On Faction Imbalance in Random Battleground Populations

The fantasy of Warcraft battlegrounds is that there are two relatively equal sides to the conflict. This isn’t just a fantasy that is pushed thematically, through lore and storytelling. This is an idea that is promulgated through the structure of random battlegrounds themselves, through the random queue mechanism that promises a similar experience to all players, no matter what faction their characters are.

However, this fantasy is false. It’s not false because of story or lore, but rather because of the interplay between three factors: 1) experience and gear providing advantages in PvP, 2) the random matchmaking mechanic itself, and 3) the separation of the pool of players into two teams.

Under the current system, faction imbalance in random battlegrounds is inevitable and leads to negative player experiences on both sides.

Let’s look at why.

THE IDEAL SYSTEM

The core idea of random matchmaking for games is: given a large enough pool of players and a large enough number of games, any given match will be equal. You should have approximately the same ratio of inexperienced to experienced players on each team, an equal distribution of gear. The goal is to allow individual performance to dictate the outcome.

In an ideal world, it looks like this:

The above picture represents the overall pool of players for random matches.

The first factor that we have to account for is that gear (on a character) and experience (of a player) both influence the success of a game. Gear is a fundamental aspect of World of Warcraft that affects a character’s ability, and in PvP it is acquired through experience. I’ve represented the combination of player experience and character gear through the relative size of the dots above – the larger the dot, the more influence that player can exert through a combination of experience and gear.

The random battleground mechanic is the next component to consider. At any given time, a group will be drawn from these pools of potential players.

When the system is at equilibrium, the queue times on both sides will be the same and the total area of all dots selected will be equal. Some flux is expected due to random selection of players, but over time the result should be solid queues and a 50/50 split in wins.

At this point we have the ideal state of random matchmaking.

Now let’s introduce faction into the population.

Now there are two pools. There’s limited fluidity between the pool – a player can choose to take their dot and go to the other side, either by rerolling (with a smaller dot) or faction changing (with the same size dot) – but that’s limited by the barriers of time (leveling) or money (faction change fees). So we’ll assume resistance to change within the pools unless there’s a reason to change.

Ideally, faction shouldn’t matter. But by splitting the source population in two, it creates a situation where not only is equilibrium impossible to achieve, it becomes something players rationally choose to avoid, creating bad experiences on both teams.

INTRODUCING IMBALANCE TO THE SYSTEM

The key to making the above system work is that the two pools that feed the teams in matches need to be equal. Any imbalance between the two affects the teams in a match, which in turn introduces a feedback loop into the population pools. Over time, small imbalances become magnified until the system stops working.

I’m going to assume just two things here.

1) The perception of an imbalance is more influential than the actual imbalance. Players only respond to imbalances that they perceive, through experience or communication with other players.

2) Experienced players are more likely to respond to imbalances than novice players. New players have not yet been exposed to the imbalance, and the chance that a player will respond to an imbalance increases over time.

Let’s start off with a simple case: the perception that one faction is better at PvP.

CASE 1: THEY’RE THE BETTER TEAM

It doesn’t matter how this idea originates. This could be through an early legitimate imbalance in a smaller population. It could be through a bad sample. It could be through some vocal members of the community repeating it. It doesn’t matter if the seed is real or not – all that matters is that players believe it.

If one faction is perceived to be better than the other, we should observe a slight shift of experienced players to that side.

Once this happens, a feedback loop starts with the matchmaking algorithm.

What might have started as a rumor now has evidence behind it, as the matches skew slightly in one team’s favor. As players grow in experience, they slowly react to this imbalance rationally.

The players on the losing team evaluate their performance over time and consider that maybe the other faction is better. Their experience is that they lose more than normal, that the other team just does a better job. The pressure to investigate the other side increases.

The players on the winning team feel far less pressure to change sides. Matches become progressively easier as more and more big dots join their pool. Why stop when you’re winning?

Queue times are also affected in this scenario. One side will have a larger pool of interested players than the other, resulting in long wait times for the perceived ‘better’ faction, and nearly instant queues for the ‘weaker’ faction. This amplifies the feedback loop and introduces the negative experience to the stronger faction.

Long wait times but higher chance to win, or fast queues for a probable defeat? Those are the choices you have when one faction is perceived to be better than the other.

