> Random selection is important to know if your system actually improves anything, or if the results are only good because the input was good to begin with. Ri

> Random selection is important to know if your system actually improves anything, or if the results are only good because the input was good to begin with. Right, this is the scientific approach. You have to disprove the null hypothesis, otherwise your theory is unsubstantiated.