Would it be too much trouble to post an "abandonment levels" graph that contains the actual figure you derived for the percentages quoted? This would be a much handier visual reference than just the two separate distribution graphs--it's very hard to see any by just looking at those.

I would suspect that the Shaman numbers may be more indicative of the fact that they're just a low population class overall. Even a marginal abandonment by some people who were "just trying out the class" can greatly affect the percentages when there's a rather small number there.

High shaman abandonment, especially when combined with low paladin abandonment, may simply reflect a larger pattern of people abandoning Horde.

Are you interpreting the data correctly? More low-level Shaman than high-level Shaman data points in your logs implies that a greater percentage of people are playing their new Shaman character now than the percent of those who originally started their characters as Shaman and since have gotten them to a higher level. This hardly sounds like class abandonment to me.

Couldn't you also argue that the class distribution in later levels represents what was popular at the time when those characters were started. And obviously the differences in the low level class shows the more recent trend shift.

I do agree that there are several other plausible explanations. Hopefully, over time as a we get more data points, we'll get a better sense of whether this is just a trend shift as Jhonus points out.

I would bet that this will turn out to be cyclical over time (meaning, classes will rotate in and out of the high and low spots in the graph), and that it probably represents a number of phenomena rather than the single one of class abandonment.

There is a major problem with labeling this as strictly abandonment. At least some segment of this must be represented by the development of alts. It's also critical to view this from the perspective of a recent game release versus a game that has been around for years. Right now, WoW has been around pretty much exactly long enough for a large segment of the dedicated player base (those players who are regularly spending a fair amount of time each week playing the game) to have at least one level 60 character.

Another extremely important point is the recent release of the battlegrounds. PvP endgame play (and I do not put that in quotes, because as much as developers may hate to admit it, it's endgame play) and PvP play period tends to enormously overwhelm class choice for players interested in that aspect of the game. Classes that were interesting for players when there was little PvP are often discarded in favor of those that are perceived as superior in PvP.

Noting this trend, in my opinion, actually has the potential to point out a far more interesting phenomenon; that of player perception affecting gameplay. It can become a very Schroedinger proposition to examine this. Developers build and balance classes on highly stringent and often limited criteria, the most salient and pervasive aspect of which depends wholly on a reasonably well-proportioned population across all classes. Balance is triangulated and highly rock-paper-scissors; each class has certain types of damage they are very vulnerable to, and vice versa. However, it is very rare for class populations to be nicely scattered. There is a clumping and groupthink phenomenon that dramatically affects gameplay.

If a player sees (or thinks he sees) more of one kind of class than others, he will become more interested in that class. Reading forums - also a time-to-release aspect; players usually play for a while before committing to the extent of reading forums regularly, excluding powergamers & diehards - then multiplies that effect tenfold because all gaming forums have the same advantage as statistics: you can always find one that supports your own personal opinion.

I think the data here is good, but largely because it points at underlying phenomena that I personally think is vastly under-studied.

On a completely unrelated note, I was originally visiting to see if you had any data on a., inventory slot capacity in various current MMOGs, and b., the possible effect of same on in-game economy. Specifically, I find WoW's approach to inventory to be both mystifying and inconvenient (very small capacity), and I was wondering about the broader ramifications of such a choice, because it has many. Particularly for new players.

Upon quick review, I realize that I should expand a little more on the point in my second paragraph: development of alts.

One of the most frequently-heard statements out of gamers' mouths is "I want to try a BLAH," or often "I like my BLAH but want to try a THING." This is usually reflected in the choice of alt class. The over-population of certain classes at high levels could actually be one of several things:

1. The high-percentage high-level characters in the second graph (representing classes-abandoned-for, presumably) could actually be OLDER than the the ones in the previous graph: players who are spending some time on alts could be bringing those characters back specifically for endgame PvP.

2. The difference in high level content for these classes - and playstyle - needs to be correlated here also. Particularly role of class and level of complexity of combat. To associate these percentages with the contained players' PvP flags would also be critical.

3. The opposite of #1 could be true, and nearly entirely support the original hypothesis presented; namely, the classes in the graphs could represent "second choices" - alt choices.

4. Changes in the game itself will invariably cause class spikes.

In terms of supporting the abandonment hypothesis, I think a long-term sample is critical. Data collected after a patch will be different than data collected two weeks before or after. Data collected after a major feature release - which almost certainly affects this particular data set - such as PvP battlegrounds will absolutely affect the numbers. Data also needs to reflect, in timeframe at least, if other data is not possible to collect in a clean manner - the development of alts versus mains.

I agree with all of Katie's points above: consideration of alt characters, groupthink, PvP endgame desirability, and patch timing are all influential.

For some firsthand data: I've played WoW for about 4 months, for a total of 14 days of gaming time, all on one character, now a level 52 Paladin. Thinking about trying a Warlock next. =)

Katie,

Very thought-provoking comments. At present, there are multiple interpretations of the data, and one of the reasons we are blogging these initial results is the hope that the research community as a whole can come up with ways to tease out the correct interpretation from the others. Even then, we will still be hampered by two major limitations: (1) We can only see the projected persona of the game character, not the player behind. We do not have the server side data to know that one player is switching between two alts. (2) Our data is from a limited time period. Hopefully, this will grow, but we are aware that the game character demographics are constantly changing, and as you point out, adapting to the changes in the game, both real and perceived.

As to inventory restrictions, I'm afraid there is nothing we can offer to help your research. As a player, I've felt the same inconvenience you have, and have wondered myself at the game utility of having inventory so severely restricted. "Hmm. I'd love to have this item just for RP purposes, but is it really worth an inventory slot?"

I believe the abandonment stems partly from two factors. In my case, being a paladin I'm now building a druid character to replace my main.

1) Drop rates on class items are too low. (29 scholomance runs without seeing my helm drop once makes the game feel like work rather than a game)
2) Paladins are too numerous and too many of them suck. I find myself fighting for a spot to do a raid all the time unless it is organized by my guild who knows I can play. A healer or warlock on the other hand almost always gets invited to high end stuff. I dont' want to spend 2 hours waiting on people, as soon as I get on I want to play.