I've been writing my own programs for years. My last couple were about break even. I just finished this last one over the last couple weeks. I've been testing it mostly on historical data and it looks really good.
But if you've ever done any probability before with historical data, you know you can always tweak a program to find a profit. The real test is how it does looking forward.
It looks good so far. But ask me again in about a month or two.