Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Calibration of algos?

Hi guys,
How do you calibrate the parameters of your algo?
In a simple example for example, how to choose the moving averages you want to run a simply MA crossing algo. I know you run into the risk of simply fitting the parameters to the historical data window you use but in any case it's always useful to see that kind of analysis (without having to manually modify the parameters and rerunning the backtests for a range of values for your parameters).
Is there functionality for this in Quantopian or do we have to write something to do that?
Thanks,
J

2 responses

Yep, that's a major pain point with Quantopian right now in my opinion. There's no way to do automatic optimization. I was told that they are working on a research environment to address this and looking forward to that but in the meantime, I use Zipline to code up my algorithm and do some automated optimization. Then, I try them out in Quantopian. Even though Zipline's data (Yahoo) is somewhat different from Quantopian's data, in most cases, an offline optimization in Zipline provides similar results in Quantopian.

Thanks Mete, you're right yes, I've seen the research environment they are planning to roll out. I think that's very much needed if they want this to work well!