Which class gets invited as what?
by Allison Robert 
Aug 2nd 2009 at 4:00PM
Veritable Avarice, a new blog on moneymaking in WoW, took a break from financial discussion and looked at class representation in tank, DPS, and healing roles by filtering and comparing data available from WoW Popular. Spec population was then checked against class population data available from Warcraft Realm's census and three live realms. Data differences, according to VA, weren't statistically relevant, and he/she is pretty sure that the numbers are at least a ballpark representation of which class is most likely to be filling a particular role within a group.
I play a Druid, so that's really what I feel comfortable commenting on here. While I can't speak to the ultimate accuracy of the numbers, I do a lot of pugging and have to admit that VA's data seems pretty close to what I've seen on my own server. The tank numbers are also consistent with a few things Ghostcrawler's mentioned recently concerning the overwhelming population advantage still held by Warrior tanks, although I wonder whether the Feral statistics are somewhat inflated here by the overlap between Bear and Cat specs. Feral tanks have all but vanished from 5-mans on my server, and it's not uncommon for me to get comments from healers that I'm the first Bear they've healed in months. Less surprising is the representation advantage held by Druid healers. Trees are insanely good in Ulduar, and between this, the rise of the Death Knight, and the de-suckaging of the Protection Warrior spec, that probably accounts for the gradual disappearance of the Bear. Also thought-provoking is just how few Druids hold a share of the DPS pie.
I'd love to hear from members of other classes on the data and how closely it dovetails into their own experience. There's a quick note for Warrior players (or anyone interested in the DPS graph) past the cut, as there's a small mistake on the relevant graph.
Quick note for all you DPS Warriors out there: when we received the tip, Matt Rossi noticed that Warriors seem to be missing from the DPS pie chart, but I think I've figured out where they went; there are 10 classes represented on the chart color-wise, but Warriors were accidentally left off the list to the right; they'd have to be the light-blue share at the very top, and so clock in at around 4.02% representation among DPS.
Filed under: Analysis / Opinion, Instances, Classes
Reader Comments (Page 1 of 6)
Eggy Aug 2nd 2009 4:04PM
Just no,
Seaborn Aug 2nd 2009 6:34PM
This is Def. not accurate on my realm. I don't think I have had a warrior tank in 2 months.
Grubba Aug 3rd 2009 2:41AM
Okay, after having watched the comments here for the better part of a day and having read through the linked article (and its comments section) several times, I'm putting this comment here so it will be seen.
Shame on you, Ms. Robert. Not only do you clearly have no understanding of statistical information and how it can be used (which is somewhat excusable considering the number of people who don't), but you have actually included information in your article that is either speculative (which should be marked as such) or outright wrong.
"...he/she is pretty sure that the numbers are at least a ballpark representation of which class is most likely to be filling a particular role within a group."
Nothing like this exists in the article. Nothing. This idea is the core of your entire article, yet it does not exist in any form in the article that it's supposed to have come from. I don't know if that's just sloppy or outright deceptive, but I really find it distasteful considering what a charged topic class balance is already.
That the author's methodology is haphazard and extremely unscientific is bad enough, as it leads to drawing conclusions very likely to be inaccurate in the first place, but to insert conclusions of your own that the author doesn't attempt to make and then credit assertions that the author *never* makes is very poor form.
I think you're one of the better writers on this site generally, but I would call this work reprehensible.
Allison Robert Aug 3rd 2009 4:25AM
Warning; this gets long.
With deepest respect, Grubba, because I've seen you comment previously on the site, I think you're reading something into this that isn't actually here.
I am no great mathematical genius, but I took two years' worth of statistics. While I wouldn't do it professionally, I try not to draw erroneous conclusions from numbers, and data always says *something* -- just not necessarily what it purports to say. So this was my thought process when we got the tip about the article:
I'm primarily interested in the tank and healer data (as you may have noticed, I didn't spend a lot of time commenting on the DPS numbers). I think this data is likely to be at least slightly more accurate because players rarely spec into tanking or healing builds for the purpose of solo play; tanks and healers are inherently group-oriented specs (although, as I noted, the Feral Druid numbers are probably hazy here due to the heavy overlap between bear and cat specs).
What interested me most about the linked article is that the tank numbers seem to confirm something said very recently by GC concerning the massive population advantage still held by Warrior tanks. He also observed that there is a relative lack of Feral tanks in Ulduar, which could be problematic for Blizzard's effort to balance tanks because they face the prospect of having to nerf what is (direct quote) an "unpopular spec." VA's data in no way disputes this, and it just so happens to be within shouting distance of the last information we had on spec population courtesy of multiple Armory data sites. We knew, going into "Wrath," that Protection Paladins and Feral Druids had both lost "market share," as it were. We also knew that Protection Warriors became significantly more popular. The Death Knight, another plate hybrid who, like the Warrior, can DPS or tank but can't heal, is introduced in this period, is *enormously* popular, and in the months after "Wrath" goes live you start to see frequent complaints on the Tanking forums that there's a glut of tanks.
