Back to Community
Major Algorithm Startup Issues This Past Week - Has This Been Resolved Or Not?

I've been trying to run two algorithms (separate robin-hood accounts). Both rely on opening positions just before close and closing them @ open - holding though regular market hours is not at all profitable. They have both failed on various days (>70% of the time) due to an error on quantopians end all week and thus both accounts have both fallen 10% (they open positions, then hold them for days, not overnight like they are supposed to - because the algorithm fails in the morning, and whenever I put it up, it fails again). Back-tests suggest that they should be up 2%.

This isn't a big deal in the long run - as long as it isn't going to continue to be an issue. the following link says that the issue is "resolved"

"We've been managing a recurring resource problem affecting our
algorithm startup process. This morning the problem was particularly
bad, and this limited our ability to restart disconnected algorithms.
We are working to mitigate the problem."

That was posted @ 5:21 pm EDT.... I had a live algorithm fail due to an error on quantopians end @ 5:38 pm EDT

So is the problem solved? Or do I need to start waking up @ 5:45 am (MST) just too make sure I'm not about to lose hundreds/ thousands in a single day again?

If the problem hasn't been fixed, that's fine, I just need to know.

1 response

I had the same problem on monday morning with an algorithm that I started on sunday. It resolved itself by mid-morning after numerous restarts. Then on Tuesday I got an email saying that my algo needed to be reconnected to RH but quantopian didn't show this on the algorithm's dashboard. The algorithm wasn't making trades on Tuesday until I restarted it. I restarted it under a different name (this is important) and it worked perfectly.

Something I think I've been thinking about lately as a potentila cause of errors is that every time I restart an algorithm and it errors out, even though it doesn't tell me why I suspect that it's caused by leftover context variables. I can't think of any other reason that I would get an error in live trading when I restart an algorithm vs backtesting and paper trading which never get an error on startup.

According to Alisa here:
"If a live algorithm has a restart, it will return to the same state provided that "context" was used to save state. "

Now going forward as potential solution I've changed all my context variables to go into a separate dictionary inside context and I wipe this dictionary clean at the beginning of my initialize() funcion. I suspect that restarting your algorithm under a different name (you can change the name when you click trade live) also prevents it from using the same context variables but I'm not sure.

like this:

context.var = {}  
    context.var["allow_unlimited_daytrades"] = False  

instead of:

 context.allow_unlimited_daytrades = False  

I'll let you know how it turns out next time I have to restart the algorithm, probably next month