Hack – delete many strategy files easily that don’t meet complex criteria

I produced over 5,000 strategies.  Then I did a Retest.  I exported the to CSV file and then did analysis on the numbers.  I was left with a list of 500 files to be deleted out of the 5,000 because they did not meet my criteria.  I went down the list in SQ, found the strategy I wanted removed, checked the box next to the strategy, and then clicked delete.

 

After spending hours doing this, I decided there must be a better way.  I found the way, typed up the process in the attached document so you can benefit as well.  As I am sure many people come across this same problem.

 

Note:  

This does not apply to the simple criteria.  For example, you want to delete all strategies that have a profit factor of 1 and below.  That can be done by sorting the stragties by that criteria, then easily selected all the ones at this number and below.  

 

This method is for the more advanced analysis.  For example, you want to remove all the strategies where the Out of Sample Ret/DD is less than 50% lower than the In Sample Ret/DD.  This criteria requires you to do formulas in excel to identify all those strategies that need to be removed.  In cases like this, you would follow the attached process to easily remove them.

 

Wish you the best of luck.

 

-Stearno

 

Updated: I attached the document.  I realized it did not let me post a word document.  So I put it into pdf and now it is attached.

That sounds good Stearno, but I don´t see a attached file with the description.

I’ve been pushing filters to more and more extremes as time goes by, and have found weighted fitness styles that really meet my standards depending on what I want the generation to accomplish. With those at hand, I set the databank to hold only 10 strategies max, and only the highest fitness remains after ~ 10million random gen. That’s atleast my own style of going about finding ways of saving time.

Dito here, weighted fitness formula, only keep 30 best results just in case, but usually place #1 of the weighted fitness formula is exactly the one I´d chose manually out of all of them too. Took months though until the weighted fitness formula was exactly like I wanted it and I hope we can easily port our weighted fitness formula from SQ3 to SQ4 🙂

My situation that this solution was developed for was not for the strategy generation.  It is applicable to me during the strategy retest, which the weighted fitness formula, the evolution filters do not apply to the filters I am wanting to do on Retests.

 

An example of what a criteria that applies to this situation is on roubust test results.  Someone wants to remove all strategies where the 95% RST result’s profit factor is less than 50% of the actual strategy’s profit factor.  So I export the test results into excel, run a formula comparing the two, and now I want to delete those strategies in SQ.  Using this method, now that can be quickly/easily done.

 

-Stearno

I’ve been pushing filters to more and more extremes as time goes by, and have found weighted fitness styles that really meet my standards depending on what I want the generation to accomplish. With those at hand, I set the databank to hold only 10 strategies max, and only the highest fitness remains after ~ 10million random gen. That’s atleast my own style of going about finding ways of saving time.

 

That is good.  Yes, this works well with the standard filters. Be interested to learn how you correlate your objectives for the strategy with the filter and levels you chose in your custom fitness calculation.

 

-Stearno

My situation that this solution was developed for was not for the strategy generation.  It is applicable to me during the strategy retest, which the weighted fitness formula, the evolution filters do not apply to Retests.

 

Weighted Fitness applies to Retests as well. Just sort by it just like in the building section, exactly the same there, using it all the time on the retests as well.

Weighted Fitness applies to Retests as well. Just sort by it just like in the building section, exactly the same there, using it all the time on the retests as well.

 

Cool. I stand corrected.  I changed what I said in the above post to what I actually meant and added a further example.  What I am talking about is formula comparisons of one field to another, not one field against a specified level (which is weighted fitness).

 

-Stearno