thanks for reading, I agree, its very tough to find an algorithm to predict the market to generate excess returns. Have you tried using technical analysis with machine learning? To me technical analysis is essentially feature engineering so it makes sense to let ML to decide the weight and subset of features to use. I have read a few papers which have taken this approach and produced supposedly decent results. I am generally skeptical because I think a lot of these studies really just capture the momentum effect, which is a replicable risk premium and not alpha. I would be keen to hear any thoughts you have on the subject.