That's a pretty common technique. The problem I've found is that it risks incorrect results be assumed correct. If I refactor your algorithm, and fix a bug in t

That's a pretty common technique. The problem I've found is that it risks incorrect results be assumed correct. If I refactor your algorithm, and fix a bug in the process, I'll be very confused if your tests start failing. The benefit of calculating it yourself first is obvious, you get to cross check your result. That tends to highlight minor errors such as -/+ errors. This is not a problem if stability is more important than correctness of course, but that's not where I am.