Results depend on the chosen symbol, period, price data, modelling method, trading costs, execution assumptions and EA settings. Compare tests using the same documented assumptions; a favourable result alone does not prove the setup is realistic.

Backtesting evaluates selected settings. Optimisation tries multiple settings to find better historical results. Selecting the best historical combination can overfit the past.

For example, a test that starts with 10,000 and ends with 12,000 shows a net gain in that simulation. It does not show how much the account fell along the way, how much capital was exposed or whether the result depended on only a few trades. Compare the equity curve, drawdown, trade count and distribution of results. These numbers are examples, not a return target or a recommended strategy.

A balance curve mainly reflects completed trades; an equity curve also reflects unrealised gains and losses. A smooth balance curve can therefore conceal substantial open-position losses. Check which curve and which drawdown measure the report uses before comparing two reports.

Confirm the exact symbol, date range, data coverage, modelling method and cost assumptions. Spreads, commissions, swaps and execution delays can change results. Do not assume that every test includes all costs or reproduces current live execution. Missing data or a different price source can also change the sequence of trades.

Keep MT4 and MT5 tests separate. An MT4 EA and an MT5 EA are not automatically the same program; the platform, EA version and settings must match the report being reviewed. If a test produces no trades, inspect the test log, available data and EA inputs before concluding that the EA is faulty or that its strategy never trades.

If you repeatedly change parameters until one historical period looks good, that period is no longer an independent check. Reserve a different period before selecting parameters, and evaluate the selected settings there without tuning them to that period. Repeatedly choosing the best result can create an attractive historical report that does not generalise.

Keep a reproducible record: platform and EA version, symbol, period, initial test balance, parameter set, modelling method, costs and the report. Compare changes one at a time. A backtest, an optimisation result and subsequent demo operation answer different questions; none alone establishes that an EA is suitable for your circumstances or will make money.

Review drawdown and the number of trades as well as profit. Keep a separate period that was not used to choose the settings, and compare performance there. Demo testing adds operating experience but still does not guarantee live results.