使用資訊科技方法研究虛擬世界-探勘多人線上遊戲中的玩家活動資訊

國

立

交

通

大

學

資訊科學與工程研究所

博

士

論

文

使用資訊科技方法研究虛擬世界 –

探勘多人線上遊戲中的玩家活動資訊

Using an Information Technology Approach to Investigate Virtual Worlds:

Mining Demographic Data in MMOGS

研 究 生：謝吉隆

指導教授：孫春在 教授

使用資訊科技方法研究虛擬世界–

探勘多人線上遊戲中的玩家活動資訊

Using an Information Technology Approach to Investigate Virtual Worlds:

Mining Demographic Data in MMOGS

研 究 生：謝吉隆 Student：Ji-Lung Hsieh

指導教授：孫春在 Advisor：Chuen-Tsai Sun

國 立 交 通 大 學

資 訊 科 學 與 工 程 研 究 所

博 士 論 文

A Dissertation

Submitted to Department of Computer Science College of Computer Science

National Chiao Tung University in partial Fulfillment of the Requirements

for the Degree of Doctor of Philosophy

in

Computer Science

June 2008

使用資訊科技方法研究虛擬世界–

探勘多人線上遊戲中的玩家活動資訊

學生 : 謝吉隆 指導教授 : 孫春在 國立交通大學資訊科學與工程研究所博士班 [email protected]

摘要

本論文的主旨在於提出一套資訊科技的方法，以觀察並分析虛擬世界（線上遊戲） 中玩家所產生出來的社會、群體與個人行為。虛擬世界是相對於實體世界而言，通常是 指電腦中介或網路中介的環境，也因其環境的關係，使得想觀察在虛擬世界中使用者活 動的研究者難以在其中進行研究。隨著個人電腦的普及、網路的發展與便利、以及線上 遊戲的盛行，線上遊戲世界幾乎成了虛擬世界的代名詞。在一開始，線上遊戲沈迷甚至 被認為是青少年在教育上與行為偏差上的重要因素；然而，隨著線上遊戲內容的發展， 國內外學者或遊戲開發者均意識到線上遊戲內容包含了人際交往、交易、合作、組織、 學習等可能性，因此線上遊戲成為除了家庭與工作場所外最重要的第三場所（The Third Space）[1]。然而，由於線上遊戲世界為網路與電腦中介的環境，並且使用者是透過自 己所創立角色，在一個遊戲公司所設計的遊戲主題下與其他使用者進行互動。因此，過 去的「線下」（或者說「遊戲外」）的研究方法，例如質化的訪談或量化的問卷法，難以 觀察並分析這些玩家在遊戲中的行為動態。而本論文的主要目的，則是利用現在線上服

務的 Web 2.0 趨勢，以資訊科學資訊擷取的取徑，直接取得玩家在線上遊戲世界的活動，

以克服在虛擬世界的環境下進行研究的困難。

Web 2.0 其中一個主要的現象為服務提供者除了在提供服務本身之外，尚提供一套

程式語言，讓電腦終端的使用者能自行開發片段程式以取得服務的內容，或依照使用者

自己本身的喜好來寫作、修改、下載、掛載其所需的插件。這些特色可以從以下目前在

網路上十分流行的 Web 2.0 服務中發現，例如 igoogle, Firefox Extention, 與 Yahoo!

