WFO

Hi all.
I'd like to train my H4 strategy fixing the following parameters.

1) length of test period in bars: 6*22*4 (4 months)
2) EndDate = 20181031
3) DataSplit = 90
4) LoockBack = 200

I've tried to play around with some parameters using the setWFO2 recomended function too. But I have found erratic results.
The impression was that Zorro did not calculated correctly the StartDate of the simulation.

Any hint? Thank you

How have you set StartDate?

No fixed StartDate leaved as default. The idea will be to vary the datasplit and the test period in order to check strategy robustness. I can set a start date way in the past like 20090101 if needed... on the contrary it shoud be calculated I guess....but why?

by default, startdate is current year minus numyears i think. If you need a certain date then you must set it.

yes, I guess so. The question is how to calculate it from the EndDate...

For calculating it from EndDate, you need to set NumYears. If nothing is set, a 6 years backtest is assumed.

Thank you JCL, can you articulate a little more?
How to keep the same EndDate, NumWFOCycles and the Length of Test period, while changing the dataSplit parameter (not in the same trainig, obviously)?

thank you

You can either set the length or the number of cycles, but obviously not both at the same time.

http://manual.zorro-project.com/numwfocycles.htm

Thank you I am spending some time on that manual page...
why not at the same time?
There is not the way to configure Zorro in order ti train 12 cycles of 16 weeks each with an assigned DataSplit percentage?

You can configure anything you want, as long as you know what you want. In your example, the simulation period is 12x training time plus 11x test time plus lookback.

jcl, thank you.
I know what I want: train a strategy for exactly 12 WFO (the numbers of core available for training in my PC) with a fixed EndDate and fixed DataSplit.
I am not confident how to code it and I was requesting some hints.

Anyway I understand that you can not provide snippets for everybody...
I'll try to code a pimped version of the setWFO2 reported in the manual end see what happens.

Ciao

There is no snippet because you normally don't change the start date. Otherwise your results can be all over the place. But if you want to do it, simply subtract the calculated simulation period from the end date. The result needs not be totally exact. Zorro will slightly adapt the cycle periods until it fits.