I read about an interesting strategy yesterday in Bloomberg Markets, by Douglas Edler and Jonathan Lin. I can't find a link to it at the moment. Basically run a linear regression (6 months?) on the positive and negative market return days. This gives a bull and a bear beta. The goal is to invest in stocks with positive "convexity", where the bear beta is 1.2. Over the last five years, it seems to have produced a legit 7%+ excess CAGR. When quantopian allows screening over big/consistent universes we can replicate it here!