Back to Community
Quantopian backtest speed - does it vary by time of day, etc.?

Does the Quantopian backtest speed vary by time of day? Day of week? Load on system? Phase of the moon?

On, other than the Web Response Time (which means what?), there is nothing that indicates how fast backtests are running.

4 responses

Subjectively, yes, substantially. PST evening hours tend to be slower than e.g. morning.

Also, backtests seem to be farmed out to multiple different processes/servers in the background. This can cause multiple instances of the same backtest to run at different speeds.

Thanks Alex,

I didn't clock it, but yesterday during the day things seemed really slow. Then, at 5 am NYC time today, backtests seemed to be running faster. Guess each user doesn't get his own copy of the databases and there is a bottleneck when pulling data?

I haven't clocked either but I live in Europe and the backtests I launch on the morning (middle of the night for the US) run on average faster than the one I launch at the end of the day.

I also experienced some cases of heavy computation backtests that have crashed because of the 60 seconds limit when launched on the evening but successfully completed when launched the next morning without any modification. I don't have a big enough sample to reach any conclusion though

Yeah I've had the same experience (UK based), and the math/data-heavy algos don't crash in the morning for me.