Second, it also tells us that the lifespan of patch 4.3’s content was less than six months for a lot of players. Around the release of patch 4.2, Blizzard’s CEO attributed churn to the lack of new content. Yet Blizzard did not release any new content for nine months after patch 4.3. In expansion 5.0, they released four patches with new content at intervals of roughly three months. Yet the number of subscribers still fell from 10 to 7.6 million between October 2012 and 2013. If lack of content is not to blame, what could be?
WoW’s lead game designer believes that players burn out because the game has trained them to complete quests as quickly as possible, rather than taking the time to explore the world [10].
It seems that MMOs need more than just new content to retain players. What could MMO designers do? A naive solution would be to make MMOs more challenging to CN players and easier to US players. Player behavior researchers have developed techniques to assess whether the difficulty of a game situation in a First-Person Shooter game is appropriate [9] to the player’s skill, or if a player’s character loses enough life during a boss encounter [2]. MMO designers could apply such techniques to better tailor raid encounters to their audience. However, tweaking the difficulty on a regional scale is out of question for games with worldwide competition such as guild and PvP rankings. This poses an interesting localization problem: can the same content be made easy enough to not frustrate US players, yet challenging enough to not bore CN players?
In a given month, the total player base consists of four categories of players: those who stay in, stay out, churn, and come back. Since most people stay in or stay out, presence is not a behavior relevant to predict. Rather, we built a simple regression model predicting the 10% of the total player base who churn every month. This model achieves very poor precision, but it may be improved by adding more in-game data that was not collected, such as the number of words typed by a player during the month, the profits made in the auction house, or real-life information such as a sudden change in lifestyle. Building a model to predict the 5% who come back every month seems even trickier because the features triggering the comeback are likely to be absent from the game. But it is an important question, since for two players who churn, one is coming back. This question may be more easily answered qualitatively.