In mainland China, players prefer Horde over Alliance.

In terms of faction preference, we found a curious pattern. In the EU and the US, players prefer Alliance over Horde. And this trend is consistent for both men and women. In the CN, this pattern is flipped. Players, both men and women, prefer Horde over Alliance.

This regional difference isn’t simply due to the difference in age. In both CN and the EU, age and ratio of Horde characters is not correlated (r = .008 and .03 respectively). In the US, there is a negative correlation (r = -.10). Thus, younger players in the US are slightly more likely to play Horde. An analysis comparing region while controlling for age showed that the effect of region was still significant (p < .001).

CN players have half the number of active characters compared with US players.

CN players have fewer active characters than players in the US and the EU. In fact, CN players only have about half the number of active characters compared with US players. Across the regions, we also see that women have more active characters than men.

The correlation between age and number of characters is not significant in CN (r = .02) or the EU (r = .04), but it is significant in the US (r = .16, p < .001). This means that older players have more characters in the US.

In an earlier blog post, we saw that CN players report playing more hours per week than US and EU players. In combination with having fewer characters, this implies a more focused play style—more time spent developing fewer characters. We’ll continue to explore this thread in future blog posts.

Overall, we collected data from 1,795 WoW players—640 from mainland China (CN), 279 from the EU, and 876 from the US. The gender distribution varied from region to region. The US had the highest female ratio at 39%. The EU had a female ratio of 29%. And CN has a female ratio of 15%. Incidentally, the female ratio for the US seems to have been slowly increasing over the years. In our Phase I data from early 2010, the ratio was 32%. And back in 2005 from the Daedalus Project data, it was 16%.

The average mainland Chinese WoW player is more than 10 years younger than the average US WoW player. In our sample, only 2% of mainland Chinese players were over the age of 30.

Due to human subject restrictions, we were unable to gather data from minors in this study, but the age distribution curves strongly suggest that the main bump for all 3 regions comes after the age of 18, and  the overall distribution of the curve is still very noticeable. There were some striking differences between the regions. In terms of averages, the average age in the US was 34.2 (SD = 10.7). For the EU, it was 32.6 (SD = 8.6). And for CN, it was 22.3 (SD = 3.6). Thus, the average WoW player in CN appears to be more than 10 years younger than the average WoW player in the US. The variance in age in CN is also much smaller than that in the US and the EU. In CN, there is a rapid drop-off after age 30—only 2% of our CN sample was over the age of 30, whereas in the US, there is a very gradual slope through age 60. In fact, of the 30 participants in the study sample over the age of 57, only 1 is not from the US.

These age differences are consistent with those we found in the Phase I data from Hong Kong (HK) and Taiwan (TW).

While the data points from HK, TW, and CN are consistent, what remain unclear are the factors that drive this age difference. These factors are likely intertwined and relate to perceptions of video gamers and the place of video games in adult life, but as of now, we do not have the cultural insight to unravel and understand these factors in any detail.

In terms of self-reported hours played per week, the average was 22.5 (SD = 18.3) in CN, 18.9 (SD = 12.8) in the EU, and 18.1 (SD = 11.5) in the US. Here and in other variables, the EU data tends to track fairly closely to the US data. In a sense, this is comforting in that it shows the smaller EU sample is not wildly skewed and seems to parallel a larger sample. These hours played per week differences are also consistent with data from HK and TW, where we saw that HK+TW players spend more hours playing per week than US players.

In the survey, we also asked about marital, child, and occupational status, as well as whether participants usually play at home or an internet café. Many of the differences in these variables are natural consequences of the age differences, but they help us think about how these community demographic differences may lead to differences in game-play norms which we’ll see later in the behavioral data.

97% of mainland Chinese players do not have children. 46% are full-time students.

Given how much younger the CN participants are, it is not a surprise that 87% of the CN participants are single, whereas this was true of only 47% of the EU participants and 42% of the US participants. 97% of CN participants do not have children, compared with 75% of EU participants and 69% of US participants. And finally, 46% of CN participants are full-time students, compared with 13% and 14% of EU and US participants respectively. Again, while these stats are somewhat natural consequences of the large age difference, what bears pointing out is how strongly these factors may change the local player cultures. In the US, players often talk about or mention their spouses or children during game-play (usually preceding an AFK or GTG), and family obligations constrain raiding opportunities, which then lead to different guild dynamics and practices. But imagine how different things would be in a player community where 87% of players are not married and only 3% have children.

For the average mainland Chinese WoW player, 70% of players they encounter are within a 4 year age difference.

It also bears pointing out that CN players are in a community where the majority of other players (about 70%) are within a 4 year difference of their own age (estimated from the standard deviation of CN age). In the US, the standard deviation is almost 11 years. Together, these community demographic factors are likely to have a strong impact on player norms. In the US, we often joke and talk about how difficult it can be to be in a group or guild where high school students, middle-age professionals, and retirees have to work together. In CN, WoW players are highly likely to be playing with someone very similar to their own demographic background.

Finally, we also asked whether players tend to play at home or in internet cafes. In the US and EU, 98% of participants only play at home. In CN, 9% always play in internet cafes and 8% play in both internet cafes and at home.

The WoW Armory contains over 4,500 behavioral variables for each active character.