Widget 等。這些現象符合了 Web 2.0 創建與分享的特徵。而在數位遊戲的開發上也有如

此的特徵，例如世記帝國（Age of Empires）與魔獸爭霸三（Warcraft III）等區域網路連

線性質的即時戰略遊戲均提供地圖編輯器給使用者自行設計地圖並與其他使用者分

享；而線上遊戲第二人生（Second Life）則自行開發一套腳本語言（Script Language）

提供使用者設計虛擬環境中的場景、建築、物品、和角色的衣著與動作；線上遊戲魔獸

世界（World of Warcraft）也允許使用者修改、寫作與載入遊戲的使用者操控介面。

以魔獸世界的使用者自製介面來說，其功能不僅能夠改變玩家操控介面的外觀，尚

能夠和 Web 2.0 服務的特徵一樣，透過服務提供者所開放的應用程式介面（Application

Programming Interface: API）來擷取在服務背後使用者在線上所產生的活動紀錄。而本

論文則利用這樣的特色來長期不間斷地紀錄玩家在線上遊戲中的活動以觀察玩家在遊

戲中的互動、成長與動態。和過去遊戲外或線下的方法比較上，以資訊方法進行玩家的

線上環境而難以研究的議題，例如玩家的分工、合作與溝通、玩家在遊戲內經驗的成長、 線上遊戲中組織的成長與消退、玩家參與遊戲內玩家自組組織的趨勢與行為、乃至於大 範圍跨文化的比較（例如在本論文中會提到的台灣與美國的玩家動機行為推論比較） 等。以線上遊戲資料擷取為方法，本論文內容包含了博士生涯的數項成果，主要為 1) 和 過去在社會學的組織區位學的研究成果相較下，以資訊蒐集進行量化研究更能夠掌握到 在遊戲中的組織變化與動態；2) 遊戲中的玩家會因為遊戲內容的影響與來自其他玩家 的壓力而決定加入或離開遊戲內的組織；3) 和台灣的玩家相較之下，美國的玩家更重 視在線上遊戲中與人互動的娛樂性。 關鍵字：人工智慧、資料探勘、線上遊戲、虛擬世界、數位遊戲、虛擬社群、虛擬組織

U

SING AN

I

NFORMATION

T

ECHNOLOGY

A

PPROACH TO

I

NVESTIGATE

V

IRTUAL

W

ORLDS

:

M

INING

D

EMOGRAPHIC

D

ATA IN

MMOG

S

Student : Ji-Lung Hsieh Advisor : Chuen-Tsai Sun Department of Computer Science

National Chiao Tung University [email protected]

Abstract

A growing number of researchers are looking into ways that online virtual and game

environments are affecting human activities, including communication, interpersonal

relationships, and community interactions. Whereas the first game researchers focused on the

potential for virtual world “addictions,” they are currently accepting the premise that virtual

space and online games are evolving into a collective “third space” in competition with family,

work, and school [1]. However, those researchers must deal with significant barriers based on

the nature of network environments, online games, and cyberspace. For example, the large

majority of users interact via their avatars and from divergent computer terminals, therefore

dynamic online and in-game behaviors are difficult to observe and analyze by conventional

off-game approaches (e.g., surveys or interviews) [2]. This dissertation will describe a method

As network services have evolved over the past two decades, a growing number of service

providers have started using easy-to-understand programming languages such as markup

language XML and script language Lua, and are now providing application programming

interfaces (APIs) to give users the power to refine and develop their own user interfaces (UIs).

These user-designed UIs are often executed as add-ons or plug-ins attached to main

applications. Examples include the Mac OS Dashboard, Yahoo! Widget, and Google Desktop

Sidebar. This flexible feature has also been adopted by the designers of World of Warcraft

(WoW), currently the world’s most popular Massive Multi-player Online Game (MMOG).

WoW players can design or modify their own UI add-ons to supplement in-game controls,

create guides and maps for solving missions, and collect in-game information on other players

or game environments. The third use will be the focus of this dissertation—specifically, taking

advantage of the UI feature to collect analyzable data on human behaviors in virtual space. The

resulting data can be used to perform quantitative longitudinal analyses, as opposed to

restricted qualitative analyses of data gathered via interviews and surveys aimed at specific

groups. In this dissertation I will highlight the advantages of the personalized UI feature for

investigating virtual worlds and give three examples of potential investigative uses: how

in-game communities grow and decline, how players join and leave guilds, and how Taiwanese

and American gaming cultures differ. As background I will discuss the characteristics of Web

between content providers and users as well as between game designers and players. My results

indicate that (a) compared with conventional research on organization ecology, the proposed

method is capable of capturing in-game guild evolution dynamics; (b) players usually leave

guilds or quit group play due to group mission pressure; and (c) compared with Taiwanese

players, American players put more emphasis on recreation.

Keywords: Artificial Intelligence, Data Mining, Massively Multiplayer Online

Acknowledgement

“Wow!” With the support from so many beloved people around me, my doctor

dissertation was finally born! Firstly, I would like to express my most sincere gratitude to Prof.

Chuen-Tsai Sun, my respectable advisor. He always respects my research interests and

provides possible directions for me while I am wandering. He provided me a lot of opportunity

to travel around the world for publishing my paper, shared many life experiences with me

beside my research, and trained me to know how to be a good teacher and leader.

