I'm trying to implement in Zorro this R code provided in this site: IGM strategy

The code looks like this:



The TEST phase works correctly, but the TRAIN phase optimize the last asset only, then the TEST issue an error that no .par file is available for each asset.
The manual explains that the TRAIN runs each asset with all available parameters value, then continue to the next asset.
In this specific strategy the position is taken only after all assets are ranked by their returns, in my understanding the current TEST approach can't handle such situation.
Am I correct or there is some hack to overcome this situation?

Is there a specific reason in this script to not use the assetloop?

I'm using while(asset(loop... optimize( ...
If I use asset - loop to calculate the returns, I got the same results, no optimization is made.

The manual suggests using an while (asset(loop()) - while (algo(loop())) structure for optimizing params per component. Try adding the "algo" loop.

Hi Zheka, I don't understand how to add a while(algo(loop when there is only one algorithm in this strategy.

I'm not sure that Zorro can optimize a rotation strategy. It seems that Zorro's optimization algorithm optimizes each asset separately.

Well, Zorro can optimize a rotation strategy, but then if you don't want to optimize every asset seperate, you'll need to move out your optimize calls from the assetcalls.
You need to manually define the optimize steps and values to optimize for.

Also looking at the newest beta, offers you plenty of options to optimize, at least latest with the version >=2.33 you can optimize everything.

Thanks danatrader.
I do not agree with you, I think Zorro (at least in the free version), uses in its optimization algorithm a "vertical" approach.
When optimizing a portfolio of assets, Zorro looks at each asset separately, and no asset influences the optimization of other one.
In a rotation strategy one might optimize in a sense that there are influences in the optimization process between the asset ("horizontal" optimization).
In a horizontal optimization one would end with different parameter values for each asset.

If I would proceed with the R code provided in the attached link, and add an optimization code, I could easily get a different "lookback" period for each of the 6 assets candidates.
I'm not saying that optimizing each asset look-back would be the best approach for this strategy, but at least it's dame a trail.

I am sorry, may you explain in different words?

You say, "It seems that Zorro's optimization algorithm optimizes each asset separately."

Then you state, "In a horizontal optimization one would end with different parameter values for each asset."

Isn't that contradicting?
Then again, with beta 2.33 you can use R-Code for optimization.

danatrader thanks for your feedback.

Yes it's contradicting. Zorro in my poor opinion optimizes in a "vertical" sense. And I'm looking to optimize in a "horizontal" sense.
I'll look into beta 2.33 to see if it can help me.

You maybe want to create a synthetic asset, optimize it and then select based on the optimize timeframe of the synthetic asset choose the best performing, the one with the most weight within the synthetic.