Yeah, mainly I was thinking like giving the machine bunch of market features and it tells me what is random and what is not, and what has happened before or not. I am wondering how trustworthy that would be considering issues such as overfitting or just imperfect training data. But I agree human...