Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Simple wizards for building statistically valid models

http://blog.quantopian.com/algorithm-builder/ got me thinking, is there a way to make a wizard that would build an algorithm which corrects for (rather than accentuates) cognitive biases towards extrapolation and data snooping?

I am thinking of variations on algorithmic implementations of the Quantitative Value screen, perhaps using factor model monte carlo rather than historical performance... thoughts?

3 responses

Hello Simon,

Interesting topic. In my mind, there's a very general problem of determining, given a data set, if there is any useful information available (sort of a signal-to-noise ratio). This would be a kind of low-level screen that would not rely on a specific trading model. It's a pretty murky idea to me at this point--I'll do some poking around to see if there's any guidance out there.

Grant

Yes, that is a must-read paper.