WL7 utilizes multithreading when optimizing (and that’s great), but not when backtesting. Can it be changed?
My everyday routine suggests using backtesting very often and I like doing it with datasets, not single tickers. Although I am now using mostly Daily TF, for 300+ tickers dataset backtesting process is not a question of a couple of seconds. So I would not refuse being able to do it multithreaded as well. Is it possible?
I’m afraid it’s by design and this single threading is because of WL7 being able to look at another symbol’s positions and so on, so nothing can be paralleled. But I must ask this question).
My everyday routine suggests using backtesting very often and I like doing it with datasets, not single tickers. Although I am now using mostly Daily TF, for 300+ tickers dataset backtesting process is not a question of a couple of seconds. So I would not refuse being able to do it multithreaded as well. Is it possible?
I’m afraid it’s by design and this single threading is because of WL7 being able to look at another symbol’s positions and so on, so nothing can be paralleled. But I must ask this question).
Rename
Since WL needs to step through each bar and run Execute for each symbol for every bar, we cannot run each symbol in parallel. We could possibly run each symbol in parallel during each pass of Execute but It would be a major overhaul and might not provide much boost especially with the required locking that would be involved. Feel free to submit it as a request though and once it reached critical mass we will work on that change.
We already run the data collection in parallel which takes the bulk of the processing time in most cases.
We already run the data collection in parallel which takes the bulk of the processing time in most cases.
Your Response
Post
Edit Post
Login is required