Guilds, Raiding and the Retention of the Casual Gamer Majority in MMOs

Most massively multiplayer online games (MMOs) today incorporate some form of social networking system into their gameplay, whether through social guilds, raiding, or group-based player versus player combat (PvP). There are a number of incentives to join these social networking organizations as membership allows the player greater access to resources and allies than solo play, and enables players to socialize and make friends with other like-minded individuals.

Players tend to join groups which have gameplay goals that align with their own, and effectively tailor their gameplay experience to their preferences. Despite this tailoring, the majority of high-end content in MMOs today seems to be geared mostly towards a specific portion of the player base, leaving others with indeterminate goals and thus less incentive to continue playing.

Undoubtedly, MMOs have a significant social component, and research has suggested that a player’s involvement or lack of involvement in social networks (whether formally through in-game organizations, or otherwise) is a predictor of character abandonment (Ducheneaut, Yee, Nickell & Moore, 2006). The reasoning behind this follows that a player who has made friends in the game world will be more hesitant to quit for fear of alienation. This is especially true for players who are involved in player run organizations such as guilds or corporations, as members of these organizations often come to be responsible for a number of specific roles and tasks. Thus, a player who quits may consider themselves not only to be alienating their friends, but reneging on a social responsibility as well. A player who holds membership in an in-game social organization has more to lose by quitting than a player who has few ties and no responsibility to others.

In Blizzard’s World of Warcraft it was found that lack of guild membership was a significant predictor of character abandonment (Ducheneaut, Yee, Nickell & Moore, 2006). Given that it has been stated that as a few as 40% of new subscribers play MMOs for more than 2 months (Mulligan & Patrovsky, 2003, cited in Ducheneaut, Yee, Nickell & Moore, 2006), it is in the interest of the developer not only to promote early player membership in guilds, but to foster the growth and continuation of pre-existing guilds.

Phases of Gameplay

In World of Warcraft there is a distinct social shift in gameplay style when a player reaches the maximum level, which divides gameplay into two distinct phases. The first phase involves embarking upon quests in small groups and slaying monsters in order to gather experience which in turn causes the character to increase in level and power. When the character reaches the maximum level (80, as of writing), or the ‘endgame’ phase, the player must then engage in high-end raid events and PvP in order to gain new equipment to increase their character’s power. Often these events require a much larger group, up to as many as 40 players and a high degree of in-group organization and coordination.

As of 2006, the majority of high-end content in World of Warcraft was geared primarily towards large and influential raiding and PvP guilds, which constitute a minority of the game’s population. Research implies that roughly 30-35% (varying by server type) of characters in guilds are members of smaller organizations with 60 or less members (Williams, Ducheneaut, Xiong, Zhang, Yee & Nickel, 2006) with an average guild size ranging from 11.26 (on RP servers) to 17.29 (on PvE servers).

A guild of exactly 60 characters would need approximately 66% of it’s members  (assuming one character per player) to be online at the same time for a period of several hours, which would be a very rare occurrence even for a scheduled event. Furthermore, endgame raids often require a sizable portion of time to complete, which makes it very difficult for the majority of smaller guilds, which may be composed of casual gamers, to participate in.

It was found after a month of observation that only 3.6% of players spent more than an hour playing high-end raid content (Ducheneaut, Yee, Nickell & Moore, 2006). Since the majority of endgame gameplay is specifically crafted to favor this 3.6%, this is a highly inefficient use of the developer’s resources. Favoring this minority comes at a risk of alienating the larger group of casual gamers in smaller guilds. Left to their own devices, it leaves them with no distinctive or realistically achievable goal thus forcing them to come up with their own reasons to continue playing. Eventually such players may start alternate characters to regain a lost sense of purpose in the game, or even become fed up and leave.

