I actually checked out your algorithm today during work (I work at a big bank, aka dead end for a "technologist"). It's an intriguing concept, but I still feel

I actually checked out your algorithm today during work (I work at a big bank, aka dead end for a "technologist"). It's an intriguing concept, but I still feel as though classifiers are a poor technique for P&L optimization for a given portfolio. Out of curiosity, what type of data set are you using for backtesting, and what time frame? It's not entirely clear that asset (i.e. stock) prices follow Brownian motion/Weiner process, rather they may be discontinuous.