I'm completely math retarded, but wouldn't it be more accurate to just do a statistic using raw data instead? Take the total number of players at 1850+ rating (or 2200 + rating). Then take into account class representation as a fraction of the entire population of players above said rating. Class A, B, C are at X % and are fine...class D, E, F are below X% and are in need of buffs to bring up to par. You get the idea.
I'm genuinely curious about which approach would be the most accurate method of measuring class representation above a certain bracket. I got Cs in my stats classes. :(
No. Class population has to be taken into account or else the data is fubar. Active arena class population anyway. If Blizzard is including low level alts and people who don't participate in Arena at all then they're pretty stupid.
Anyway, here's why (for the people who don't seem to have a grasp on it).
Lets assume that a base % of total players are "good" say 1%, but the number doesn't matter. So 1% of all warlocks are good, 1% of all druids are good, 1% of all warriors, etc... Good players would succeed at any class. We have to make these assumptions because the skill of a player isn't measurable except when filtered through his class.
Now lets look at an example. Lets say there are 100 warriors total population and 20 druids total population. Small numbers because they're easier to work with, and extreme total imbalance to make the point more obvious. Of those 100 warriors there are 10 over 2000 rating and of those druids there are 5 over 2000 rating. If we don't account for total population it would seem that warriors are twice as viable than druids. But, accounting for total population we can see that only 10% of warriors are above 2000 rating whereas 25% of druids are above 2000 rating. This throws a totally different spin on things.
You can see that we can draw 1 of 2 conclusions from this data. Either druids are more powerful or that warriors are. However, which makes more sense? If we want to ignore total population we have to make the assumption that there are just more good druid players, which is not only absurd, it contradicts our initial assumption. So this conclusion has to be thrown out. And the only other option is that druids are more powerful than warriors based on a % representation in 2000+ rating.
Of course everything gets more complicated if you want to take into account class compositions, synergy, counter comping, etc... So much so, that I'd argue that it makes any conclusions we want to draw irrelevant. All these statistics allow people to do is misinterpret them and tweak them to support their own ignorant and biased opinion.