Sometimes I'll throw out times over 1 or 2 stdevs away from my average, and consider the average after doing that more accurate and useful. It tends to remove any particularly bad solves which make my average seem worse than it should be, while also removing any lucky times I also wouldn't consider to be typical. Sort of like removing the best/worst times from an average, except this is probably more useful for large averages. It usually doesn't impact my averages much, but it can, for better or worse. I even built this feature (averages with outlying times removed) into the timer I made.