Lessons and Suggestions

While trends specific to World of Warcraft in 2006 cannot necessarily be generalized to other games, issues of player retention, social networking and endgame play are concepts which are relevant to all MMOs. Ducheneaut, Yee, Nickell and Moore write that the aforementioned 3.6% of players who participate in high end content are likely highly vocal hardcore gamers, and conclude that World of Warcraft, in the end, is not as casual a game as it claims to be.

The fact that the high end content was capitalized by a minority of the player population suggests a need for a different means of progressing at endgame other than raiding or PvP.  CCP’s EVE, for example, introduces a system of progression which is based upon learned skills, not level. Skills are learned based on time and may have prerequisites. The skills are so numerous that players cannot hope to possibly learn all of them, and must therefore decide what to specialize in. A character may excel in leadership and commerce, but be completely inept when it comes to combat. No matter how skilled a character may be, there is always room for improvement. This results in a more fluid definition of ‘endgame,’ which changes depending on updates, in-game politics and what the player desires.

Similarly, in the original incarnation of Sony Online Entertainment’s Star Wars Galaxies, players could choose from a number of professions, some of which were entirely based upon social interaction. Most exemplar of these were the entertainer classes, who healed wounds of the psyche, and the politician class which allowed for the administration of player owned cities. Other classes, such as the architect, engineer, tailor, chef and trader also allowed for similar social interaction. Endgame in the original Star Wars Galaxies came when a player ran out of skill points with which to develop their character. At that point the character could no longer gain further skills, and depending on their profession, was free to participate in raids, factional PvP, city building, or whatever their vocation allowed them.

None of this is to say that EVE and Star Wars Galaxies are or are not plagued by the same social and endgame problems that World of Warcraft experiences, but they are examples of alternative styles of play which directly reward social interaction outside of the primary game mechanic of slaying enemies. Considering that some 60% of participants of a 2006 study considered their guild to be primarily social (Williams, Ducheneaut, Xiong, Zhang, Yee & Nickel, 2006) this is an important consideration.

Since 2006 Blizzard has significantly expanded the endgame phase of World of Warcraft. As of current writing raid groups in World of Warcraft range from 5-25 people, and can be challenged at different levels of difficulty, which increases accessability to the game’s larger population of casual gamers. Unfortunately there is a significant lack of contemporary empirical research on massively multiplayer games with respect to gameplay issues, and so it is difficult to authoritatively say what other effects these changes have created.

The difficulty of designing a game which has something to offer both to casual and hardcore gamers is not trivial. Any changes which cater to casual gamers may inadvertently upset hardcore gamers and visa versa. In the end hardcore gamers will always have the advantage over casual gamers, as they are generally able to devote more time to playing games, and will in turn have access to a greater share of in-game resources than their casual counterparts.

Overall the most important lesson is that while it is common practice to appease power gamers by implementing large scale PvP and raids, these events should not be developed to the exclusion of content for smaller scale organizations and casual gamers. The more players developers can keep entertained, the less they have to worry about acquiring new ones.


Ducheneaut, N, Yee, N, Nickell, E, and Moore, J, R. (2006). Building an MMO with mass appeal: A look at gameplay in World of Warcraft. Games and Culture, 1(4), 281-317.

Williams, D, Ducheneaut, N, Xiong, L, Zhang, Y, Yee, N, Nickell, E. (2006). From tree house to barracks: The social life of guilds in World of Warcraft. Games and Culture, 1(4), 338-361.

Special Thanks

To pwfffff and ITJC68 on Slashdot for updating me on the current state of World of Warcraft’s endgame, and pointing out an error.

Using Personality Models to Make Better Games

Personality models are used today in a number of fields for a variety of tasks, from diagnosis in social work, to determination of an individual’s strengths in the workplace. This paper proposes that such tools could be used to help shape user experience in virtual worlds and suggests two methods to integrate these models in games.

In a 2008 article on the subjective well-being of online gamers it was suggested that gameplay could be tailored to the player to better accommodate his/her personality type (Chen, 2008). This would greatly increase the quality of each user’s subjective experience, and in effect increase both the longevity and value of the game. However, as of yet, there seems to be very little implementation of personality modeling systems in games. Rudimentary systems have been used in order to determine a user’s desired playing style, but they often are only used at the start of a game in order to suggest a character’s class, and have little or no further effect upon gameplay.