I would also like to thank my friends in my Learning Sciences and Technology laboratory.

Chung-Yuan Huang always unselfishly provided his help and encourages me to do research

and publish my idea. Chia-Yin Cheng always helped me overcoming any problems as possible

as his can when I stayed in our laboratory. Finally, I would like to thank my parent, Lien-Fen

Kang and Chuen-Zong Hsieh, for your love and encouragement, I can express only an

inadequate acknowledgement of my appreciation. I dedicate this dissertation to you. And,

Jen-Yi, my sweetheart, I thank that you always accompanies with me, supports and warms

INDEX 

摘要 ... III  ABSTRACT ... VI  ACKNOWLEDGEMENT ... IX  INDEX ... X  FIGURES ... XII  TABLES ... XIV  1.  INTRODUCTION ... - 1 - 

1.1.  MASSIVE MULTIPLAYER ONLINE GAMES—VIRTUAL WORLDS ... -1- 

1.2.  APPROACHES TO INVESTIGATING VIRTUAL WORLDS ... -6- 

2.  BACKGROUND ... - 9 - 

2.1.  PLAYER ... -9- 

2.1.1.  Player modeling ... - 9 - 

2.1.2.  Power Leveling ... - 10 - 

2.2.  MMOGCOMMUNITIES ... -11- 

2.2.1.  Informal and Formal Groups ... - 14 - 

2.2.2.  Game Element for Forming Groups—Guild ... - 16 - 

2.2.3.  Game Content for Promoting Group Play—Raiding Missions ... - 20 - 

2.3.  WOW(WORLD OF WARCRAFT) AND CLIENT-DESIGNED USER INTERFACE ... -22- 

2.3.1.  WoW Success ... - 22 - 

2.3.2.  WoW Game World ... - 23 - 

3.  ENVIRONMENTS AND APPROACHES ... - 25 - 

3.1.  NEW TRENDS FOR PEER DESIGNERS TO INVESTIGATE DIGITAL WORLDS ... -25- 

3.1.1.  Open-Data Trends ... - 25 - 

3.1.2.  Client-designed User Interface ... - 27 - 

3.2.  WEB 2.0TRENDS IN GAME COMMUNITY ... -30- 

3.3.  INVESTIGATING WOWUSING THE CLIENT-DESIGNED UIFEATURE ... -33- 

4.  RESULTS ... - 36 - 

4.1.  DATA COVERAGE ... -36- 

4.2.  DYNAMICS AND EVOLUTION OF MMOGGUILDS AND OTHER ORGANIZATIONS ... -38- 

4.2.1.  Introduction ... - 38 - 

4.3.  DYNAMICS OF PLAYER LEAVING AND JOINING GUILDS ... -47- 

4.3.1.  Experiment and Results ... - 48 - 

4.4.  COMPARING PLAYER BEHAVIORS IN DIFFERENT CULTURES ... -52- 

4.4.1.  Introduction ... - 53 - 

4.4.2.  Experiment and Results ... - 57 - 

4.4.3.  Level Upgrading Efficiency ... - 60 - 

4.4.4.  Faithful Participation in Guild Missions ... - 62 - 

5.  PC GAME EXPERIENCE: EVALUATING PLAYER EXPERIENCE IN STARCRAFT BY REPLAY ANALYSIS ... - 65 - 

5.1.  INTRODUCTION ... -65- 

5.2.  REPLAY ANALYSIS ... -70- 

5.3.  LEARNING PLAYER STRATEGIES BY CASE-BASED REASONING ... -73- 

5.3.1.  Results – Calculating the percentage of strategies accurately predicted by the trained decision system - 79 -  6.  CONCLUSION AND DISCUSSION ... - 82 - 

