Two journal papers interested me today. The first is "101 Formulaic Alphas" that was published some members of WorldQuant team. They explicitly list 101 "real-life trading alphas used in production-" that they have discovered through data mining.
Flipping through them, what jumps out at me is the ubiquity of cross-sectional rank.- maybe it turns out to be a good proxy for momentum and that's why it turns up all over the place.
Anyhow, I'm going to recreate these alphas, probably on Quantopian so anyone can play with them.
The prevalence of cross-sectional rank spurred some google searches, which led me to "Jumps in Cross-Sectional Rank and Expected Returns: a Mixture Model." The TL;DR for that is that they first modeled the likelihood of a jump in the next timestep and then modeled the conditional return distributions and used the two to create a nonlinear model of returns which was used in a trading rule with promising results. I'll have to come back to this one when I'm done with 101 alphas.
No comments:
Post a Comment