It doesn't take a rocket scientist to figure out that there is probably some pressure on Paladins and Druids in this period to respec to their healing trees, because: a). Warriors are still the most popular tank: b). DK's are, by universal acknowledgment (including that of the developers), the best hard mode tanks in Tier 7 and early Tier 8, and: c). Neither of them can heal.
Fast-forward to July/August 2009 where Blizzard is saying that the vast majority of people tanking today are still playing Protection Warriors and that the developers believe Feral Druids are under-represented in Ulduar. At this point, whether you believe VA's data is plausible or not (I'm not even going to touch the issue of accuracy for the moment -- I'm just going to deal with whether it makes SENSE) hinges entirely on your faith in what the developers are saying about tank representation. And, as I said below, I think it's pretty foolish to build an argument on the premise that the developers know less about class and spec representation than we do. Patently this is not true. VA's data dovetails into what *we* knew about spec popularity in late 2008/early 2009, fits the narrative on what happened to the tanking population in the transition from BC to "Wrath," and confirms something written within the last two weeks by GC.
Are the numbers accurate? We have no way of knowing.
But are they plausible? I would argue that, in light of what I've just written, they make sense and merit discussion.
Moreover, I would argue that my gloss on the linked article is an accurate means of summarizing the question the author was asking while building this data set. VA's first paragraph is ENTIRELY about the tanking situation I just described here! He/she writes that it's about "expected group composition" based on spec popularity, i.e. how likely your healer is to be (class) and your tank is to be (class), and so on and so forth, and that can certainly be paraphrased to "Which class gets invited as what?" To be frank, I'm confused as to why you think I behaved irresponsibly, and would welcome some clarification.
Beyond that, anything in an article that is accompanied by the statement that "I can't swear to the accuracy of the numbers" is inherently speculative in nature, and readers aren't stupid. If that statement isn't enough to tell them that the numbers may not come from Reality Land, most are already aware that WoW fan site numbers are, at best, educated guesses. This is not the word of God, handed down from on high in tablet form. It's a lot of numbers that may or may not say something important, many of them are consistent with recent statements by GC, and so consequently I think they merit discussion.
airman_dopey Aug 3rd 2009 6:16AM
Bam!
blacklobo Aug 7th 2009 12:58PM
Allison,
So being right for the wrong reasons is still right, is that what you are saying? I mean the logic in the original article was sketchy at best. Just because some of data seems to match up with what the developers have stated lately that means that all of it is correct? Just because a couple data points on a graph line up with something you have seen doesn't mean the graph is a perfect (or even good) representation of the facts.
If you want to have an editorial about class representation and/or balance have at it. However to base an article on flawed logic detracts from the point you might be trying to make.
LilBanshee Aug 3rd 2009 8:47AM
According to http://www.warcraftrealms.com/census.php which is one of the alleged sources of informaton for those statistics, 15% of all level 80s are death knights and 10% of all level 80s are warriors.
The chart claims there are 6 times more warrior tanks than death knight tanks. To try and justify that absurd statement, let's pretend for a moment that every single warrior in the game was a tank, so 10% of all level 80s are warrior tanks. Even with this extreme view, that would still mean only 1.66% of all players are death knight tanks while 13.33% of all players are DPS death knights.. Yeeeah, no. The dps versus tanking disparity among death knights is not THAT extreme. Especially since it is widely agreed that death knights are currently the best endgame tanks and are required for any hardcore raiding guild.
Oh, and you know what else I just noticed? Click that "class representation in tank, DPS, and healing roles" link and take a good hard look at those charts. Notice how only 9 classes show up in the dps chart? That's right, according to this information there are in fact ZERO dps warriors in the entire game!
Just remove this whole article, it's utter nonsense. Some guy ran a blatantly flawed analysis on some already nonrepresentative data. That happens in a hundred forums every day, please stop encouraging it. Make people defend their claims before you glorify them in an article.
Jon Do Aug 3rd 2009 9:02AM
In my own case, I my priest is dual spec. Any given day you're likely to find me specced as shadow - that is my solo / daily questing / questing spec.
However, I almost never run in groups as shadow (unless I'm running friends through an instance, or running lower-level content as DPS). I've never entered a raid as shadow spec.
I raid and run heroics as holy. I probably spend less than 10 hours a week in holy spec. I run with guildies and am almost never in LFG.
