Just to contradict myself, things like http://cs229.stanford.edu/proj2013/TakeuchiLee-ApplyingDeepLearningToEnhanceMomentumTradingStrategiesInStocks.pdf are somewhat interesting in that they concretely show an application of data.
I personally expect that if something like that worked at all, it’d get incorporated into the standard algorithms and the effect will go away (per the mentioned EMH) within a few months of publication (and note that it’s a 2013 paper).
Just to contradict myself, things like http://cs229.stanford.edu/proj2013/TakeuchiLee-ApplyingDeepLearningToEnhanceMomentumTradingStrategiesInStocks.pdf are somewhat interesting in that they concretely show an application of data.
I personally expect that if something like that worked at all, it’d get incorporated into the standard algorithms and the effect will go away (per the mentioned EMH) within a few months of publication (and note that it’s a 2013 paper).