I experimented with the features until I got good results in my evaluations. If I am getting favorable guesses, how does that point to bad features. Favorable guesses is what I was going for in order to assist in trading decisions.

I use a 70/30 split of training/eval