I have serious doubts that this methodology would correctly classify my priest.
LilBanshee Aug 3rd 2009 9:18AM
Okay I noticed farther down the reply chain that the alison notes warriors are shown as 4% of the dps meter and there's just a missing label saying so. Alright, let's give that due consideration (and resist the urge to dismiss my point outright due to a missing label throwing me off).
The table below the piecharts says 70.64% of all players were dps and 9.43% were tanks. 4% of the dps are warriors (2.83% overall), 19.77% of the dps are death knights (13.97% overall). 65.27% of the tanks are warriors (6.15% overall), 10.97% of the tanks are death knights (1.03% overall)
So...
2.83% DPS warriors + 6.15% tank warriors = 8.98% warriors overall
13.97% dps death knights + 1.03% tank death knights = 15.00% death knights overall
Therefore this article claims that only one out of every 15 death knights has a tanking spec. I'm sorry but no, that just defies reason. My most generous interpretation of this data is that the author of the analysis didn't know how to distinguish between dps and tanking specs for death knights.
Wyred Aug 3rd 2009 9:56AM
@John Do. In the original article it states that it takes each person's main-spec as their 'spec', and discounts their secondary spec. Doesn't matter which you log out in. Of course, this could still result in mis-representation, but without people actually saying 'I tank in raids' there's no real way around it.
Jon Do Aug 3rd 2009 11:26AM
After looking more closely at this, my conclusion is that this amounts to speculations represented as pie art, not actual data.
Unfortunately, as such it is likely misleading people into incorrect conclusions.
Naix Aug 3rd 2009 11:32AM
Hmm
Conflicting info here on wow.com. First Mike cries about the warrior class be unrepresented and now we see it jump to 65%?
Does the warrior class need a tissue from the other classes that have been nurfed to the ground over the last two years?
styopa Aug 3rd 2009 11:46AM
@ lilbanshee - your comments would be much more persuasive if they were supported by, well, ANYTHING.
"To try and justify that absurd statement..." Absurd, why? Because you see a lot of death knight tanks? How is your experience universal?
"...Even with this extreme view, that would still mean only 1.66% of all players are death knight tanks while 13.33% of all players are DPS death knights.. Yeeeah, no. The dps versus tanking disparity among death knights is not THAT extreme...." Again, according to...you? Sorry if I don't automagically assume you are infallible. I have 6 80s and another 3 toons that are 70+ across 4 servers, and have been playing since Eldre'Thalas opened...roughly WoW month 3. From *my* experience, I'd say there are currently somewhere at least 4x-5x as many DK DPS as tanks. (And, from my experience of DK tanks, more of them should stay dps, but that's just me...) Now, my experience is ALSO not ipso facto universal, but it's enough for me to categorically dispute that your "gut" isn't right, either.
".. Especially since it is widely agreed that death knights are currently the best endgame tanks and are required for any hardcore raiding guild..." This might be the core of your misunderstanding. If you're a hardcore raider, I'd suspect that you're going to see a higher proportion of DK tanks, for the reasons you state. But I'd submit that the % of people actually playing in endgame raiding guilds is quite tiny compared to the bulk of WoW's millions of players.
LilBanshee Aug 3rd 2009 12:30PM
One of the first steps in critical thinking is assessing the initial plausibility of a claim. Sure, there's a degree of opinion involved there, but that doesn't mean you turn off common sense altogether, especially when there are no supporting facts given for the opposing claim.
Claiming that two of every three warriors are tanks isn't that unreasonable, it's only a little bit off of an even distribution between their two possible roles. However, claiming that only one out of every fifteen death knights are tanks - that is disproportionate enough that it should raise a few eyebrows and not be blindly accepted without some supporting evidence.
Do I offer evidence against it? No, and I don't have to, the burden of proof falls on the person with the more outrageous claim. It's just as if I claimed a spaghetti monster ate ghostcrawler and has been making forum posts in his name; the burden of proof would fall on me, the person making the outrageous claim, not someone who challenges it.
Grubba Aug 3rd 2009 12:54PM
Allison, thank you for your response. I will give you a proper reply later today when I have time to give it the attention it needs.
Allison Robert Aug 3rd 2009 1:40PM
LilBanshee, the article isn't arguing that 2 out of 3 Warriors are tanks. It's saying that, if you join any random group, the tank you get is disproportionately likely to be a Warrior. Those are two different assertions.