CASE 2: THE STREAMS THAT FEED THE POOL

Let’s go back to our original state of equilibrium and look at a different variable – new players.

Player population is never static over time – people pick up and put down Warcraft all the time. A certain amount of churn – loss of experienced players – is expected within the pools as players either stop participating in PvP or stop playing Warcraft entirely. Churn is offset by new players joining the pool, either new subscribers or experienced players who are trying PvP for the first time.

So here’s another assumption: dots of all sizes can drop out of the pool, but only small dots join the pool.

Because of the gear structure and scaling present in instanced PvP, no player can start out as a really big dot. They might be a great player with experience in the class, but even with the best heroic raid gear the character will be undergeared for PvP. Best case: an experienced player with great PvE gear tries PVP and enters as a medium dot. That’s pretty rare, so we’ll assume only small dots enter the pool.

In a state of equilibrium each faction will have equivalent churn and growth rates, resulting in equal-sized pools. If one side churns or grows faster, imbalance will be introduced.

Let’s say that one side is slightly more popular than the other. Not popular in PvP, but overall more popular with the entire player base.

Over time, one pool will get bigger than the other – but not necessarily any better. The ratio of small dots to big dots is maintained over time as players improve and gear up. As long as the flow is consistent, equilibrium is not disturbed by overall faction imbalance. That’s great!

But what about when you have events which disrupt that flow?

The more popular side finds itself at a temporary disadvantage. More small dots entered their pool than the the other side, yet there hasn’t been any time for them to mature into big dots. The numerator of big dots hasn’t changed, while the denominator of little dots has gotten bigger.

Ironically, an influx of undergeared toons affects the more popular faction more adversely than the unpopular faction. Team quality declines on the popular faction, causing more losses. More losses means more churn as new and old players alike get frustrated.

Conversely, the unpopular faction weathers the influx better and their small dots grow and mature in a victorious environment. The popular side has a double whammy of initial frustration with their teammates followed by better-geared opponents.

And that brings us back to Case 1’s feedback loop.

I’ll just leave this here with a note that undergeared, inexperienced boosted level 90s are about the smallest dots you can represent on these graphs.

CASE 3: BETTER GEAR

World of Warcraft is a game of gear acquisition. Even with the experiments with uniform gear scaling in Mists of Pandaria (Challenge Modes, Proving Grounds), there’s no indication that random PvP will move to an entirely uniform set of gear across players.

So what happens when one side has an advantage in gear acquisition?

PvP gear is generally acquired through three methods: crafting, honor points, and conquest points. Everyone has access to crafted gear, honor gear is available through both random and non-random BGs, and conquest is available from random BGs and rated PvP play.

There’s no immediate faction advantage with the above gearing strategy, especially not from a state of equilibrium. If everyone starts off equal, with the same access to gearing opportunities, there won’t be a problem. But as soon as a problem is introduced, the gear system throws another wrench into the works.

The key is the rewards for winning a random battleground.

Over time, the faction which dominates the random BG queue will acquire more Honor and Conquest than the side which does not. Rated play is essentially factionless, as is crafting – so those two methods are effectively a wash. But control of the random BG reward allows that subset of players who don’t do much rated play to gear up faster than their opponents.

In the dot model I’m working with here, the really big dots get big at the same rate no matter what, but the small and medium dots grow into bigger dots at a faster rate, causing a feedback loop independent of faction changes.

In the North American servers, we see an additional layer of complexity to this problem. Alliance PvPers dominate only the two largest BGs – AV and IoC. Horde dominates the random queues. Alliance PvPers therefore queue specifically for those two BGs (and ONLY those two) so they can gain some honor with a victory, and those BGs reward a lot of honor anyways. The Horde is able to queue for random BGs specifically excluding those two maps, therefore ensuring both that they’ll both gain better gear faster and that the Alliance will continue to dominate those maps. The only reason to venture into AV or IoC as NA Horde is for achievements.

My understanding is that the situation is reversed in the EU, but ultimately it doesn’t matter which side is on top. Both sides suffer because of the feedback loops introduced by gearing strategies.

