MetaTrader 5 ships with a genuinely capable backtesting engine — the Strategy Tester — that can replay years of historical price data through an EA and report every trade it would have taken. Used honestly, it is the most important tool you have. Used carelessly, it produces beautiful equity curves for strategies that will lose money from the day you turn them on.

What a backtest can tell you

A good backtest establishes a strategy's statistical shape: how often it trades, its win rate, its average win versus average loss, its worst historical drawdown, and how sensitive it is to costs. This shape is the baseline you compare live behavior against. When a live strategy starts behaving unlike its backtest — more losses, deeper drawdown, different trade frequency — that divergence is your early-warning system.

What a backtest cannot tell you is what returns you will get. The future draws from a different distribution than any historical sample. Everyone knows this sentence; almost no one prices it into their expectations.

The four classic self-deceptions

  • Overfitting: tuning inputs until the backtest is perfect. With enough parameters, you can fit any historical noise. The result memorizes the past instead of learning structure. Antidote: test the tuned strategy on data it never saw during tuning.
  • Cherry-picked windows: testing 2023–2026 because the strategy "was designed for current conditions." Include ugly periods — 2020's volatility, flat grinding ranges — or the test is an advertisement.
  • Ignored costs: spread, commission, and slippage turn many profitable backtests into losers. Model realistic spreads for YOUR broker, not the tester's optimistic defaults.
  • Modeling-quality blindness: bar-level simulation guesses what happened inside each candle. For strategies with tight stops, use real-tick data and treat anything else as an approximation.
Rule of thumbIf small changes to an input produce large changes in backtest profit, the strategy is fragile. Robust strategies degrade gracefully as parameters move.

A sane testing sequence

Backtest across at least several years including hostile conditions, with realistic costs. Then forward-test on demo for weeks: the strategy runs on live prices with no history to peek at. Compare the demo results against the backtest's statistical shape. Only when live behavior matches tested behavior does real capital — small, at first — enter the picture.

This sequence is slow by design. Every shortcut through it is a transfer of risk from the past, where it is imaginary, to your account, where it is not.