Hi

I asked this already but I am not quite sure if I am doing it right and I think it is not discussed in the tutorial.
The problem is how to train some algos and some assets to develop a portfolio strategy.
Lets say there are 5 algos to be trained with 10 assets and ZorroS.
What is the right way to train them?
Way1:
All together like workshop6. Then the question is which hedge to use and how it influences it.
Way2:
Do 5 scripts, one for each algo and use the 10 assets in each script
Way3:
Do 10 scripts, one for each asset and use the 5 algos in each script. I guess hedge plays a role here.
Way4:
Do 50 scripts, one for each algo asset.

In way1 everything is already together. In the others, after the train is done, the par,fac files have to be combined in on file and the script has to contain all algos and assets. So then doubts emerge since algos are now affecting each other and asset/algo performance can change.

Certainly the most efficient way is to train them all together. Set Hedge = 1 or 2 (depending on what you want your system to do).

Next, there are many ways you can optimize the portfolio component selection process. If you are comfortable exporting and importing balance curves for each component, you can read them into another script and perform your optimization there. On that basis, you could then set up your trading script. You could also get really clever and perform your portfolio optimization in real time in the one script (I haven't done this myself, but I do believe it is possible).

If you don't want to bother with importing and exporting balance curves for each component, another way to do your analysis and optimization is to set the LOGFILE flag during the backtest. Then, open the testtrades.csv file in a spreadsheet program and create a pivot chart from the data. You can then sort the pivot chart to show the daily profit and loss of each portfolio component separately. You now have all the data you need to perform your portfolio optimization.

If you use the software, you can perform the above in R where you have access to all sorts of powerful optimization and econometric tools that would be tricky to implement in Excel.

Does that help?

Hi

Yes it does. Knowing that train all together is the best way is actually really helpful. I was not completly certain about it before. But all agree on that? I mean, there is a lot of information about trade signals and indicators out there but when it comes about portfolio there is not that much.

Thanks for the info about R and econometrics. I think it is highly recomended to jump into it to create a pro strategy.