I should note that the current PvP gear system at endgame is an improvement over previous seasons and twink brackets, when individual items could enter imperfectly into a faction and tip the balance quickly. The level 85 twink bracket experienced an influx of 3 Lute of the Sun-Kings from the BMAH in 5.2, causing a dramatic increase in the relative power of the side which possessed them. This, in turn, caused the opposing team’s twinks to abandon the bracket, which destroyed the competitiveness of the bracket.

I mention the fate of that bracket only as a cautionary tale.

SOLVING THE PROBLEM

There are a few different ways to address the imbalances caused by faction in random BG matchmaking. Some work better than others.

1) Remove gear as a factor entirely. This leaves experience as the only determining factor between teams. It reduces the impact of new players joining, but doesn’t address the perceived imbalance between factions, which is the more pernicious long-term problem. Also it runs completely counter to the core idea of Warcraft.

Since this is offered as a solution a lot, I think it’s worth pointing out that while eliminating gear as a factor in PvP would fix some problems (mostly with class balance), it won’t address faction imbalance. Players will still gravitate to the perceived better faction.

2) Remove the bonus for random battleground wins. This has several drawbacks, most notably that it reduces the overall pool of players for BGs. It also only stops the gearing feedback loop (case 3), and doesn’t address anything about cases 1 or 2. Ineffective and possibly counterproductive.

3) Allow players to group cross-faction. I’ve been a strong advocate of this for rated play – let me do Arenas and RBGs with my opposite faction friends, already, no one cares about faction in rated play! – but for unrated random battlegrounds, it’s actually counterproductive to solving faction imbalance! This removes the obstacles for faction switching AND puts you in a premade group within a match, further skewing the results. Experienced players on the weaker side would just jump into groups on the stronger side, resulting in further domination.

(I still think this needs to happen for rated play, but that’s a different discussion.)

4) Eliminate factions entirely from the matchmaking algorithm. The implementation of this could include giving players the appearance of the opposite faction or not, but completely removing faction from random selection solves the problem completely. With no perceived faction advantage, players will no longer migrate. Queues become optimized and extremely fast. Best of all, matches become random again. You can enforce rules like role selection (X number of healers per side) and gear logic because your pool is doubled in size.

This is a massive paradigm shift and runs counter to the idea that Warcraft is a game of factional combat. Adding options like “queue as mercenary” help address this somewhat, but not completely, since the population will still be segregated into non-random sets.

5) Rig the system. Give the weaker side a behind-the-scenes buff to their abilities. Use real-time data to see what a faction’s overall performance has been and calculate buffs to tip the scales back to equal. Perhaps this is dynamic scaling instead of flat scaling – one side might scale up to 510 while the other scales up to 504.

This is hard to implement right. It adds in another level of variability and addresses some of the weaknesses of case 1, all of the issues of cases 2 and 3, but it’s a tremendous amount of effort to get in place and will require maintenance and constant tweaking. The best system is one that self-adjusts, but that requires time and development resources which could be spent on new content.

6) Bribe players on the weaker side. A CTA-style bribe bag (like is offered to tanks and healers to queue for heroics) doesn’t incent good players to queue on the weaker side – and the weaker side already has a surplus of players. Any solution needs to get the good players off of one side and onto the other in a completely random fashion, and bribe bags actively work against that.

My personal opinion is that the correct solution is to get rid of factions entirely from the random matchmaker. Every other solution keeps some type of imbalance which will inevitably cause a feedback loop to skew the balance one way or the other. Eliminating factions brings queue times back down while equalizing opportunity for victory. It solves the major complaints of both sides of the faction divide.

This solution will not be popular with many parts of the playerbase, entirely for thematic reasons. I get that. For the NA region it means Horde players win less often (but have shorter queues) and Alliance players win more (but give up dominance of the PvEvP battlegrounds.)

Making the game more fair isn’t going to make everyone happy – everyone loses something.

But until the mechanics of random battlegrounds change, dramatic faction imbalance is not just a possibility – it’s inevitable.

—-

If you like this kind of analysis and think you could use someone like me on your team, drop me a line. I am a kickass IT professional with an emotional need for thorough analysis, and I’m recently unemployed. My brain is for hire. 🙂

Class Distribution Data for Patch 5.4.2

 

 

Oh. Hey, how’s it going? Enjoying the end of Mists of Pandaria? Leveling a lot of alts, are we?

