Exactly. I can use the exact same set and have a variation of 50DPS simply because of human error, latency, good vs bad damage ranges, critting all my high potency skills, the list goes on.

The only, true way, to test stats and different "builds" is in a vacuum where RNG and Crit are normalized, with human error removed from it.

Any other method isn't scientific. You need to eliminate as many variables as possible when conducting tests.