Skip to main content

Best Practices And Tips

This page gives a practical workflow for testing BananaEA settings without turning the process into curve fitting.

The purpose is not to create a perfect backtest. The purpose is to decide whether a setup is worth testing further.

Testing Workflow​

Use this order:

  1. Define the setup you want to test.
  2. Run a baseline backtest.
  3. Review the main metrics and chart behavior.
  4. Optimize only if there is a clear reason.
  5. Validate selected settings on out-of-sample data.
  6. Run sensitivity checks.
  7. Forward test on demo.
  8. Start live only with small risk, if you choose to use it live.
  9. Monitor and keep records.

Before You Test​

Write down:

  • EA version
  • Symbol
  • Timeframe
  • Broker
  • Spread and commission assumptions
  • Preset or .set file
  • Date range
  • Initial deposit
  • Risk settings
  • What you are trying to learn

This prevents accidental cherry-picking after you see the result.

Baseline Backtest​

Run a baseline before optimizing.

Check:

  • Does the EA run without errors?
  • Does it trade the expected symbol and timeframe?
  • Are stop loss and take profit distances sensible?
  • Does the trade count look reasonable?
  • Are spread and commission assumptions realistic?

Do not optimize until the baseline setup is understood.

Optimization Rules​

Keep optimization small:

  • Optimize one to three meaningful settings.
  • Use practical ranges.
  • Avoid tiny step sizes.
  • Avoid optimizing cosmetic or notification settings.
  • Review the top group of results, not only the top result.
  • Look for stable parameter areas, not one perfect number.

If the result only works at one exact setting, treat it as suspicious.

Out-Of-Sample Validation​

Out-of-sample data must be untouched during optimization.

After selecting settings:

  1. Lock the settings.
  2. Test them on the later unseen period.
  3. Compare profit factor, drawdown, trade count, average trade, and consecutive losses.
  4. Reject the setup if it fails badly or behaves very differently.

Do not keep re-optimizing on the out-of-sample period until it looks good.

Forward Testing​

Forward testing means running the setup on a demo account or very small live size under current market conditions.

Check:

  • Broker execution
  • Spread behavior
  • Order rejections
  • Slippage
  • VPS stability
  • Whether trade frequency resembles the backtest
  • Whether drawdown stays within your limit

Forward testing takes time. A handful of trades is usually not enough to confirm a setup.

Live Deployment​

If you use a setup live:

  • Start with small risk.
  • Use one preset at a time until you understand behavior.
  • Monitor the platform.
  • Avoid changing settings after every loss.
  • Pause if lot size, drawdown, or execution looks wrong.

Live trading introduces real execution issues that Strategy Tester cannot fully reproduce.

Common Mistakes​

Avoid:

  • Testing only the best-looking historical period.
  • Ignoring losing periods.
  • Optimizing too many inputs.
  • Trusting a small trade sample.
  • Treating net profit as the only metric.
  • Ignoring spread, commission, swap, and slippage.
  • Changing broker or symbol without retesting.
  • Treating a backtest as a promise.

Testing Journal Template​

Test ID:
Date:
EA version:
Symbol:
Timeframe:
Broker:
Model:
Spread:
Commission:
Date range:
Preset:
Changed settings:

Net profit:
Profit factor:
Max drawdown:
Trade count:
Average trade:
Max consecutive losses:

Notes:
Decision:
Next action:

Monthly Review​

Review live or demo performance regularly.

Ask:

  • Is the EA behaving as expected?
  • Are spreads and execution similar to the test?
  • Has drawdown stayed inside your limit?
  • Are losses part of normal variance or caused by setup errors?
  • Has the broker changed symbol specifications?
  • Do you have enough trades to judge anything?

Do not re-optimize automatically every month. Re-optimize only when there is a clear reason and enough data.

When To Re-Optimize​

Reasonable reasons:

  • Major market regime change.
  • New production EA version with relevant setting changes.
  • Broker symbol specifications changed.
  • Long-term demo/live behavior no longer resembles testing.
  • You are building a new preset for a different symbol or timeframe.

Poor reasons:

  • One losing trade.
  • One losing day.
  • Boredom.
  • Wanting a smoother backtest.
  • Trying to force a setup to pass.

Final Checklist​

Before treating a setup as a candidate:

  • Baseline test completed.
  • Optimization used limited, meaningful inputs.
  • Out-of-sample test completed.
  • Parameter sensitivity checked.
  • Spread and commission assumptions reviewed.
  • Trade count is large enough to review.
  • Demo forward test is planned or completed.
  • Risk setting is small enough for live uncertainty.