Base portfolio idea, bad?

Ok, I am sort of new here. Still running trial, but am definitely buying Strategyquant and the other apps.

 

I wanted to hear your opinion about this thing I am working on.

 

I am trying to find the best strategies for all major pairs on H1 timeperiod 1987-present (using asirikuy 1m data). Selection criteria is mainly ret/dd.

 

The idea is to put these strategies together in a robust base portfolio, even if they are all mostly using mean reversal they will hopefully not correlate too much.

 

Of course a strategy like thos will not give as good roi when going live as if you used a more resent timeframe for testing, but it might be more robust in the long run, and could serve as a solid base strategy.

 

I know its not an original idea, I am sure most of you have contemplated the same idea already.

 

I guess what I am really asking is this: Should I instead be focusing on more recent time, e.g. 2010-present?

I don’t do forex, and have less data, but I use the oldest data (in my case ~1997/9/2000 depending on the exact instrument – eminis are comparatively recent) to fairly recent (~2012/13). Then I use the more recent data (~2013) to fairly recent (~2015) for multiple OOS/robustness tests, with the most recent (2016) for walk forward.

 

In general, the longer the data you use to generate. For me, when I put a strategy on my platform, knowing it was generated over many years, then tested ok OSS for even more years up to present, gives me confidence (but they still fail during incubation much, much more than they go live).

 

For data, set yourself up a work flow that uses everything you have for either testing, robustness, WF, etc….use as much data as possible.

your criteria should be not so strict otherwise you will find just a very few strategies. i have only one straegy which has stagnation around 1 year for so long history. all other are in stagnation around 2000. for such a long history, stagnation of few years is acceptable.

you can use just newer data for generating and old data use to retest the robust strategies to see how it performed in past. 

 

of course that creating a robust diversified portfolio with the longest possible backtest is not any newie 😉

Yeah I noticed I have a hard time finding any worthy strategies at all on the 1987-present time periods other than on the EURUSD pair perhaps.

 

I can see several past topics on that issue.

 

So I will probably give up this idea, and focus building strategies from 2010-present instead for my all-majors portfolio.

That was my initial mistake as well, you can wait for days and there wouldn’t be any strategy going into your data bank.
So just relax your filter so as to fill your data bank, and then do retest with M1 or Tick and you might find unexpected differences in results.

I don’t do forex, and have less data, but I use the oldest data (in my case ~1997/9/2000 depending on the exact instrument – eminis are comparatively recent) to fairly recent (~2012/13). Then I use the more recent data (~2013) to fairly recent (~2015) for multiple OOS/robustness tests, with the most recent (2016) for walk forward.

 

In general, the longer the data you use to generate. For me, when I put a strategy on my platform, knowing it was generated over many years, then tested ok OSS for even more years up to present, gives me confidence (but they still fail during incubation much, much more than they go live).

 

For data, set yourself up a work flow that uses everything you have for either testing, robustness, WF, etc….use as much data as possible.

Hi Cujo, where do you get your emini data from?

 

Thanks.