While Patch 5.4.2 wasn’t a major content patch, it marks a good point to take a look at the middle of this long end-of-expansion period and see how classes are doing in World of Warcraft. This data follows the same methodology of my previous census posts (previously, previously, previously, previously) in that it is snapshotting census data from  Realmpop and World of Wargraphs,  then analyzing that data over time.

All data is available in my Warcraft Class Distribution – Mists of Pandaria 5.4.2 Google spreadsheet. Previous versions are linked in the previous census posts.

This snapshot was collected on 1/8/14 and 1/13/13. I’m going to call it Patch 5.4.2 data even though the patch dropped on December 10th, 2013. (I got busy during the holidays, sorry.)

Go ahead and open up the spreadsheet in another window, and let’s get started.

OVERALL POPULATION TRENDS

 

The expected flow of an expansion is that players will level their main characters to endgame first, with alt characters following later on as time permits. The final content patch of an expansion allows many players to focus their attention on other characters after completing the final raid, resulting in people leveling alts. My theory is that we should see a more even distribution of classes as the expansion ages and players level those alts.

In general, I think we’re starting to see that. The 6 point disparity between Paladins/Druids and Rogues/Monks from the beginning of the expansion has lessened. Monks and Rogues have both gained a point, Paladins and Druids have lost a point, but everyone is starting coalesce around the 9.19% mean.

The one exception to this rule is Hunters; the Hunter class continues to steadily hold at 11.5%. My guess is that the Hunter class continues to be friendly to new players, Hunter mains, and folks leveling alts, so we aren’t seeing any drop off there. 

 

The chart above is the percentage of a class which has leveled to level 90, which gives us an insight into the overall health of a class through the leveling process as well as who people picked for their initial mains in Mists of Pandaria. Hybrids tend to level to 90 more frequently than Pure DPS classes, so we can evaluate them in two groups.

Hybrids (tanks, healers, DPS) tend to all be around 27-28.5%. Hybrids who can perform all three roles (Druids, Monks, Paladins) aren’t really that different from those who can perform only two roles (Death Knights, Priests, Shamans, Warlocks) in this respect, with one exception: Warriors. Warriors continue to be the least leveled-to-endgame hybrid class, and are the second least popular overall. Considering the class which overlaps them the most (Death Knights) is the second highest on this chart, that continues to be alarming.

Pure DPS tend to all be around 26%, with one exception: Rogues. Rogues are still the least leveled-to-endgame class, continuing the trends we saw in 5.4. There’s been the same slow improvement as before but it really seems to be a niche class for both mains and alts.

We’re now seeing a consistent trend throughout Mists – players are not choosing to play Warriors and Rogues at the endgame. If you look at the overall population figures you can see that Rogues are much more represented in the 10-85 pool – 8.85% – but then they fall off dramatically in the 86-90 range and never recover. Interestingly, Warriors do not have this drop in population – instead, they’re just an unpopular hybrid at all levels (compared to other hybrids). 

Without substantial changes to the class play style or a substantial carrot offered to players of these two classes, I don’t see any real substantial changes for Rogues or Warriors until the next expansion, Warlords of Draenor. 

DOING THE HARD STUFF

Class representation in Heroic PvE (6+ bosses) and Rated PvP (1800+ rating) is always a tricky thing. It’s all well and good to look at the classes to see what people are using when Doing the Hard Stuff, but the presence or absence of a class is not in and of itself a sign of a problem. Raid mechanics might favor one spec over another, PvP burst might favor one class over another, etc.. We have seen times when representation in hard mode Warcraft was absolutely fine, DPS was fine, and yet the class was seriously struggling (Warlocks in Cataclysm) – and we’ve also seen times when certain specs have been vastly underrepresented, yet people do amazing things with them. (Demonology Warlocks were abysmally represented in Rated PvP at the same time a Demolock won the World Arena title.)

So use this data cautiously. Think before you jump to any conclusions with it.

 

Here’s the total population vs endgame vs PVE vs PVP chart. The general idea behind this one is to point out areas where a class might be over or underrepresented in an area of the game relative to its overall endgame population. Paladins, for instance, are 11.1% of the endgame population and 11.7% of the heroic raiding population – relatively balanced. Priests, on the other hand, are 9.5% of the endgame population but 11.8% of heroic raiding and 12.8% of the ranked PvP data set. They are performing very well at the hard stuff.