7.  RELATED PUBLICATION LIST ... - 86 - 

FIGURES 

Figure 1. Second Life screenshots. All objects are user-created. Two buildings at the top of the screenshot are embassies, one for the Maldives and one for Sweden. Avatars in central section are involved in an online meeting. Second Life players have the power to design game actions, including sexual behavior. ... - 3 -  Figure 2. World of Warcraft Screenshot. Similar to most other MMOGs, WoW avatars use weapons to win experience points and to earn gold. ... - 4 -  Figure 3. Relationships and interactions among service providers, web services (virtual worlds or MMOGs), and researchers. ... - 27 -  Figure 4. Yahoo! Widgets. Users can design their own add-ons or insert add-ons made by third-party service providers and designers. ... - 29 -  Figure 5. iGoogle. Users have the power to add gadgets to the iGoogle panel. Since it is a webpage, it requires a browser. ... - 29 -  Figure 6. An original player-controlled UI provided by WoW. ... - 30 -  Figure 7. Modified UI created by third-party source. Client-designed UIs can be found and downloaded from player-sharing community websites. ... - 30 -  Figure 8. Screenshots of Warcraft III mods including two types of gameplay: tower defense and two-team combat. ... - 31 -  Figure 9. Screenshot of a WoW player-designed UI called AltasLoot, which provides

probability of loot dropping. ... - 32 -  Figure 10. Structure and hierarchy comparisons between physical and virtual worlds. ... - 35 -  Figure 11. Avatar level distribution for World of Warcraft in Taiwan. Data were collected on 02/10/2006 for 641,805 avatars on 60 servers. More than half of all avatars between levels 16 and 20 were guild members. ... - 41 -  Figure 12. Differences between new and mature servers. A mix of growth and decline is noted in 2b. ... - 42 -  Figure 13. Guild scale transformation. Array thickness represents transformation probability potential. Data were collected four times per day on average between February and April, 2006 from an average of 40 PvE and 20 PvP servers. ... - 43 -  Figure 14. Life cycles for six basic guild types. ... - 45 -  Figure 15. Guild life cycle after Blizzard Entertainment launched its official expansion pack in Taiwan (September-December, 2007). ... - 47 -  Figure 16. Player changing guild behaviors. The thickness of arrow stands for amount of players changing behaviors. ... - 51 -  Figure 17. Dynamics of player staying at original guild and following with their guild growth. ... - 51 - 

Figure 18. Dynamics of player leaving and joining guilds. ... - 52 -  Figure 19. Data collecting on avatars who leave original guilds but don’t join other guilds. ... - 52 - 

Figure 20. Average playing time for players on the Taiwanese and American WoW servers used in the study. ... - 58 -  Figure 21. Size distributions of the Taiwanese and American guilds used in the study. .... - 59 -  Figure 22. Distribution of level upgrading efficiency for players on both servers. Only avatars with levels 11 or higher were considered. ... - 62 -  Figure 23. Average weekly raiding time distribution for guilds on both WoW servers (single avatar guilds deleted). ... - 64 -  Figure 24. Example of possible building sequence in StarCraft. ... - 74 -  Figure 25. Building sequence representing 100 combat incidents between a “Terran” species player and a “Zerg” species player. Link thickness indicates number of times the same

strategy was used. Node ID number indicates building status. ... - 80 -  Figure 26. Predictive accuracy of the first 100 rounds for player using Terran, Zerg, and

Protoss species. After 20 rounds of training, predictive accuracy was generally higher than 50

percent. Graph indicates large variation in predictive accuracy for player using Zerg species due to large variation in player strategies. ... - 82 - 

TABLES 

Table 1. Categories of WoW communities according to definitions from Bartle (2003) and Taylor (2006). ... - 16 -  Table 2. Example of WoW avatar behavior data collected using the game’s player-designed user interface. ... - 34 -  Table 3. Datasets. Different dataset recording frequencies, periods, and scales were used for different purposes. ... - 38 -  Table 4. Data collected from Taiwanese WoW servers between February 10 and April 10, 2006, including total numbers of avatars, average avatar levels, percentages of level-60 avatars among all avatars on individual servers, total numbers of guilds, average guild size, and Alliance/Horde ratio for PvP and PvE servers. Data were collected four times per day from an average of 40 PvE and 20 PvP servers. ... - 39 -  Table 5. Descriptive comparison of Taiwanese and American WoW servers. ... - 58 -  Table 6. Comparison of guild-related characteristics for the Taiwanese and American WoW servers. ... - 59 -  Table 7. Data on level upgrading efficiency and playing time per day for players on the Taiwanese and American WoW servers used in the study. ... - 61 -  Table 8. Selected actions and corresponding descriptions. To preserve anonymity, players are named “Player1” and “Player2” ... - 72 -  Table 9. Example of current case and equivalent pass cases. The best choice should be case 1. ... - 78 -  Table 10. Number of building states and construction strategies for three species. Statistics were obtained before and after100 rounds of training from each species ... - 80 -  Table 11. System predictive accuracy. Statistics were obtained after training 100, 200, and 300 rounds of play for each species (Replays of each species belong to only one player). - 81 - 

