Currently there is a NumSampleCycles which reshapes the price curve and allows to see how the strategy perform in different conditions ( then if the PF is similar between the cycles, then it still has an edge ).
The strategies are built using a WFO fix window. It means that eventhough this is OOS, in reallty the oos periods are completly biased. If the strategy could be tested over different wfo cyles, then we would really know how it really behaves. Then for exmaple, something like NumMaxWfo=10; and NumCyclesWfo = 5; would do 5 backtest of the strategy with wfo as 10,9,8,7,6 each.
Re: Votes for future features
[Re: nanotir]
#467773 08/29/1716:3308/29/1716:33
I believe you can do that already. Use NumTotalCycles, and change NumWFOCycles at any cycle. I haven't used this for WFO cycles yet, but in theory it should work.
Re: Votes for future features
[Re: jcl]
#467779 08/30/1720:1808/30/1720:18
+1 for python bridge because I am much more familiar with pandas, sklearn, etc. than I am with the equivalent methods in R.
I would also like advise(NEURAL) to accept way more than 20 signal values. I know that I risk overfitting when using huge numbers of features, but I do need to export the features in order to run feature selection algorithms on them.
+1 also for Notepad++ integration with auto-compilation.
Re: Votes for future features
[Re: johnnyp]
#467847 09/04/1713:4709/04/1713:47
Error 032 means that the number of WFO or sample cycles changed not from cycle to cycle, but inside a cycle. The number must remain constant after the INITRUN. If you can't fix it, post the code in another thread on the forum and I'll check it.
Re: Votes for future features
[Re: boatman]
#467868 09/05/1700:1209/05/1700:12
Minor: This setting reads like it can be automated. If two Zorro instances run strategies for the same API, then it could simply auto enable it, otherwise turn it off. Thus the user would not have to worry.
NOLOCK Don't synchronize multiple trading Zorro instances. Synchronization prevents simultaneous API access, but freezes all Zorros while a single Zorro waits for a response from the API. Set this flag for speeding up API access when the trading Zorros don't share the same API.
Re: Votes for future features
[Re: sqirtle]
#467898 09/06/1715:5509/06/1715:55
I think it would be nice with a mode flag that would enable brute force optimization, which is useful when data mining. I know it's possible to do this now as well, but itīs VERY easy to make a mistake when you have many parameters to optimize. Already with three parameters you get something like this
Code:
int ZPeriods[8] = {20, 50, 75, 100, 150, 200, 250, 500};
var EntryThresholds[5] = {0.75, 1, 1.25, 1.5, 2};
var ExitThresholds[4] = {0.5, 1, 1.5, 2};
int Index = optimize(1, 1, 8*5*4, 1)-1;
int Period = ZPeriods[Index%8];
var EntryThreshold = EntryThresholds[(Index/8)%5];
var ExitThreshold = EntryThreshold*(1 - ExitThresholds[(Index/(8*5))%4]);
Re: Votes for future features
[Re: Dalla]
#468216 09/25/1718:2209/25/1718:22