In the book "Building winning trading systems with TradeStation" there are several mechanical systems that are applied to daily charts, and backtested on approximately 20 years of data (I think it was 1989-2002, but not sure) and they are profitable. But the author also mentions that the system may produce losses for years. So theoretically there are profitable systems, if you are patient enough to wait 1-2 years for profit that is. I was also playing around with neural networks, and i had examples of networks that are consistent losers and consistent winners. I don't know why, but I also tend to be more inclined towards consistent losers, as there are greater chances that good system will become bad than the other way around (bad to become good)