1. I

NTRODUCTION

1.1. Massive Multiplayer Online Games—Virtual Worlds

Virtual worlds are computer-based simulation environments that allow users to interact

via avatars. They are often described as persistent, shared, and computer-moderated

environments in which multiple users interact with each other and make changes via individual

representations and movements in real time [3]. Due to increased time spent online and greater

complexity in online behaviors affecting physical world communication, interpersonal

relationships, and communities, virtual worlds are now recognized as real but not concrete [4].

Virtual worlds visually mimic complex physical spaces in which people can interact using

self-created avatars as alter egos [5]. Although they have other non-entertainment applications

(e.g., education, public policy making, business trades, advertising simulations) [6, 7], most

virtual worlds that are the targets of research are game-based [3], mostly text-based Multi-User

Dungeons (MUDs) and graph-based Massive Multiplayer Online Games (MMOGs). The

immense popularity of MMOGs is reflected in the number of subscribers in 2006: 14,000,000

and growing [8].

Launched in 2003, Second Life is a typical web-based virtual world (Figure 1). Since 2003

(http://secondlife.com/whatis/economystats.php). Compared to other MMOGs, Second Life

has no clear goals, missions, experience levels, or victories. The main activities are social

communication, participation in group activities, and producing/selling virtual objects and

services (e.g., virtual building construction, giving music lessons). This and similar games (e.g.,

ActiveWorlds, Kaneva, and Entropia Universe) blur the boundaries between virtual and

physical worlds—for example, the virtual “Linden Dollar” currency in Second Life can be

exchanged for real world currencies (265 Linden Dollars = $1US when this report was being

written), virtual Second Life objects have been sold on ebay, and at least one country (Sweden)

has created a virtual Second Life embassy to introduce its culture and to canvass visitors.

Another remarkable example of a virtual world is World of Warcraft (WoW) created by

Blizzard Entertainment (Figure 2). WoW is currently the world’s most successful MMOG, with

over 10 million players as of January 2008. WoW crosses regional and cultural boundaries,

providing services in North America, Europe, and East Asia. Main WoW activities include

visiting virtual “zones,” accepting and executing missions, upgrading avatar levels, buying,

selling and trading treasure and equipment, and joining guilds. Blizzard continues to develop

and distribute expansion packs and game content updates (especially high-end group missions)

as part of its major effort to retain players. It is also attracting considerable research attention in

Figure 1. Second Life screenshots. All objects are user-created. Two buildings at the top of the

screenshot are embassies, one for the Maldives and one for Sweden. Avatars in central section

are involved in an online meeting. Second Life players have the power to design game actions,

Figure 2. World of Warcraft Screenshot. Similar to most other MMOGs, WoW avatars use weapons to win experience points and to earn gold.

The dramatic increase in the number and quality of MMOGs since 1997 has impacted not

only the game market, but also multiple aspects of gaming society and culture [9]. The

popularity of gaming and the formation of MMOG online/offline communities have resulted in

the breaking down of boundaries formed by age, sex, race, and national origin. The rapid

increase in the number of online games has resulted in a steady stream of new products

developed by a growing number of commercial companies. Designers have increasingly

focused on detailed aspects of game societies consisting of player avatars, and the growing

number of increasingly sophisticated games is attracting exceptionally high numbers of new

players. The average age at which children begin to play video games has fallen at the same

As MMOG popularity has grown, players have become more immersed in virtual worlds

consisting of game situations and player communities. Individual players spend considerable

time, money, ideas, and emotional energy creating avatars and maintaining relationships with

each other. Avatars often evolve into identities that players adopt when trading game-based

equipment, knowledge, and virtual capital and property. In some cases game capital and

property take on value in the physical world, and game achievements, friendships, and

appearances take on emotional or personal value. Game-related transactions create second

lives for many players [12], leading to a description of online games as a collective “third

place” whose status competes with family, work, and school [1]. Compared to other types of

virtual worlds, MMOGs attract greater research attention because they cross boundaries

between countries and cultures, yet the location of servers in different countries allows for the

expression of unique cultural characteristics during play. Accordingly, they allow for studies of

hierarchies, societies, organizations, and individuals in both virtual and physical worlds.

