For me it is pretty obvious that bots could do alot but not everything. The most distinct difference between computers and men is the former deductive capabilities and the latter impressive inductive skills. A bot is somehow the inductive human capabilities translacted into deductive computer language. In my humble opinion , the edge resides more on money management aspects rather than in pattern recognition aspects, even though both are important. I never claimed that a trader is better off with a random signal strategy, even though I believe there is more chances of profiting out of good money management rather than out of good pattern recognition. Having said that, my point is, the cornerstone of ATS is tackling noise through diligent money management. There is no way to profit (or at least BE) from noise than having a sound MM framework. So, IMHO, the most important building block of successful ATS design is MM. Now, MM is not as easy or simple as many believe. From all the metrics and ratios available, there is one that concentrates statistical measures and the outstanding teachings of gambling. It was coined by Sterling and it was named after him, it is the sterling ratio. Managing your sterling ratio in realtime is about improving your equity curve slope and that is paramount. How cares if one's bot could make 10% today if tomorrow will lose 10% and put itself bellow day one? Who cares about what sharpe is if your equity curve doesn't follow a normal distribution? Anyway, here is some thoughts to tease you.