It has been my experience that using NNs to predict a MA is very difficult. There is simply too much variability (noise) to deliver a profitable model. This is just my experience and I'm obviously wrong if this NN model is going to win the EA contest.