Kind of a short conclusion, but after digging through it a little bit, I’m fairly convinced I still have the wrong model. There are a few signs when something isn’t there, although I can’t quite explain why. On the upside, I am a bit closer. The numbers are a lot better (about 10% better) but still not good enough for an edge. This means (hopefully) that I’m on the right track, but that I’m missing a few components. I have a few ideas as to what may work, but creating the data for that may be difficult and I may have to scrap this line of thought and try something else.
-Build data, not signals.
Back to the drawing board