Hi James,

It sorta depends on what you mean by "forward-walk testing/optimization." You can run algos in simulated live trading mode (see https://www.quantopian.com/help#overview-papertrading). Within the algo, you could do periodic optimization. So, the basic machinery is there, but without more information on specifically what you need to do, it is hard to say if the platform will work for you.

Grant

Hi James,

You might be interested in Alphalens, a tool for running forward testing on alpha factors.

Thanks, I was looking for something like the Grail walk-forward optimizer.

What does the Grail walk-forward optimizer do? I'm not aware of a Quantopian built-in API, but you might be able to "roll your own."

Grail allows testing on unseen or OOS (out-of-sample) data, you can optimize your system on a section of historical known data, run your optimization then “unmask” forward data and see how your system performs. So, as a simple example, you optimize the algo on the 12 month period between 10 years ago and 11 years ago, then walk-forward and test it on the out of sample data in the forward year(s) and keep stepping forward and repeating the process one year at a time, so that you can see how the algo performs on the 'unknown' OOS data. Generally backtested algos peter out pretty quickly or become over optimized or add two many parameters so that you wind up with an overly fitted polynomial. Sorry, if I'm just telling you what you already know.

I'd say 'yes' something like that could be done (depending on the details), except that you'd need to sort out how to set up the optimizer yourself.