03/20/2013 01:51 PM
Posted by Auqo
The data, when framed in the way you've done in those links, would have implied that fire mage was by far the strongest spec in firelands (hint: no)
Didn't play in Cataclysm, so this is lost on me, more or less. Were Fire Mages at the top of DPSBot then due to old content on farm but mediocre in current content, or something?
There's two problems with those links (actually more than two, but I'm lazy, so I'm only going to describe two).
1. The population imbalance problem (don't use top 100)
Different specs have different populations. For example, if you select all parses, raidbots currently has 26,612 samples for shadow priests, but it only has 109 subtlety rogue samples (to use the two extremes). As a result, when you compare the top 100 parses for each spec, you are (essentially) comparing the top 0.3% of shadow priests against the top 91.7% of subtlety rogues. This is not a fair comparison.
2. The fire mage in firelands problem (don't use overall dps)
Those links take the top parses for each spec across all fights, average them, and then rank the specs based on those averages. The problem is, DPS isn't constant from fight to fight - on some fights you'll be doing 200k+ dps, and others you'll be doing 100k. There could be any number of reasons for this difference - lots of AOE, gimmick buffs/debuffs, etc. Whoever takes best advantage of a tier's gimmicks (like fire mages on alysrazor in firelands) will dominate those pages, whether they're strong specs in general (e.g. assassination rogues on that page) or not (e.g. fire mages in firelands).
Basically, you can't make cross-class comparisons based on who wins the tier's gimmick encounters. Fire was a dumpster-tier spec in firelands. It wasn't used for any fight except alysrazor, because the alysrazor gimmick coincidentally played perfectly into the spec's toolkit, and it did so in a big enough way that fire was ranked as the top spec in firelands on a lot of those sites. Spec score was invented in order to minimize this effect - basically, it takes the top spec for each fight, sets it at 100%, and then everyone else is rated in relation to how far off they are from this figure on the same fight. In other words, a spec with a spec score of 82.6 (e.g. enhancement in 25N all parses) performs, on average, 82.6% as well as the top performer on any given fight in the tier. You'll notice that not even the highest spec has a spec score of 100, because they're not the top spec on every single fight in the tier.