Is there a quicker way to identify similar strategies?

Aside from going through each strategy, looking and comparing the pseudo code, is there a view column that helps identify near matching strategies. Strategies that are pretty much the same except for a couple of numbers?

Not that I know of.  I spend a lot of time clicking on strategies and removing “almost a twin of” results.

The filtering function need be improved.

In my experience, there are many strategies in databank, but some of those are just different with the stop or profit target, and the main entry and exit conditions look very same!

Maybe we can give the parts of strategy different  weight.

entry conditions and entry order > exit conditions > sl/pt/profit trailing and so on .

So we can set the criteria in ranking options, just keep to the top 5 or 10 strategies, or accroding some criteria(return/dd,SQN,profit factor….) , if those strategies have similar entry and exit conditions.

I agree would also like to see some Data Mining Bias Filtering added to seperate the chaff from the wheat so to speak

The filtering function need be improved.

In my experience, there are many strategies in databank, but some of those are just different with the stop or profit target, and the main entry and exit conditions look very same!

Maybe we can give the parts of strategy different weight.

entry conditions and entry order > exit conditions > sl/pt/profit trailing and so on .

So we can set the criteria in ranking options, just keep to the top 5 or 10 strategies, or accroding some criteria(return/dd,SQN,profit factor….) , if those strategies have similar entry and exit conditions.

I believe this exists already in some part under one of the tabs where you filter the data bank using ‘weighted fitness’ this section allows you to prioritise certain parameters over others and the combination that most fits is expressed as the highest fitness score.

Sent from my iPhone using Tapatalk

 Guys!

this is not so easy. you can not just delete similiar strategies if you are not sure if they are robust. what if the software delete good robust straegy and that one which stay with better RDD for example will be not robust?

this function should be option.

Patrik

Patrik,

We’re not looking for ways that SQ will just automatically delete similar strategies, only a way for us to easily identify similar strategies (whether at the robustness stage or wherever) so we can delete them manually ourselves.

If I have 100’s of strategies that have passed various stages of testing, I’d like to know if one dozen are very similar, and then if another dozen our very similar, etc etc…so out of each dozen I can only be left with the best and most robust of that batch, rather than dozens of the same.

But yes, you’re right…this should definitely not be an automated process, as our input and evaluation of the strategies is more important at this stage.

Do we mean similar in rules (building blocks used) or similar in terms of equity curve, number of trades, win/loss ratio?

For example, strategy 1 and strategy 2 have same entry rules, same order types and same exit rules, but very different parameters, so strategy equity, no of trades etc is very different.

Strategy 3 and 4 have different entry rules, different order types and different exit rules, but equity curve, number of trades etc is not that different.

Which are similar?

Well ultimately my reason or need for this is to avoid highly correlated strategies that share similar positive and negative performance during the same periods of time when building a portfolio.

I think both of your examples are important…

But if I could easily identify your first example where entry, orders, and exits were similar, it might help me determine which strategies I want to add to a portfolio evaluation.

I could group those similar strategies together, and then more easily ‘at a glance’ have a quick look at other performance metrics to see if there are greatly varying differences. <— This is the bit I can't do right now because I don't think there is a way yet, without manually going through each pseudo code file.