Speeding up the Portfolio Optimization

Speeding up the Portfolio Optimization No. 1

Please allow the Equity Curve with less points, so a detailed equity curve has less data points in portfolio optimization. That means that the user can define how much he narrow the data points from detailed to less detailed equity curve. Of cource the maximum intraday drawdown should be shown nevertheless. But with less data points an optimization on portfolio level is much faster. E.g. on 1 minute trading systems with many of trades it can be useful that only every 5th trade is shown as data point e.g. or that only one point or one bar per day (or week for daily systems) is shown as equity curve. After downsizing the data points on an equity curve by the user he then can do much faster an portfolio optimization because less data points per trading systems have to be considered in calculation. This is neccessary because with lots of single trading systems, e.g. 50 or 100, you need to speed up the optimization, because it still takes too long with a genetic approach, I have been experiencing. Thank you.

 

Speeding up the Portfolio Optimization No. 2

Please allow to set a multiple portfolio optimization before a first portfolio optimization starts. E.g. you want to test 100 strategies. You cannot do this at once because it takes too long. Then you divide them into 5 times 20 strategies you can test in one portfolio optimization. After such an optimization the user defined criterias choosing the best portfolios ( 1 or an X number of his portfolios from the optimization table), should be saved into chosen portfolios. After the 5 (with 20 single systems each) portfolio optimization then all the chosen portfolios comes together in an overall portfolio optimization. In that way a large number of single trading systems can be handled for a portfolio optimization. Of course the user should define how he wants to define the several parts of a bigger portfolio optimization. This would save time. Actually all must be done by hands. 
E.g. another approach could be that one single system with the best fit is chosen than a second single system is been added to a portfolio to see if this new portfolio is better than the first single system together. If not then the next one single system is chosen until all single systems has been grouped together to achieve a portfolio with the highes user defined fit. Also this can only be done by hands actually.

Please realize this feature request to save time for the user. There are a lot of advantages that comes up if this is realized. I cannot show them all here. I think it is even more important than request “Speeding up the Portfolio Optimization No. 1” because of more possibilities.

thank you for the suggestions.

 

as for No. 1 this would not solve the issue. Yes, it will speed up the stats computation a bit, but the main problem is that there is simply extremely big number of combinations.

 

as for No. 2, I’m not sure I follow, it seems to me that you’ll miss the biggest portion of possible portfolio combinations from consideration.

 

 

There is a better solution for this, it is using genetic search for portfolios, and it is already planned, it will be ready in one of the next updates of QA.

 it will be ready in one of the next updates of QA.

 

When When When? 🙂

I’m sorry, but it will take some time, we cannot work on everything at once.