Hi Tamas,
https://strategyquant.com/doc/programming-for-sq/example-plugin-a-complete-custom-project-task
I have read this document carefully.I understand the existing design and usage methods, and there are also bugs in the existing modules.
Bug1:
The current method does not consider filtering the correlation of strategies within the databank Target, which means that the correlation of the filtered results in the databank Target can be greater than the preset value.
Here is an example to illustrate.
I currently have 39 strategies waiting to be filtered based on correlation. I put these 39 strategies in the Databank Source.
If the Filter by correlation plugin on Databank is used for these 39 strategies, when the correlation period is set to Day and Max correlation is set to 0.2, there are still 33 filtered results.
I selected 10 out of these 33 strategies and placed them in the Databank Existing. Then, I launched the Filter by Correlation task in the project, which resulted in 26 strategies in the Databank Target.
Looking at the correlations of these 26 strategies, it can be found that some of them have significantly higher correlations than 0.20. When we used the Filter by correlation plugin on Databank to filter these 26 strategies, we found that there were exactly 23 left, with 10+23=33.
Please refer to the attachment for the relevant documents.
In addition,It is strongly recommended that If the number of strategies in the Existing databank is 0, then directly filter the correlation of the strategies in the Source databank (The same as the function implemented by the Filter by correlation plugin on the databank) and save the results to the Target databank.
Bug2:
The databank includes Source, Target, and Existing. Every time I modify the corresponding databank, as long as I simply switch between tasks or projects and return to Filter by correlation task, I will find that the previously set Source databank returns to the default Results databank.Unable to save modified databank.
Feature1:
Add ranking criteria such as Rank Strategies by Net Profit to ensure that higher profit strategies are not discarded in correlation filtering.
Feature2:
Referring to the design of the Automatic Portfolio Builder task, there can also be data range settings.
Best regards.