a) I think that this behaviour arises mainly when the time frame is also optimized
b) In case the time frame is optimized, there is a 2nd bug which is: The optimization never reaches the 2nd generation, it always stops at the end of the 1st generation..
To show you proof of this, I provide two samples with pictures.
Optimizations performed on " Sample MA crossover"
example #1
=========
Fast 5;200;1
Slow 5;200;1
Data Series (minute) 15;60;15
Jan 1 2006 to Dec 31 2016
CL 09-16 (Merge Back adjusted)
Data source: NinjaTrader continuum demo (CQG)
Optimizer: Genetic (generation size 400), the rest parameters as stock.
Optimize on: Max Net Profit
Keep best # results: 50
Mismatched Result:
Parameters: Period:60m Fast:15 Slow:23
Results: Total Net Profit: 225330$.
BUT Trades Analysis & Graph show Cum.net profit: 6090$
example #2
=========
Fast 5;200;1
Slow 5;200;1
Data Series (minute) 15;60;15
Jan 1 2006 to Dec 31 2016
CL 09-16 (Merge Back adjusted)
Data source: NinjaTrader continuum demo (CQG)
Optimizer: Genetic (stock params).
Optimize on: Max Net Profit
Keep best # results: 50
Mismatched Result:
Parameters: Period:15m Fast:35 Slow:38
Results: Total Net Profit: 122340$.
BUT Trades Analysis & Graph show Cum.net profit: -19720$
Best regards.
Comment