Lies, Damn Lies, and Statistics

While there are countless forum posts, blog entries, and debates that talk about how to fix various class imbalances, it is pretty rare that I find many discussions with which I agree. I've played mage almost exclusively for about as long as possible and I think my class bias is transparent in everything I write. I think my perspective is similar to Ming's on class balance in many ways, and this perspective includes:

• Balance is when every class can succeed with similar probability in every arena bracket and has a "reasonable" chance against any other class 1v1.
• Every spec of every class need not be viable in every bracket, but classes should have multiple balanced specs that are effective in various arena brackets.
• Balance should scale up from 1v1 and 2v2 up to 5v5.

I'd like to briefly address the point made concerning how the previous "weighting" methodology was being calculated incorrectly. While the weighing is naive, I disagree with it being less valid than some of the other proposed solutions. By dividing by team size, you basically normalize each arena bracket size into being a synthetic one man team; while there are certainly issues and assumptions implicit in doing this, simple taking a simple average of the three brackets isn't mathematically robust...at all. Weighing the brackets as blizzard weights points is illogical to me since most of the serious competitive play (CGS and WSVG) are in the smaller arena brackets. The point has also been made that there are many more viable combinations in 5s as opposed to the small brackets -- this is somewhat supported by data. Adapting the class weights based on total number of people who actually play the class is reasonable, but I have some concerns as to how valid the warcraftrealms data really is and how relevant data really is. The data certainly isn't necessarily representative of the arena PvP class breakdown.

I think if someone wants to extend this sort of thing into a real statistical analysis they should develop a compromise method of normalizing the arena brackets and discuss the class percentages relative to different arena "tiers." Compare the class makeup of top 20 teams vs top 500 teams, what classes seem to fall out? I doubt you'll ever get a ton of agreement on how best to collect and analyze numbers like this as everyone is extremely anxious to draw conclusions from them to support their own agendas. In the future, I'll provide a few different "normalizations" when I try to define a single metric for arena class composition, but recognize that this sort of thing is as much art as science.

Regardless of how you feel about the normalization, the individual bracket data to some degree speaks for itself. I'll briefly discuss EXTREMELY GENERALLY what I take away from looking at numbers like these as I'd like to later talk about more specific solutions:

Hunters and Mages both are extremely powerful, well-balanced classes...in open fields. There have been a million billion discussions on why LoS is unfair to either class blah blah /emo whine whine blah. The problem is not in either class, but is simply an issue of environment design. If there was a fourth arena that was largely open where it was impossible to escape hunter's range, hunters would become somewhat more prevalent simply to their power in this one locale. (Or picture a map with a single bridge and an open area on each side) Including a map with less LoS 'abuse' would shift the class composition perhaps better than trying to buff or nerf various abilities -- I picture something like the STV arena when I think of "open" arena environments.

The relative melee class balance is a significant concern. Warrior/paladin fusion is simply too strong and it creates problems for warriors, paladins, and rogues. While so much emphasis is always placed on buffing rogues to make them more viable in the larger arena brackets, I think the lopsided nature of warriors is at least as big an issue. The whole "warriors are horrible 1v1 but great in group PvP" is not acceptable. Warriors need greater independent strength at the cost of their scaling in group PvP situations. Some sort of self-freedom mechanic at the cost of chain mace stuns and a reduction in the scaling damage through group buffs is needed to compensate. Paladins are in the same boat. Trainable repentance, a better non-direct healing option, and a reduction in the potency of their group abilities would yield a class that performs more evenly from 1v1 to 5v5 and paladins might be finally liberated from simply freedom+spam healing their warrior buddies. As for rogues, I'm more of a fan of the "Expose armor reduces target's resilience to zero" type of change than trying to tweak the potency and duration of the many cooldowns. And mana burn poisons? =P

The druid/shaman relationship is also hard to deny. Druids are prominent in the small arenas and absent in 5s, while the opposite holds for shamans. This is difficult to address; ever since BC rumors first leaked, I thought the idea of elemental summoning totems that had no use in arena was a stupid idea. Perhaps scaling the potency on the totems such that one could be used in arena would elevate their status in the smaller brackets. Their prevalence in 5s comes from their place on both 4DPS caster teams and "2345" teams, but it isn't like shamans are particularly overrepresented dramatically in 5s. Druids are extremely rare in 5s as its hard to include a healer into the lineup that does not bring a defensive dispel. I don't really have a nonspecific improvement to this problem (and if you haven't noticed, I've been avoiding specific class changes as much as possible for the sake of general discussion), but I'm not really convinced yet that druids really need any sweeping changes to be powerful in 5s.

There were some great comments on the last post, and I do plan on putting out similar data in the future to which I'll certainly credit those who've contributed to improving the methodology -- for now, you know who you are. I'd like to talk again about specific class improvements so I'm anxious to read what sorts of tweaks everyone thinks would improve overall arena gameplay. Think of how changes don't just make a particular class better or worse in a single bracket, but how they affect the class from dueling all the way up to 5s.