Originally Posted by
Endus
It isn't. It never is. The problem with "top 100" is that it artificially inflates higher-RNG specs thousands of DPS higher than they realistically should be. DPS distributions are bell curves; most people perform close to the median. Top 100 only looks at the very highest percent; for Fire mages, for instance, there's almost 10,000 samples at 25-normal, so top 100 is the top 1%.
To use some numbers, imagine you have two specs, A and B, A averages 102k, B averages 100k (numbers made up for convenience). A's the better spec, right? The problem with "top 100" samples is, you have to consider RNG: if A averages 102k but has relatively low RNG, only varying +/- 1k or so based on random chance, their "top 100" are going to be close to 103k. And if B has much higher RNG, varying +/- 4k, their top 100 will be closer to 104k.
By looking at top 100, you're thinking Spec B is averaging 1k better. When in fact they're averaging 2k worse. They can only do better than A if they're in one of those 1% of lucky streaks. This isn't a workable model for class balance. It's artificially giving inherent value to RNG, when RNG should have no value at all.
The numbers will get more reliable as we get more parses, but Top 100 is flawed right out of the gate.