Replace similar strategies in databank

Voted for this task.

PS that builder task was only trying to build 20 strategies, so finding 8 x DeMarker + PSAR strategies out of 20 is REALLY problematic!

Voted for this task.

6 / 15 results are duplicated in this image

5 / 9 in this one

19 out of 65 !!!

Genetic settings trying to create as much diversity as possible based on reading help and e-mail discussions, but still not working that well. 

This is literally about 5 minutes later from my 19 / 65 post and now I'm up to 44 / 94 with the same entry criteria!

If this does get included, can we please have 3 options:

1. Keep all strategies (do not delete similar strategies)
2. Keep only strategies that have the same entry and exit criteria, but different variables and pick the best one for each 
3. Keep only the best strategy for strategies that have the same entry and exit criteria

The "best" strategy will be based on Fitness.

For option 2:

1. Example keep: If 2 strategies have the same entry and exit criteria (databank columns), but each have different other variables, then keep both
2. Example replace: If 2 strategies have the same entry and exit criteria (databank columns) and have the same variables for everything else

For option 3:

1. Only keep the best strategy where displayed entry and exit criteria are the same (from databank columns)

Hi,

I think this could be a good new feature. But should be an easy option to activate or deactivate. I mean, do not overcomplicate it with setup and parameters just check the correlation %. Simply; i) add a boolean on the Builder to be able to On/Off this option, ii) New field where you can setup which is the maximum % correlation allowed.

Voted for this task.

discussed here many times - https://roadmap.strategyquant.com/tasks/sq4_5945, https://roadmap.strategyquant.com/tasks/sq4_3998

i dont think so, that its reasonable to solve the similarity only on used building blocks...different parameters and mostly different used SL ant PT values means that even the strategy is using "same" building blocks could be totally diferent

for me the similarity is more equatation of correlation or RDD vs number of trades

and for now we have bigger problems - because there is not solved the simplifier of strategies - because we can see building blocks in the strategy which is doing nothing - https://roadmap.strategyquant.com/tasks/sq4_3807

All this needs is a quick text comparison, which is EXTREMELY fast in any database solution.

These 2 strategies (attached) are exactly the same strategy (one strategy). Both have exactly the same variables, but different parameters. I only need one of these strategies to remain in my databank, not both

because I will optimise these parameters during WFM assuming this strategy passes all robustness tests

This is good but it would need to be done in crosschecks ofcourse. Real diversified strategies have diffrent entry codes but there are ways to trade when just off setting the entry points actually works great especially in trending markets which is the only time breakout strategies makes money  using the same entry so totally removing it when building is not an option and also impossible since would be to slow but as a crosscheck would be great.

Voted for this task.

Agreed mabi. A correlation crosscheck would be the best

Voted for this task.

Voted for this task.

Voted for this task.

there already is check for similar strategies, but the similarity is evaluated by the similarity of trading results, not by the used indicators. 

Small chanes in parameters can create strategy that trades very differently.

yes, this could be true, but we are still waiting for some simple correlation duplicities remover, which we still dont have in SQX

fit to existing portfolio cant be used for our purposes