Is there a way to test the impact of a single indicator/parameter when optimizing by turning it off or excluding it as part of an optimization run?
For example, I've found a number of indicator filters using the Analysis Series visualizer of my backtest results, but they only test one indicator on one backtest run at time. I'd like to run an optimization where I can test which combination of multiple filters ultimately has the most benefit.
To do this, I believe there has to be some ability to backtest not just a range of parameter values but also the impact of if the indicator/parameter was turned off for a run as well. Manually, I would go to design surface and disable the block and retest, or optimize a range of values that I know would cover a backtest run where the parameter would have no impact (RSI < 70 to 100 by 10), but both of these feel inefficient.
Is there a way to automatically or more efficiently do this? If not, I would like a button for this to be a #FeatureRequest
For example, I've found a number of indicator filters using the Analysis Series visualizer of my backtest results, but they only test one indicator on one backtest run at time. I'd like to run an optimization where I can test which combination of multiple filters ultimately has the most benefit.
To do this, I believe there has to be some ability to backtest not just a range of parameter values but also the impact of if the indicator/parameter was turned off for a run as well. Manually, I would go to design surface and disable the block and retest, or optimize a range of values that I know would cover a backtest run where the parameter would have no impact (RSI < 70 to 100 by 10), but both of these feel inefficient.
Is there a way to automatically or more efficiently do this? If not, I would like a button for this to be a #FeatureRequest
Rename
Yes, you control which optimization parameters to use with these check boxes...
This disables running the parameter through its values (keeping just the default) - not testing if it were also off. Given your picture example, what I'm looking for would be equivalent to testing each profit target value (profit target included) and then also testing if there were no profit target (profit target excluded). But the goal would be to be able to do this with multiple parameters to find an optimal usage combination, where if it were just one I could disable the building block.
This can be accomplished in a Coded strategy for sure. You'd just have to structure the code and logic the right way so your parameters control what you want.
I'm concerned here about further complicating the already complex optimization UI.
I'm concerned here about further complicating the already complex optimization UI.
Ahh. Well, I'm not much of a C# coder, so if you ever get around to adding that feature for non-coders, that would be a high value-add in my opinion. Optimizing for not just which combination of parameter values is best, but for what combination of parameters themselves actually improve a strategy would be really beneficial to know.
Well another way could be using a Multi-Condition Group. You could put a few filters in the MCG and then configure it to "Condition number N is true." And that "N" could be optimized.
Your Response
Post
Edit Post
Login is required