Optimisation passes inconsistent
Created at 28 Nov 2024, 11:46
Optimisation passes inconsistent
28 Nov 2024, 11:46
Issue: When optimizing the same EA with identical parameter ranges but different data types, we're getting inconsistent and unexpectedly low numbers of optimization passes.
Test Case Details:
- Parameters being optimized:
- Risk per trade: 0.5-1% (step 0.5)
- Fast MA Period Trending: 10-20 (step 1)
- Slow MA Period Trending: 30-50 (step 5)
- Fast MA Period Ranging: 8-15 (step 1)
- Slow MA Period Ranging: 18-25 (step 1)
- Take Profit ATR Multiplier: 2-5 (step 0.5)
- Stop Loss ATR Multiplier: 1-3 (step 0.5)
Results:
- With Real Tick Data: 1,525 passes
- With M1 Bars from Server: 1,300 passes
Expected Behavior:
- The theoretical number of combinations should be 24,640 passes
- The actual number of passes should be consistent between data types
- If there's parameter filtering logic, it should be documented and consistent
This suggests either a bug in the optimization engine or undocumented optimization behavior that needs clarification.
Question for cTrader team: What determines the actual number of optimization passes, and why does it differ between data types when using identical parameter ranges?
PanagiotisCharalampous
29 Nov 2024, 07:17
Hi there,
What optimization method do you use?
If you use genetic, then it is expected that the passes will not be consistent between optimization runs, even if everything remains the same.
Best regards,
Panagiotis
@PanagiotisCharalampous