One common mistake is to take the average of run times (for identical runs). I think taking the minimum should be better under reasonable assumptions. Edit: I

One common mistake is to take the average of run times (for identical runs). I think taking the minimum should be better under reasonable assumptions. Edit: I now see that this was mentioned elsewhere here. Good!