Moreover, I don't doubt for a moment that the vast majority of Death Knights are still DPS. DK's may be overrepresented as tanks on hard-mode encounters, which is the likely result of their early strength in the age of cooldown-based tanking, but there really aren't *that* many hardcore raiding guilds out there; they are still a distinct minority among the player base. In the transition from BC to Wrath, all pure DPS classes lost population. Rogues and Warlocks in particular went from being extremely common classes at 70 to being among the least common at 80. Unless these people disappeared down a black hole somewhere, it was fairly obvious who was statistically likely to be playing a DK as a new main.
Now, while I also don't doubt for a moment that many pure-DPS rerolls to a DK did wind up tanking in some capacity, it does argue that most people playing a Death Knight as a main are used to playing the game as DPS and DPS only. This does argue that we should realistically expect to see more Death Knights DPSing than tanking. We have no idea if fansite numbers on this can be trusted, but the basic assumption is a sensible one.
LilBanshee Aug 3rd 2009 2:14PM
Reality check time. I just finished reverse engineering the data, and here is what the author based his charts on:
First, he went on the wow census site and determined these class breakdowns: 9% Warrior, 15% DK, 11% Druid, 9% Hunter, 9% Mage, 12% Pally, 9% Priest, 9% Rogue, 9% Shaman, and 8% Lock (these numbers don't match what show there today, so presumably they've shifted a little. What matters is that these are the numbers he used). This is a reputable site, and he simply pulled data from it with no bias, a perfectly legitimate action.
Second, he dug through wowpopular and determined that 6/19 of warriors are dps, 13/19 of warriors are tanks, 27/29 of DKs are dps, 2/29 of DKs are tanks, 1/8 of druids are dps, 3/4 of druids are healers, 1/8 of druids are tanks, 17/19 of pallies are dps, 10/29 of pallies are healers, 2/29 of pallies are tanks, 13/23 of priests are healers, 10/23 of priests are dps, 5/9 of shamans are dps, 4/9 of shamans are healers, and obviously all other classes are all dps. These are the exact numbers he or she used, not fractional approximations.
The problem here? WoWPopular doesn't tell you what percentage of people fill which roles. It simply highlights the specific specs that are the most common. If you don't see the distinction there, then follow me here for a moment: let's pretend exactly 50% of death knights are tanks and the rest are dps (obviously that's not the number, I'm just pointing out how WoWPopular will distort the number). Let's say that tanks have a lot of flexibility in their talent choices, so there are 30 different tanking builds. Let's say dps death knights are more limited in their talent choices, and there are only 5 common dps specs. What will WoWPopular say about death knights in this scenario? It will have the top 5 specs be dps specs with 4-5 stars each because each of those specs is so much more popular than any individual tanking spec. Likewise the tanking specs will struggle to even hit 1-2 stars each because they each individually have such a small share of the population. All the author sees is the stars, so he makes up a system for converting those stars into percents by adding up the stars in the top 10 or so builds and finds 29 stars, only 2 of which belong to tanking builds. So, he or she wrongly concludes that the vast majority of death knights are dps. Complete and utter misinterpretation of the data.
LilBanshee Aug 3rd 2009 2:28PM
I just re-read that and the first couple sentences sounded a little harsh, I don't mean to belittle you or your article, I'm just saying that this "The Grinch" fellow performed a flawed analysis of the information and it should nto be cited as a legitimate source of information.
Jon Do Aug 3rd 2009 2:34PM
Agreed.
There is no reason that we should assume that because someone published a pie chart on the web, it is accurate or credible.
In fact, it is riddled with obvious errors (such as the DPS pie chart is missing a class - the same class that appears to be overrepresented in the tank pie chart).
And the pie charts lack any supporting data (generated from stars, not actual numbers), and makes ridiculous underlying assumptions such as ‘every level-80 toon plays the same amount of time and runs 5-man instances with the same frequency’, which is what these chart imply by ignoring those factors.
It’s utterly ridiculous.
Rob Aug 3rd 2009 2:59PM
Well, after looked at the data, I also think it's utter rubbish. While I am not a statistician by trade I do perform lots of stats in my day job (and have stats training). It was a nice attempt to look at the numbers, and it was appropriate and nice for Allison to link it to us, but there are so many problems with the methodology we can't make heads or tails of it. As Allison notes, the numbers make sense, but that's a far cry from being accurate. For one thing, we don't know what percentage of the time people actually play in that spec. For another, we don't know how active that character is. Thus, for any given PUG tank, if you put these two probabilites together, you can get a likelihood of a given class.
I think most people have DKs but most level 80 DKs are not tanks. Anyway until we can get these two probabilities and some other data its going to be extremely hard to say whats what. However given my experience, the dps to heals to tanks ratio is about 5 to 1for any given character. Thus take a random 80, they are much more likely to be dps than anything else, and even if they could tank or heal doesn't mean they do.