1. Match the intended market and timeframe
Select the symbol and timeframe the EA was designed for. A result on a different market or timeframe may say little about the production setup.
2. Use the strongest price model available
For short-timeframe systems, detailed tick data is generally more informative than coarse bar-only modeling because entries, stops and spreads can be sensitive to intrabar movement.
3. Make trading costs realistic
Spread, commission and swap can materially change an automated strategy's result. If the backtest assumes unrealistically cheap execution, the result can overstate what the strategy could have captured.
4. Record the exact settings
Save the EA inputs, test period, initial balance, leverage, modeling mode and broker data source. Without reproducible settings, a backtest is difficult to audit or compare.
5. Read more than net profit
Review trade count, profit factor, win rate, average win/loss, maximum drawdown and the distribution of results. A strong headline return paired with extreme drawdown can represent a very different risk profile from a lower-return test with stable exposure.
6. Avoid tuning to every historical wiggle
Repeatedly changing parameters until one period looks perfect can create an overfit strategy. Use parameters that have an economic or behavioral rationale and look for stability rather than one isolated optimum.
7. Move to demo validation
A historical test cannot reproduce every live condition. After backtesting, use a demo account to check broker symbols, spreads, execution, news feed behavior, licensing and notifications before considering live use.
Smooth Operator FX reference test
The latest supplied reference tests use EUR/USD M1 in MT5 Strategy Tester from January 1, 2025 through September 22, 2026 with 100% reported history quality. The headline $200,000 runs produced 1,073 Prop-Safe trades and 1,076 Compound trades; separate $10,000 runs from the same dates are also documented on the backtest page. See the full backtest page for the documented metrics and disclosure.