MMOGs and online/offline social interactions are drawing considerable attention from

players, game developers, and researchers in the computer and social sciences [13]. Topics

attracting the greatest research interest include the effects of a player’s social context on

choosing games to play, common online and offline resources used to solve problems during

processes that form social norms in virtual societies, how players gain experience and expand

their social networks inside and outside the games they play [14], how players form informal

communities and highly structured organizations [15], and how players create or design new

game content [16]. Today’s players not only come together in the same virtual space to play,

but also form offline player communities. Today’s players frequent game websites and

discussion boards to share their experiences and to discuss game-related issues.

1.2. Approaches to Investigating Virtual Worlds

The large majority of users enter virtual worlds from divergent computer terminals,

thereby posing several considerable challenges to online game research. Most efforts to date

have entailed qualitative approaches (e.g., individual and group interviews, secondary data

study [17], participation in games by researchers who then reflect and report on their

experiences) and a handful of quantitative efforts (e.g., surveys[2]) to investigating player

motivations and activities. However, statistical analyses and experiments with small numbers

of research subjects generally fail to detect trends or the strong effects of variables [5].

Combined, these challenges and factors have resulted in deficiencies in understanding

individual player behaviors and the nature of interactions within online communities.

Furthermore, long-term MMOG trends in such areas as in-game organizational change and

increases in one avatar’s level, even when using longitudinal surveys and interviews. Choosing

survey samples that represent certain population segments is also difficult due to the potential

for respondents to misrepresent their age, gender, and other demographic characteristics.

In summary, the online virtual world properties of persistence, closed communities, being

based on the Internet, and multiple user properties pose five important challenges to

conventional quantitative or qualitative approaches to investigating player behaviors: (a)

interview and survey samples can be biased in terms of representability and generality; (b)

players find it difficult to describe what they are doing when they are immersed in games, with

different players offering different descriptions of the same avatar behaviors; (c) conventional

approaches are ineffective in providing cross-sectional and longitudinal observations at the

same time; (d) researchers currently cannot obtain in-game data without the permission of and

support from game companies; and (e) cross-cultural studies are particularly difficult to design

and execute for virtual worlds.

With the help of computer science and information technology, some researchers are

using digital video recorders [13] and computer monitor recording software to collect data on

in-game activities. In some cases, they are using computer monitor recording software to

record the online behaviors of avatars and digital video recorders to collect data on how players

monitor recording software to assess human-computer interface efficiency. These methods

demands long hours of replay viewing and analysis, and results are highly dependent on the

ability of researchers to observe consistent behaviors and to analyze motivations behind them.

Some researchers are beginning to design their own programs to collect online

information from such sources as Bulletin Board Systems (BBS) and transaction logs for

online auction websites such as Yahoo! Auction and ebay. Recent developments in markup

languages are allowing researchers to retrieve and integrate information from web pages,

although such efforts are limited to text-based online resources. For closed, non-text based

MMOG environments, users and researchers are limited to filtering network packets to retrieve

in-game information—for example, designing a “bot”1 that automatically kills monsters,

upgrades avatar level, and earns money. This method is somewhat inconvenient because it is

best done when the target network address can be obtained; it is less effective when a user’s

long-term behaviors are monitored by an onlooker. Another challenge to using bots is that

game designers are increasingly concerned about transactions involving virtual goods on ebay

and other auction websites, with most (as well as many players) arguing that the use of bots is

unfair to the spirit of online game playing. Accordingly, some game companies are starting to

encrypt network packets in an effort to block bot applications. While this may satisfy the

1

A “bot” is a third-party intelligent agent capable of automatically controlling avatars to slaughter monsters or non-player characters. Bots are capable of upgrading avatar experience levels when players are away from their computers. Most designers and players complain that bots ruin the fairness of game, and some game designers encrypt network packets to block bot activity.

desires of players, it serves as another challenge to researchers.

