How to Backtest a Trading Strategy in MetaTrader

Open the Strategy Tester from the View menu or with Ctrl+R. Select the expert advisor you want to test, the instrument, the timeframe, and the date range for the historical period. Choose a model — Every tick based on real ticks is the most accurate but requires downloadable tick data and is slower to run; Open prices only is faster but far less precise about intrabar behaviour.

Set the starting deposit, leverage and spread to match a realistic account rather than default values, then run the test. Results appear as a summary of net profit, drawdown, win rate and profit factor, alongside a full trade-by-trade report and an equity curve chart showing how the account balance moved over the tested period.

A backtest is only as reliable as the data behind it. Many brokers' free historical data has gaps, incorrect spreads, or missing tick-level detail, particularly on less common instruments or during volatile periods, and a strategy tested on poor data can show results that would never occur live. MT5 allows you to download higher-quality tick data through its history centre, and it is worth checking the quality report the tester produces, which shows what percentage of ticks used were modelled versus real.

Modelling assumptions also matter beyond raw price data. Spread is often held fixed in a backtest even though real spread widens during news and thin liquidity; slippage is frequently not modelled at all unless you specifically configure it; and swap costs on positions held overnight need to be included for any strategy that holds trades longer than a day. A backtest that ignores all three will systematically overstate real-world results.

Curve fitting happens when a strategy's parameters are adjusted repeatedly against the same historical data until the results look strong, effectively tailoring the rules to that specific period's noise rather than to a genuine, repeatable edge. A telltale sign is a strategy with many finely tuned inputs — an unusual stop distance, an oddly specific moving average length — that performs exceptionally on the tested period and considerably worse the moment the data changes even slightly.

A useful check is to split your data: optimise parameters on one portion of history, then run the same unchanged settings on a separate, later portion the optimisation never saw. A strategy that holds up reasonably well out of sample has a better chance of reflecting something real; one that collapses has likely been fitted to noise rather than to a genuine pattern.

A backtest, however carefully built, cannot capture everything about live trading — execution quirks, your own behaviour under real risk, and conditions that have not occurred in the tested history. Forward testing on a demo account, watching the strategy trade in real time without financial risk, is the next step, and it should run long enough to see the strategy behave through more than one type of market condition, not just a single favourable week.

Once a strategy has held up in forward testing, moving to a lower-cost evaluation account before committing to a larger funded programme is a sensible way to confirm it holds up under real drawdown pressure and genuine psychological stakes, which no backtest or demo account fully replicates. Treating the evaluation stage itself as part of the testing process, rather than skipping straight to maximum size, tends to produce steadier long-term results.