It is quite hard to find a bias that has lasted for quite a few years and which will hopefully work on into the future. Even then any bias that an algorithm works may disappear at any time leaving the algorithm useless if not causing losses.
I don't care about win loss ratio or anything like that in my algorithms. I focus on expectancy. I find that it doesn't matter if it only wins 10% of the trades or up to 95%; as long as it has a good expectancy than it should be theoretically profitable. (As long as the bias doesn't disappear, market change, etc.)
I also calculate in excessive costs into my algorithm; for example I calculate in 4 ticks/round turn slippage and experience an average of less than 1 tick in some of my algorithms. I reprogrammed the slippage calculation for NinjaTrader because I didn't like the default back tester.
To be honest it wouldn't matter me if it's day trading or swing/longer term trading that I was doing, as long as at the end of the day I'm still making money.