Dynamic Universe and Volume SMA - Backtest and Log issues

Hi Eoin.

As the error describes, it is thrown when the algorithm takes too long to finish a loop. I do not know why you got this error specifically, as I did not encounter this error, even though I ran the algorithm multiple times. Maybe you are entering a loop that takes to long to finish, due to the high amount of stocks being picked?

To answer the first question, you can log the added and removed stocks with the following code:

    def OnSecuritiesChanged(self, changes):
        for added in changes.AddedSecurities:
            self.Log(f"stock added {added.Symbol}")
            
        for removed in changes.RemovedSecurities:
            self.Log(f"stock removed {removed.Symbol}")

Lucas

Hi Lucas,

Many thanks for your advice - I entered your OnSecuritiesChanged method for logging universe changes and actually managed to get it to backtest without getting the error as described in my original post, though it still took a number of tries.

I suspect you're right when you say “Maybe you are entering a loop that takes to long to finish, due to the high amount of stocks being picked?” . However, I just don't understand why the error is throw only some of the time?

May I ask what backtest node type you're using? I'm using B-Micro and what to see if a more powerful node might solve my issue, if I can't find another workaround.

Successful backtest attached for reference.

Kind regards,

Eoin

Hi Eoin

I use the B2-8 note (the one that cost 14 bucks a month). Yes, it is faster than the free note, but it should not take more than 10 minutes to finish a loop. It takes 50 seconds for me from when i hit “backtest”, to the backtest is completed. It should be noted that a lot of securities is going through the algorithm, which might hit some RAM limits on the free notes.

A workaround (if you are willing to change the algorithm)  would be to have tighter constrains in the algorithm, before enterering the first for loop. The history request is what is most likely causing the trouble. Try taking the 100 most liquid stocks (or something like that) and see if that help. You could do something like this

selected = sorted([x for x in coarse if x.HasFundamentalData and x.Price > 10], key = lambda x: x.DollarVolume, reverse=True)[:100]

and then sort by price after, or what else you wanna 

Lucas

Hi Eoin,

The attached should be stable. I certainly used Lucas' code for OnSecuritiesChanged as it was proper.

Your strategy to calculate volume averages for all assets in (2$, 10$) inherently takes time and memory…

Fred

Hi Lucas, Fred,

Thank you both very much for your input, it's been a great help.

I've implemented both of your changes and limited the amount of securities returned by the coarse filter to 50, in the interests of quicker backtesting. Attached is a recent successful backtest for reference. 

I also upgraded my backtest node to the B2-8 (mainly out of curiosity) and am pleasantly surprised by how much faster it is, even after scaling the algorithm up a bit. 

I am, however, running into the following issues:

• I'm having trouble logging the number of symbols currently in the universe (my attempt is at line 42 in the attached). I know in the attached code the result will obviously be 10, but I'm looking to implement this for future backtesting purposes. The code I currently have doesn't print anything to the log.

• I'm also having trouble logging the value of the the 30 day Volume SMA for each symbol (my attempt is at line 52 in the attached, commented out). I keep getting the error “name 'selected' is not defined”. I've tried defining ‘selected’ as an empty list in the Initialize method (among other things), but haven't been able to fix it.

I'd be grateful for any advice either of you (or anyone else happening upon this thread) might be able to give me. 

I'll slow down on the questions for a while after that, I swear.

Eoin

Hi Eoin,

return selected[:10]
        
self.Log(f"Number of assets: {len(selected)}")

In that segment, you write self.Log() after the return - it will never be executed. In fact, selected does not exist anymore there.

Use this:

self.Log(f"Number of assets: {len(selected)}")
return selected[:10]

In:

self.Log(selected(symbol).thirtysma.Current.Value)

selected is not defined. It is defined in CoarseSelectionFunction but it is then not visible in OnSecuritiesChanged. You want to use your self.averages dict:

self.Log(f"Averages: {self.averages}")

Fred