Shortened test period after WFO NumWFOCycles=5

Posted By: SBGuy

Shortened test period after WFO NumWFOCycles=5 - 06/29/18 17:27

I have a script with the following test period.
StartDate = 20130101;
EndDate = 20171231;

Used NumWFOCycles = 5 to optimize two parameters, clicking on Test produced results only for the period 11/1/2015 - 12/31/2017.

Is this normal or am I doing something wrong? I have to comment out the NumWFOCycles line, Save, and Test to apply the parameters for the full test period of 1/1/2013 - 12/31/2017

Thanks
Posted By: Spirit

Re: Shortened test period after WFO NumWFOCycles=5 - 07/01/18 21:59

With no WFO you get a biased backtest, which is way worse than a shorter test period.

https://en.wikipedia.org/wiki/Walk_forward_optimization
Posted By: SBGuy

Re: Shortened test period after WFO NumWFOCycles=5 - 07/02/18 17:26

Thanks Spirit. I'm not sure what you mean by "no WFO". I did use WFO with NumWFOCycles = 5.
Posted By: SBGuy

Re: Shortened test period after WFO NumWFOCycles=5 - 07/02/18 18:10

My observation is that the final simulation run, as shown in result plot, looks like it is only for the last segment of the WFO, instead of the entire desired test period of:

StartDate = 20130101;
EndDate = 20171231;

If I want to to apply the optimized parameters, from the 5 WFO cycles, for the desired simulation period of StartDate thru EndDate, I have to comment out the NumWFOCycles line, Save, then click Test. The script then reads the parameter values from the .par file and applies it for StartDate thru EndDate.
© 2024 lite-C Forums