Advanced Filtering doubts: Correlation and “Extreme Strategies”

Hello,

I’m trying to build and finetune my Custom Project to be as automatic as well as sharp as possible.

I’d like to ask the following questions on these 2 filtering topics:

1) Correlation: I see a very interesting option in the Builder to filter-out strategies that are correlated, but have some doubts:

– When strategies are correlated, how does the software behaves? Does it remove both of them? Or only one? And with which criteria to keep one or the other (e.g. Net Profit)? I see an advantage in reducing correlation, but would expect only one out of two is removed.

– Does adding a correlation filter slows down building/generation? If that is the case, Is there an option to filter by correlation at later stages (e.g. retester), when strategies are way less?

2) Extreme strategies: I see often the strategies that pass my protocol are quite extreme, e.g. 100% short or 100% long. I don’t feel comfortable with these strategies and want to remove them. While I still want to feature a moment of human review of trade analysis stats, I would love to remove automatically at least these very extreme strategies to reduce some of the work.

– Is there a way I can filter strategies by a certain threshold of long or short trades? My first thought was to use the “Ranking” tab –> Custom filters –> Select “# of trades”, full period,  value as % –> Use as a condition >60%. Would add 1 condition for long and 1 for short. So this would me would have meant: “When # of long (or short) trades in % is higher than 60%, filter”. But this doesn’t seem to work. Is there a way to reach my objective?

Thank you so much in advance for your support!

Diego

 

 

 

Hello,

Can anybody help with advice on this?

Many thanks again!

 

-probably it chooses the highest fitness one and keeps that one but not sure…

-slows it down ALOT!

you may check this out, something is fishy about the function:
https://roadmap.strategyquant.com/tasks/sq4_8586

apparently, there is a module that can filter by correlation but I don’t know too much about it or how useful it is yet:
https://roadmap.strategyquant.com/tasks/sq4_9296

no worries you can filter by “trades symmetry” stat for the number of trades ratio or “symmetry” stat for the performance of long vs short probably based on net profit ratio or something similar I forget. I use the former often.

in 135 there a lot of symmetry bugs most are fixed in the current dev build but there are still a few problems:

This list includes the codebase indicators but still a handful of built in indicators are not symmetrical…

Dev 5 + newest versions of extras from codebase
All indicators: Builder not respecting parameters: https://roadmap.strategyquant.com/tasks/sq4_9294
DEMA: MT5 doesn’t match: https://roadmap.strategyquant.com/tasks/sqp_0184
DEMA: Applied price issue and pseudocode issue https://roadmap.strategyquant.com/tasks/sqp_0177/edit
Ulcer: Internal short side issue https://roadmap.strategyquant.com/tasks/sq4_9121
Schaff Trend Cycle: Minor problem when slow and fast period are same https://roadmap.strategyquant.com/tasks/sq4_8663
Bulls Bears Power: symmetry levels issue https://roadmap.strategyquant.com/tasks/sq4_9287
Aroon: falls / rises issue https://roadmap.strategyquant.com/tasks/sq4_9128
FIBO: Bearish readings gives bullish levels instead https://roadmap.strategyquant.com/tasks/sq4_9109
VWAP: MT5 Terminal mismatch https://roadmap.strategyquant.com/tasks/sqp_0180
DVO: significant rounding issue of internal calculations can be a big problem on M15 and lower: https://roadmap.strategyquant.com/tasks/sq4_9313
Disparity Index: is/rising/falling: https://roadmap.strategyquant.com/tasks/sq4_9251

**Aroon just got a fix for dev 6 but its not out yet…

Also, if you want 100% true symmetry you will actually need to use a template:
https://roadmap.strategyquant.com/tasks/sq4_5170

But watch out, you’ll need to work around this:
https://roadmap.strategyquant.com/tasks/sq4_8856

Thank you for your help!!