Strategy tester

 
When increasing the accuracy on historical data from "1 min OLC" to "every tick" to "every tick based on real data": If I get less and less successfull results, what could that indicate? I mean wouldn't the result be random, rather really one sided?