How to forward test in MT5! Do this if you want to know the truth about you trading strategy!

Main ideas, concepts, and lessons

• Forward testing is essential for real trading viability

  • Backtesting evaluates a strategy on historical (past) market data.
  • Optimization searches for the “best” EA parameters to produce strong backtest charts, but this often leads to curve-fitting (overfitting) that may not work in live trading.
  • Forward testing evaluates the EA on out-of-sample data (data not used in optimization), which is closer to what will happen on a real account.
  • The speaker argues that many traders conclude Forex is a scam because strategies can look profitable in backtests/optimization but fail or blow accounts when deployed live.

• Live-deployable EA development is hard

  • The speaker describes creating an EA that works live as extremely complex, requiring multiple testing methods beyond basic backtesting:
    • forward testing
    • walk-forward testing
    • Monte Carlo analysis
    • full portfolio testing
  • The speaker claims >99% of strategies fail, and that many traders don’t test with high-quality data.

• How to set up forward testing in MT5 (core methodology shown)

  • Use MT5’s built-in forward testing feature.
  • Configure:
    • how long to forward test
    • how much data is kept out of sample
  • Example approach described:
    • Optimize for 10 years total
    • Run the EA for the last 1 year as out-of-sample (not used during optimization)

• Why out-of-sample matters

  • Optimization may look excellent on backtest data but perform poorly out of sample.
  • The speaker warns that seller results and “beautiful curves” from optimization can be misleading compared to live outcomes.

• Testing choices and performance evaluation

  • Test on open prices to make testing faster (explicitly mentioned).
  • Use a fast genetic-based algorithm to avoid extremely long optimization times.
  • Use a reasonable number of optimization iterations rather than full optimization:
    • Example: target about “~1000 optimization” iterations instead of “full optimization”
    • Rationale: full optimization often increases curve-fitting; the “best” parameter region often appears in the early-to-mid portion (speaker claims roughly 1/3 to 1/2 of the full run).
  • The speaker also advises controlling what parameters are visible/considered (e.g., selecting specific “factors/indexes”) so the evaluation is focused.

• Interpreting the forward test vs backtest chart (important explanation)

  • The speaker addresses a misleading-looking MT5 graphic where equity may appear to rise and then “collapse.”
  • Clarification:
    • The backtest portion corresponds to the optimized period.
    • The forward test portion is separate (visually separated by a line).
  • Because the chart segments represent different evaluation periods, the graphic can look confusing/misleading.

• Demonstration of how forward test outcomes change parameter selection

  • The speaker shows that selecting parameters based only on a single metric (e.g., profit) can still yield negative forward performance.
  • They demonstrate filtering and comparing using multiple criteria:
    • profit
    • maximum drawdown
    • recovery factor
    • profit factor
    • Sharpe ratio
  • Recommendation: use a combined / custom criterion approach.
    • A “Complex Criterion” is presented as helpful because it combines multiple metrics into a more stable selection.
  • Main lesson: even if a parameter set looks great in backtest/optimization, forward results can be slightly positive, negative, or much worse, and one-dimensional filtering can fool traders.

• Claimed practical outcome

  • In the speaker’s demo case, the forward test outcome is described as not catastrophic for the chosen parameter set.
  • However, the lesson is that many other EAs would typically show large drawdowns and poor forward behavior.

• Call to action

  • If viewers want more detail, request additional content on:
    • forward testing
    • Monte Carlo analysis
    • other methods for ensuring live readiness
  • The speaker encourages visiting their website/blog for more content.

---

Detailed methodology / instructions (as presented)

• Define testing periods (out-of-sample split)

  • Optimize strategy parameters using a historical range (example: 10 years).
  • Reserve a portion strictly for forward testing (example: the last 1 year) so it remains out-of-sample.

• Run MT5 forward testing

  • In MT5, use the built-in forward testing function.
  • Set:
    • the forward-testing duration (how much data is out-of-sample)
    • that the forward test is not part of the optimization window

• Speed up testing (execution configuration)

  • Configure the test to use open prices only because it is faster.

• Choose an optimization strategy

  • Use a fast genetic-based algorithm to avoid very long runs.
  • Use fewer iterations for practicality (example target: ~1000 passes) instead of full optimization.
  • Avoid defaulting to full optimization, since it may worsen curve-fitting.
    • Tip from the speaker: better results often appear in roughly the first 1/3 to 1/2 of a full optimization run.

• Control optimization parameters visibility

  • Disable irrelevant parameters and focus only on the parameter groups you care about (the speaker references selecting “factors/indexes” and notes viewers should understand them).

• Evaluate results using multiple metrics

  • Compare both backtest and forward test using metrics such as:
    • profit factor
    • recovery factor
    • Sharpe ratio
    • drawdown
  • Do not rely only on profit filtering.

• Filter parameter sets carefully

  • Try multiple filters, since each can mislead:
    • highest profit → may still be negative in forward test
    • high recovery factor → may still not be ideal
    • high Sharpe ratio → not necessarily stable forward performance
    • drawdown filtering → can improve results but may still be weak
  • Use combined/custom criteria when possible.

• Use “Complex Criterion” as a combined metric (recommended in video)

  • The “Complex Criterion” combines metrics to create more stable and “balanced” parameter sets.
  • The speaker claims genetic optimization tends to find the parameter set with the best balance under that combined criterion.

• Interpret MT5 visuals correctly

  • Treat chart segments properly:
    • optimized/best-fitting region = backtest
    • segment after the separator line = forward test
  • Understand that some visuals may look like equity failure is on the “same line,” when the evaluation periods are actually split.

• Further required (mentioned) advanced tests

  • For robust EA verification beyond what’s shown, the speaker mentions:
    • walk-forward testing
    • Monte Carlo analysis
    • full portfolio testing

---

Speakers / sources featured

• Speaker/creator: Unnamed video author/speaker (referred to as “I” in the subtitles; no name provided).
• Software/source referenced: MetaTrader 5 (MT5) — especially the built-in forward testing feature and chart/criteria UI.
• Additional external references (mentioned, not shown as speakers):
  • “Meta’s developer team” (referenced regarding graphic layout/interpretation)
  • Mentions of EA types/names such as “Martingale” EAs / “Gil/Scers” EAs (no identifiable authors named)