading can cause losses larger than expected, including the loss of your whole trading balance.