But first, here’s a brief overview of how we collected the data. We began with a web survey to recruit participants. The study was publicized on web forums in the different regions along with popular international sites. The official WoW forums and WoW Insider are two such locations. The web survey, apart from gathering self-report demographic and personality data, also asked participants to list their active characters (up to 6). Once we receive the survey data, we have two automated data collection tools that run in the backend. One is an in-game census bot that runs /who commands on all the characters in our study database. If the character is online, we retrieve their zone tag and guild tag. We also track which of their guild mates is also currently online. From this data, we can generate social networks of characters and participants. The second data collection tool is an Armory scraper. The Armory contains over 4,500 variables for each active character and is updated on a daily basis. Our scraper retrieves the most recent Armory data for each character (if the Armory data has been updated).

Being able to link the survey data with the server data allows us to really understand play patterns, and especially how they vary or are consistent between age groups or geographic regions. Of course, the sheer volume of Armory data can make analysis difficult. As much as possible, we extract and derive conceptually meaningful variables from the data set to help us understand what is going on. In many of these cases, we have to figure out a way to normalize variables so we can remove variations due to time played or number of characters. In these blog posts, we’ll always start by listing the data source and in particular the sample sizes and how we sliced and normalized the data. Oftentimes, we’ll present multiple perspectives of the data to show how a particular trend is supported from different points of view.

We’ll cross-reference other blog posts to make overall trends more salient.

Given the range and scope of the data, it will often be difficult at first to understand the broader patterns. We’ll start with the basic demographic variables before delving deeper and deeper into the behavior variables. We’re aiming to put 1 or 2 blog posts each week. As much as possible, we’ll call out recurring patterns and cross-reference other blog posts to make these trends more salient to readers. To make the core findings easier to skim, we’ll use indented call-outs in the articles to highlight these findings. And as we mentioned before, we’ve come across many findings that we don’t have good stories for and your thoughts are very welcome.

To be a part of our newest research project only takes about 15 minutes to complete a web survey. We will then use an automated script to collect your characters’ data from the Armory. Last year, we collected data from the US, Hong Kong and Taiwan. This time around, we’re expanding recruitment to WoW players in the EU as well.

To participate in this study, you must be:

• An active WoW player
• Age 18 or above
• Playing on a US or EU WoW server

Follow this link to begin: http://wow.parc.com/survey/start.php

The Armory keeps track of the guild a character belongs to. Over time, we can therefore count the total number of different guilds a character has been observed in. For this analysis, we looked at the maximum among a player’s characters (i.e., one data point per participant). The chart below shows that, over a 5-month period, roughly 50% did not change guilds, 25% changed once, and the remaining 25% changed more than once.

Across both regions, men were more likely to switch guilds than women. Consistent with the days played analysis, we find that women are more loyal and engaged to both the game as a whole and to social organizations.

And across both regions, younger players are more likely to change guilds (r > -.14, p < .01 for both).

The Armory updates daily and only if the character was active the day before, so tabulating Armory updates over time provides an estimate of player engagement. For this analysis, we looked over a 5-month period and tabulated the total number of unique days a player played WoW. Thus, this metric ranges from 0 to 150. In our participant pool, the spread was surprisingly even. The average player played 65 out of 150 days, around 43% of available days. Older players played more than younger players in the US (r = .10, p = .02), but no significant correlation in HK+TW.

Interestingly, across both regions, we found that women played more days than men (p < .002).

We also looked more directly at hours played. Here, we turned to our in-game census bot that has an interval of about an hour. We found the same results. Women played more with this finer-grained measure than men (342.7 vs. 282.8, p = .01). This finding is consistent with a study of EQ2 that found that women play more hours than men. Contrary to the stereotype of women as “casual” players, they actually spend more time playing the game than men. Our earlier post on differences in achievement categories also highlight how men and women spend their time differently.

The Armory keeps track of transportation options. One of these is the number of flight paths ever used by each character. We took a look into this to see what we could find. Given that the use of flight paths is largely level independent (i.e., any level character can use a flight path), we tabulated the sum of flight paths across all of a participant’s characters. There was a significant difference between regions (p < .001). US players used more flight paths than HK+TW players. The gender difference was approaching significance (p = .08), but substantively small even if it were significant.

Age was positively correlated with flight paths taken in both the US (r = .21, p < .001) and HK+TW (r = .20, p < .001). So this means that older players use more flight paths. In terms of game-play motivations, flight paths taken was positively correlated with the Social motivation for both the US (r = .09, p = .05) and HK+TW (r = .15, p = .001). It wasn’t significantly correlated with any other motivation. So players who enjoy chatting/socializing/teamwork are more likely to take flight paths.

To drill into this deeper, I looked through the correlations between flight paths taken and the achievement ratios. For both US and HK+TW, flight paths taken was strongly positively correlated with World Event achievement ratios (r = .33 for HK+TW, r = .28 for US, p < .001 for both). Flight paths taken was negatively correlated in both regions with other achievements. This suggests that players who have taken a lot of flight paths tend to be older, social players who enjoy flying over the world and collecting quest items for and completing World Events. In hindsight, this also helps explain the regional difference. HK+TW players are younger and more combat-oriented than US players.