Well, the pattern posted a table for last post seems to have a least 4 channels to it, so it may be is a channel a certain multiple and if that’s the case it won’t cut more then probably about 1/3 of the values, which doesn’t seem significant enough for the complexity it would introduce into the code.
I decided to stop pursuing that pattern and instead took a look 2 of the other parameters instead. For the most recent run, which had the period length set to a range of 28-35 days, I found I could filter two of the parameters down to just one value and and then get some average CAGRs over 20% for some of the usage period values. At level of filtering all that feeds into the average CARG is one result per stock per time frame, so that’s as low as I’m going right now since the software isn’t doing stock picking yet.
Just now I started a run that removed the second layer of metrics. Since I’ve got the second layer of metrics narrowed down now, I want to see if there is any correlation between the parameter ranges for the first and second layers of metrics. If so, great, I can just reduce the parameter ranges and I may be on to something. If not, then I’ll need to refactor the metric code so that each layer can have different parameter ranges.