General observations:

• Death Knights, Hunters, and Warriors are less represented than their general endgame population. For DKs and Hunters my guess is that they are excellent classes for casual players as well as experts, so they are overrepresented in the endgame pop and underrepresented in heroics. Warriors are a different matter. Mages struggle a little bit in heroics. Rogues struggle in PvP.
• Shaman and Druids seem to be doing very well in heroics. Warlocks seem to have a bit of an edge with mechanics in the current raid tier and have solid representation as well. Warriors and Priests are doing amazingly well in PvP.

The details for each class can be found on the PvE and PvP tabs. 

Digging into the specs (PvE, PvP) shows us some interesting tidbits, like Arms and Holy are doing great in PvP, Destruction Warlocks are dominating in SoO (due to raid mechanics, I assume?). But we also have a complete enough data set that we can start plotting Spec Popularity over Time, at least for heroics and rated PvP. 

POPULARITY OVER TIME

You snapshot data enough times and finally you can do some visualizations with it. I’ve divided the spec data from World of Wargraphs into 4 sections for each Hard Mode activity (PvE, PvP) and graphed them according to role and function within that activity. In theory, people familiar with each class should be able to pinpoint certain changes that were made either to the class or to the content that swung the herd into favoring certain classes over each other. 

 

 

 

 

 

 

 

As I look over these graphs, I can’t help but wonder about things like raids which favor absorb healers vs throughput healers, tiers which favor more tanks versus fewer tanks, encounters which reward cleave specs versus multi dot or single target specs. 

PvP is a different set of stories, presented in a similar fashion.

 

 

 

 

It’s amazing, looking at these graphs, to see the effects of certain changes in gameplay. Fix the bug in Spirit Shell, change how Vengeance works in PvP, and watch players switch in high end play. Arms is a roller coaster that is currently going off the tracks in PvP.  The BM/MM flip is startling to see. All of these changes are likely tied into changes in each patch. It’ll be neat to see if folks can correlate the data with the changes.

CAVEATS AND CONCLUSION

A final few caveats.

First, there appears to be a lot more spec data in here than in previous census posts, though really it’s the same data presented over time. I don’t think a popular class being popular in a hard activity is really all that unusual. A small class like Monks or Rogues are going to be a smaller piece of the hard mode pie. It’s when an unpopular spec is popular at an activity – or vice versa – that I think it’s worth standing up and taking notice. Take a look at the spreadsheet and see where there’s huge deltas between the general endgame population and a given activity – that’s where there’s something going on, something good or bad.

Second, take these numbers with a grain of salt. They represent the movement of the masses, and while they might help shed light in to why people are playing – or not playing – a class, they’re not a determiner of your own personal performance or enjoyment. If you’re not chasing heroic mode achievements or 2200+ ratings, the spec data shouldn’t make you feel bad about your choices. 

But there’s a flip side, as I found out in Cataclysm – sometimes it’s reassuring to know when you’re struggling with a class that it’s not just you. If you’re not feeling your Warrior right now, or are suddenly really digging your Shaman – well, other people are too. Really like your Priest? There’s a reason for it. You’re not alone, there are reasons why people are making these kinds of decisions en masse. 

I’ll leave it up to you to level those alts and figure out which ones you enjoy the most.

Class Distribution Data for Patch 5.4

World of Warcraft’s Patch 5.4 was released on September 10th, 2013. This is the latest entry in my series on population data snapshots. All data is contained in the Google Docs Spreadsheet for 5.4. The methodology is unchanged from the previous entries:

• Class Distribution Data in Patch 5.3 (spreadsheet)
• Class Distribution Data in Patch 5.2 (spreadsheet)
• The Decline and Fall of Warlocks in Cataclysm (spreadsheet)

As before, the patch dates are handy labels for each snapshot, but they mark the beginning of a period, not the end.

My hands are still in a lot of pain and I haven’t played in any real capacity since before 5.3, so my commentary is mostly about interesting facts in the numbers I find. I’ll leave it to current players to propose solutions.

