The best forward testing technique that really works! Advices from professional forex EA developer.

Tech/EA Forward-Testing Concepts & Key Takeaways

• Forward testing is presented as the most important step in developing a Forex EA/trading system—more realistic than backtesting.
• Common misconception addressed: People think forward testing means doing one optimization and one forward test (e.g., optimize over 5 years, then forward test only 1–1.5 years). The speaker argues this is not enough to know what will actually hold up in live trading.

Why single split tests fail

When you optimize on a historical window, you can “pick” parameters that look good in that forward segment. But in live trading, you can’t know which optimized parameter set will remain robust.

Recommended robustness approach

To build evidence that the EA generalizes beyond one cherry-picked chart period:

• Perform many cycles of optimization + forward testing (at least ~50).
• Use multiple time splits (example described: optimize 2000–2005, forward test on new data for ~1 year, then repeat for many years/periods).
• Run tests across multiple currency pairs (example mentioned: at least ~10 pairs).

Scale of effort

The speaker claims that trying to find a consistently working EA with only one optimization/forward run (even with an EA developer) is unlikely—robustness requires broader stress/validation.

Method evolution / EA architecture implications

• The speaker says previous EAs failed repeatedly under this stricter method.
• To fix this, they moved toward price-action–based EAs, building new/less predictable combinations from existing indicator/price-action concepts rather than relying on one known pattern.

Handling market regimes (especially ranging markets)

• Simple indicator crossing systems (e.g., a few indicators + RSI) are described as insufficient for range-bound conditions.
• The proposed solution is a more complex system including:
  • 5+ indicators (general guidance)
  • plus filters to detect ranging markets
• The current approach mentions using moving averages, complex price action patterns, and time filters.

Additional robustness features mentioned

• Dynamic stop losses
• Time filters
• Testing/stress testing software tools (implied as part of the workflow)

Critique of marketplace EA sellers

• Many sellers (especially on MQL5) supposedly optimize EAs to produce “beautiful charts” for specific periods, while real-market results don’t generalize.
• Some sellers may not fully understand forward testing, leading to products that do well in backtests but fail live.

Backtest vs real performance emphasis

• Backtesting is framed as not showing reality.
• Forward testing is framed as the “reality” check.
• The speaker notes that some EAs can look great in backtests yet be less reliable in live trading (and vice versa, depending on strategy design).

Practical buyer guidance

• Buyers should try live trading / demo vs real live accounts, not judge solely by backtest charts.
• The speaker mentions offering “rental versions” (trial versions) so users can test on live accounts before buying.

Main Speakers / Sources

• Main source: A single EA developer/trader speaking directly (no other named co-speakers mentioned).
• Referenced communities/platforms:
  • MQL5 EA market / sellers
  • Forex EA developer community (no specific individual names cited)