Much of the research around personality as it relates to playing games has been done using a commonly accepted five factor model (also known as the “Big 5”) which incorporates five different dimensions to describe a person’s disposition. These five dimensions include: Neuroticism, or how easily a person becomes distressed, Extraversion, which is a measure of how outgoing a person is, Openness to Experience, or how flexible a person is in their behavior, Agreeableness, which takes into account factors such as how trusting or how antagonistic a person is, and finally Conscientiousness, which measures a person’s dependability and organization (Costa & McCrae, 1992). Each of these dimensions are measured along a spectrum, which means that a conventional introvert would achieve a low score on openness and extraversion while a person who compulsively cleans might score higher in conscientiousness and neuroticism.

Suggested Integration

Personality types are most commonly determined by self-reporting via questionnaires. These questionnaires rely on the participant’s honest response to develop accurate results. This format makes sense in the context of social work or in increasing workplace efficiency, but not so much in an in-game environment.

The simplest implementation of an in-game personality modeling system would probably look something like the Generalized Occupational Aptitude Test (G.O.A.T.) in Fallout 3. In this example the player’s avatar is made to take an in-game aptitude test, however the test only serves to help the player choose skills for their avatar which would accommodate their playing style and has no further impact on gameplay. With some modification, it would be simple enough to have the results of such a test influence gameplay in later stages and would provide users with a more enriching experience. While this implementation is simple, it is hardly elegant and is fairly intrusive into the gameplay. In some cases, as in Fallout 3, it can be done while maintaining immersion, but in others this would prove very difficult if not impossible. It should also be noted that the current ‘lite’ versions of personality tests in games tend to be relatively short, whereas real psychological personality tests tend to be somewhat longer.

An alternate version of this implementation would take advantage of the ability of current generation console platforms to affect games via platforms such as the Playstation Network, Xbox Live or the Wii’s Dashboard. This approach would be similar to the first, in that the player would fill out a personality test, but the result would then be stored on the user’s profile, and would be usable in any game on that system that had the ability to take advantage of it.

A more elegant way of discerning a player’s personality type would require a system which observed the player’s behavior ingame, and then developed a model over time. This is a technique which undoubtedly is best suited for massively multiplayer sandbox games, but could also be applied to more linear and single player games with some creative design. Such a system could be implemented in many different ways, whether by creating quests which would probe at a player’s disposition, observing how sociable a player is via ingame statistical tools, or even observing different ingame social networks such as guilds/corporations/clans, etc. This method does have drawbacks, as it could be somewhat difficult to implement and brings into account issues of privacy. Each player should be made aware of such a system and be allowed to opt out of it if they so choose.

Once a model was established, it would then be used to shape individual gameplay towards a player’s preferences. In roleplaying games this could manifest in something as simple as rewarding extroverts for participating in social organizations, such as guilds or corporations, or even just assigning different missions or quests to a player depending on their personality type. More complex solutions could manifest in more subtle ways, and could be designed to organically reward players for playing the game in the way which they find most enjoyable.


An ingame system which tailored gameplay to each user’s personality could be immensely useful in subscription retention for MMOs, and in increasing the experiential value of regular single and multiplayer games. While such a system is most obviously useful for roleplaying games, it could also be implemented in a number of other genres with some creative game design. An acquired model could also be used for purposes not directly related to gameplay, such as intelligently targeted advertising or to help identify which users are most at risk for addiction in MMOs, but discussion of these uses are beyond the scope of this paper.


Chen, L. (2008). Subjective well-being: Evidence from the different personality traits of online game teenager players. CyberPsychology & Behavior, 11(5), 579-581.

Costa, P, & McCrae, R. (1992). Normal personality assessment in clinical practice: The NEO personality inventory. Psychological Assessment, 4(1), 5-13.