CHARTS AND GRAPHS AND DATA, OH MY

First tab is relative class popularity over time at the endgame of a given patch or expansion. Some changes to note:

• Hunters are now the number one class at endgame, Paladins are now number two.
• Death Knights surpassed Druids for the number three spot.
• Warriors fell past Mages into seventh position. If we normalized their numbers for the addition of Monks they’d be worse off than at any point since Wrath.
• Shamans, Warlocks, Rogues and Monks all gained ground (but maintained relative positions.)

The next tab, Mists 5.4, is a data collection sheet, just like in 5.3.

Rogues are probably the most interesting story here, since they continue to have more toons at all levels than at 90. Last patch Ghostcrawler confirmed that people roll Rogues, they just don’t finish leveling them, and we continue to see that here. This probably won’t change until changes are made to how the class levels, which is typically an expansion revamp.

Monks, Death Knights, Paladins, Druids, Shamans: they all get leveled to 90. That seems to be a hybrid class trait. Pure DPS struggle more. Warriors struggle as well, but we’ll talk more about them later.

The data set shrunk slightly between June and September of this year – 31.1 million toons as opposed to 31.3 million toons. I think this is the effect of characters aging off of the Armory after several months of inactivity, but it could also be a relic of the RealmPop database getting cleaned out periodically.

Next tab is patch-by-patch breakdown of class populations in Mists. You can see the drop in the populations between 5.3 and 5.4, which might be part of the overall drop in subs – or it might not.

I want to zoom in on my new favorite chart, the Percent of Total Class at 90:

There we go. Nice and big. Notice anything there?

Hybrids are around 22-23%. Pure DPS classes have caught up a bit at 20-21%.

And then there are Warriors and Rogues.

The reason I like this data so much is because it functions like a health check – what are people bringing all the way to endgame? In the first patch, it was primarily mains, but now we’ve had a chance for people to level up their stables of alts. Hybrids are still a little easier to get to endgame than pures, though that gap is decreasing. The low percentages for Warriors (19.4%) and Rogues (15.4%) puts them at the absolute bottom of the pack. Warriors, as hybrid tank/melee DPS, should by rights be up at around 22%. Rogues should be around 20%, though even 19% would look okay at this point.

There are problems with those two classes. Everyone else looks pretty good.

DOING HARD STUFF

The next few tabs are, as before, the data sets from World of Wargraphs, which uses a different data collection methodology than RealmPop. The general population numbers match, however, so while there might be some statistical variation it’s slight.

This class looks at the relative popularity of classes within Heroic raiding (2+ bosses) and rated PvP (1800+ rating).

Priests, Pallys, Shamans, Druids, yep. I have no idea what’s gone on with raiding this tier so I’ll let other bloggers cut loose on this data set. Here’s the breakdown by specialization.

Ready for the PvP? Of course you are.

In case you were wondering, Priests are really, really overrepresented in PvP right now.

The Discipline Priest numbers are … way out of control. I don’t know what’s actually gone down this season, but they are probably going to get some heavy PvP nerfs. The last time we saw deltas like this was with Arms Warriors at the beginning of Mists, and they were nerfed into the ground.

THERE IS TOO MUCH, LET ME SUM UP

• Hunters: How’s it feel to go from one of the smallest classes in Wrath to the biggest class population in Mists?
• Paladins: It’s okay, you’re still doing just great.
• Rogues: Still hurting to get to 90.
• Warriors: Having a long, slow decline throughout this expansion. Rogues have the obvious problem; Warriors have a subtle one.
• Death Knights: No, you can’t change your name to “Way Better Than Warriors.” Be nice, former servants of Arthas. Suffer well, but suffer politely.
• Priests: Yowza.
• Druids: You still have instant flight.
• Shamans: I worried about Shamans in Wrath and Cataclysm, I’m not worried about them anymore.
• Monks: surprisingly healthy for a new class, need time to mature and catch up to other populations.
• Mages: Wait, you want me to say something nice about Mages here? REALLY?
• Warlocks: you should probably roll a Warlock alt, I hear they’re lots of fun.

That’s all I’ve got for now. Enjoy patch 5.4 for me!

Class Distribution Data for Patch 5.3

World of Warcraft’s Patch 5.3 was released on May 21st, 2013, so as with previous patches I’ve taken a snapshot of population data to provide historical analysis of class popularity trends. The methodology is unchanged from my 5.2 population analysis, so the same caveats and cautions apply here. The patch dates serve as handy labels for each snapshot, but it’s important to remember that they effectively mark the beginning of a period, not the end. The 5.2 snapshot represents data at the beginning of 5.2 – or the end of 5.1.

