>That's pretty much the same with the benchmarks in hardware or major apps. Some of them score better than expected, and then in the analysis of the results the testers explain why. They usually consider these results as some kind of cheat - as in "it's using this trick to achieve that". As an user, I don't care. If that trick works when I use it, well, the better.
>

Tend to disagree more than otherwise on this topic. If that trick works for other situations, like a better usage of multiple cores via an algorithm not only for the specific test, I'd call it optimization and not trick.
If it involves hightening the CPU speed if a particular test is encountered/recognized, I'd call it cheating and also not a trick ;-))
If it is a precalculated result FOR A SPECIFIC TEST (xxx is prime number yyyy) and not a general optimization done by the compiler like loop unrolling often done in C I'd also call it cheating.