Tinkered with it, but not enough to find anything conclusive. My prior is that is very prone to overfitting so it's hard to commit to it in earnest. For long-term research, I'm more excited about using DL to extract medium-term alpha from micro-structure signals.
In all honesty, though, most...