Like previous versions, I’ve constructed a separate spreadsheet for this patch’s data. There is one new sheet looking at the realmpop population data (instead of just popularity data) which is finally giving us a view into how players are leveling through Mists of Pandaria.

INTEREXPANSION VERSUS INTRAEXPANSION DATA

The data on the first tab is relative class popularity over time at the endgame of an expansion. There are 11 classes available in Mists, so median popularity sits at 9.09% (down from 10% from Wrath). That 9% isn’t so much a target as it is a line to keep in mind – just because Mages are currently ranked #7 with 9.15% doesn’t mean they’re unpopular – but neither are they wildly popular, either.

I think it’s important to note with this graph that it’s mixing apples and oranges a bit. The Wrath and Cataclysm data is from late in those respective expansions, when people had had time to level alts in addition to mains. Starting with the third snapshot (Mists Pre-Release) the environment shifts to those characters who received priority in leveling. Nearly a year out we’re starting to see player stables fill out a bit more, until eventually it will be an apples to apples comparison.

This intraexpansion data is interesting because I think it shows in very broad strokes a picture of which classes got leveled first, and which got leveled later. My hunch is that changes in popularity percentages here, now, are more likely to be from players leveling secondary characters/alts than from new players joining or huge flaws within a class. There are no doubt exceptions and rerolls – the Warrior spike in the 5.1 snapshot was almost definitely due to PvP dominance – but I think we are seeing more alt activity than before.

CONSUMING MISTS CONTENT

Above is the 5.3 snapshot for all levels, 86-90, and 90.

Quick comparison to the equivalent one for 5.2:

Monks have continued to take a little more of the overall pie, but the overall numbers are fairly steady. Rogues have improved a bit at level 90.

The popularity data is well and interesting, but I think there’s a more interesting way to look at the question of how players are engaging Pandaria – using population figures instead of popularity figures.

These population figures come from Realmpop, and are just the class totals for the US and EU regions added together. I’ve talked with Erorus a bit about how he gets his population sample and it reaches pretty deep. The important thing is that the methodology hasn’t changed over the course of the expansion, so by looking at the sample set of 17 million toons we can extrapolate out to general class trends. (I’d like it better if the data wasn’t missing Korea and China, but we have to work with what we have.)

Keep in mind this isn’t the total population of the game for these regions, but rather a consistent subset we can use for analysis.

I’ve been adding these population figures in since 5.1 – they’re in the hidden columns on the 5.3/5.2/5.1 base tabs – but I wanted to sit on them for a bit until we had enough data points to look at trends. So this is new, and it’s pretty neat! It’s on tab 3 in the spreadsheet.

The data is laid out in three blocks – all population, 86-89, and 90 – each showing population growth  by class across the expansion.

For All Levels, we can see a few interesting things about class growth and stagnancy. Over a period of time where subs have dropped by ~1.6 million subscribers, character populations have remained stable in the US/EU. Monks continue to be created, and I expect we’ll see them level off somewhere in the 2 million range by the beginning of the next expansion. Warlocks, Warriors, and Hunters have all seen some growth over the course of the expansion, while the rest of the classes are mostly stagnant.

Rogues are the only class to have a drop in figures between patches (5.2 to 5.3) and I’m not sure why.

Let’s move on to the Mists levels.

About twice as many characters are out of Cataclysm and into Mists of Pandaria right now than at the start of 5.1. Okay, it’s a little less than twice for almost everyone but Monks – 1.74x Paladins, 1.98x Warlocks, 2.56x Monks – but the population of characters in Mists content has basically doubled in 6 months.

Since the general population didn’t double, a lot of this movement has to be attributed to existing characters moving up from pre-Mists levels into Mists. How about at level 90, the endgame?

This is probably my favorite bit of data to look at. There are two components – the population increase over time, like we did before, and the percent of a class which is level 90. The first block is very similar to what we saw before at 86-90, though most classes have more than doubled their number of 90s in the sample. Druids are the furthest behind at 1.94x, everyone else has a factor of 2.1x or greater. There’s something going on there, but I’m not sure what. I think we need more data points, and someone who can evaluate how each class does in current content, to look at that – definitely not me.

The second block is amazing, though. I love it. It’s the number of 90s versus the total population of a class over time. With one exception, classes started out in a fairly predictable spread and have coalesced into one of two packs, either at 16-17% or 18-19%.  Monks started out amazingly strong – people who rolled Monks at the very beginning were motivated to get them to the endgame – but have come in line with Paladins, Death Knights, Priests, Shamans and Druids in the top pack. Mages, Hunters, and Warlocks are in a second pack a percent or two behind, which might be a result of their pure DPS roles? More incentive to level a hybrid alt, perhaps?

Warriors are stuck back with the pure DPS for reasons that I can only attribute to their being perceived as poor tanks in current Mists content. I’m speculating there, but I think it’s notable that they’re behaving like all the other hybrids.

And then there are Rogues.

At all levels, there are more Rogues than Monks, Warlocks or Shaman. There are almost as many Rogues as there are Priests! But Rogues are not making it to level 90. The 5.14% in 5.1 could be assumed to be the Rogue mains with a job to do, but even with a healthy influx of level 90s after that, they are not playing in the endgame. That 12.42% outlier result is amazing. It’s terrible, but it’s amazing.

Some of this might be due to Rogue populations swelling in late Cataclysm for the legendary daggers. A large number of leveling PvP rogues might also account for it? I’m sure that the Rogue community will have much greater insight than I over it.

But right now, Warriors are behind the other hybrids by a little, and Rogues are behind the other pure DPS classes by a lot.

ENDGAME ACTIVITIES

The last bit of data I want to talk about is the PvE/PvP representation data from World of Wargraphs. The basic idea behind this is that if a class is more or less represented in high-end play in one sphere or another compared to their general population trends, they may be over- or under-powered for that activity.

Some general observations:

• Monks: Very underrepresented in PvP. Brewmaster FCs seem to be the only decent spec?
• Hunters: Underrepresented in PvE, dramatically underrepresented in PvP.
• Priests: Mind-bogglingly good at high-end PvE and PvP.
• Paladins: Really strong in PvE (I assume from both Prot and Holy right now).
• Warriors: Have dropped from completely dominant in PvP to underrepresented in both PvP and PvE.
• Shaman, Rogues, Warlocks – doing pretty well, all things considered!

There are detail tabs for each spec in the spreadsheet if you want to dive into this data a bit further.

I’m not sure what to say about Hunters, to be honest. They’re a hugely popular class – more Hunters overall than Paladins, and soon to be more level 90 Hunters too – but they don’t have the same popularity in high end PvE or PvP. Is it that Hunters are fine in high end play, but simply have a larger population of casual players? Or are the problems with the class such that Hunters face problems in high end play above and beyond other classes?

I think there’s evidence to support both theories, and they are not mutually exclusive.

THE PROBLEM CHILDREN

Warriors were the PvP darlings of 5.1 – no more. They are struggling in both PvP and PvE and it’s showing.

Monks, as the new class on the block, have done a great job catching up to the population numbers of the other classes. They may never be as popular as the big 4 (Pally, Druid, Hunter, DK) but they are likely going to carve out a niche for themselves. They are leveling well, and the only real area of concern seems to be their PvP viability. As someone who’s been leveling a Monk, I expect to see some much-needed cleanup to the class in the next expansion to make it a smoother, more coherent experience.

Hunters are the dominant class in World of Warcraft, yet they are underrepresented in high-end play. There are more Hunters than Paladins at all levels, and soon Paladins will be ever so slightly less popular than Hunters at level 90. The question of Hunter’s role in endgame play is an important one – are these population discrepancies indicating real problems in high-level rated play, or is it due to the overwhelming popularity of the class? Is it due to their pure DPS role versus the popularity of hybrids? There’s a serious discussion that needs to happen there.

Rogues are more popular than they seem but are struggling to make it to the endgame. Those Rogues who make it to the endgame can do well, but so few of them do compared to everyone else that there’s something abnormal with them. Rogues are less likely to experience Pandaria than any other class, and that is worth investigating.

Update June 5, 2013: Svelte Kumquat @ The Red Hatted Rogue has taken this data and is running with it for Rogues. He’s pulled some additional data from Realmpop about our sneaky friends.