Due to recent developments in computer science, especially in the concept of open source

programming, more and more online service providers are releasing application programming

interfaces (APIs) that allow users to create applications or plug-ins to access and record online

information and activities; examples include Google desktop, MSN Messenger, and

Yahoo!Widget API development kits. Furthermore, some service providers are giving their

virtual world users access to online information. For example, several real time strategy (RTS)

games are now providing access to text-based user activity logs, which allow users to analyze

and understand their own playing strategies. Two games that offer this feature are Starcraft2

and World of Warcraft; WoW allows users to access and record information and activity logs for

their own and others’ avatars.

2. B

ACKGROUND

2.1. Player

2.1.1. Player modeling

Bartle [3] notes that player demographic data are important to game companies wanting

to investigate what kinds of players like specific MMOGs (in order to know where to look to

increase membership) and how MMOGs might be altered to attract new players and to retain

existing players. Academic researchers can take advantage of the same data to learn how

players control avatar behaviors, as well as to organize their analyses in terms of player type

(e.g., newbies, casual, hardcore) [18, 19] and motivation (e.g., seeking personal achievement

or social interaction) [20]. Data on avatar behavior can also be used to identify gold farmers

(players interested in earning real-world currency) and white-eyes (players who purposefully

try to disrupt games or otherwise break gameplay rules) [21]. Bartle [22] used motivation to

categorize MMOG players as achievers, socializers, explorers, and killers; these categories

have been the bases for a large number of discussions by other researchers. Note that Bartle

used online surveys aimed at gaming communities to gather the data for his categories. For

my research I used avatar demographic data for similar analytical purposes.

2.1.2. Power Leveling

Potential level upgrade speed is recognized by players as an important factor influencing

the popularity of online games, especially MMOGs. Game design complexity affects level

upgrade speed, which in turn influences how long a player continues visiting the same game.

WoW designers clearly focused on this important aspect of their product when creating a game

that allows players to learn the game environment and gain a sense of control very quickly.

player’s experience growth or flow status to adjust mission difficulty. For example, WoW

contains missions that allow newbies to reach level 10 in less than 5 hours. Players at

subsequent levels can take part in more complex actions such as small-group raids (for up to 20

avatars), joining guilds, or riding horses. A major difference between WoW and its predecessors

is the speed at which players can reach the highest avatar level (70)—an average of two months,

but possibly as fast as one week. At one time, players had to spend at least one year to reach the

top level, and many never succeeded. Upgrade speed is strongly influenced by player behaviors:

since players in groups are required to share experience points with other avatars, guild

membership and cooperation can slow down player advancement [1]. Players gain experience

points much more quickly by repeatedly killing the same monster at a fixed location without

guild support; those interested in rapidly advancing their avatar levels can take advantage of

this tactic.

2.2. MMOG Communities

The characteristics of communities formed by online game players is a sociological issue

receiving considerable attention from virtual world researchers [3]. Designers of the most

popular MMOGs recognize that a sense of community is an important factor for player

retention, and therefore often create several community levels, from casual teams and

its members organize themselves to obtain achievement or to simply have fun. Recognizing the

range of possibilities, developers have designed (a) multi-avatar missions promoting group

play, (b) friend lists for players to maintain private in-game connections, and (c) structured

guilds and group (raid) missions to promote formal communities. For their part, players are

usually creating independent web sites for discussing online game community issues and

sharing information.

In-game communities are recognized as centrally important to retaining players. The

ways in which players organize themselves and make decisions to join or leave game world

organizations are drawing attention from real-world sociologists and enterprise managers.

Good guild leaders in an MMOG are recognized as having potential for becoming good team

managers in the physical world [23]. For this reason, game communities are also receiving

attention from researchers of organization ecology [24, 25]. Whereas real world organizations

have clear goals that resist change due to members joining and leaving, virtual world

organizations have no clear initial goals and therefore suffer from frequent member

movement. McPherson [24] has proposed an ecological model of competition among social

organizations for members, defining ecological niches in a manner that brings together

organizational geography, time, member characteristics, and social composition. Successful

organizations grow and unsuccessful organizations either disappear or are absorbed into other

members, and their growth and disappearance may be viewed as compact examples of change

in real-world organizations.

In organizational ecology terms, a competitive relationship exists between organizations

that overlap temporally or geographically. The criterion used to measure or evaluate an

organization is called a niche, with every organization having its own niche dimensions (e.g.,

employee education level, age, occupational prestige, type of service). Since MMOGs cross

geographic, age, gender, education, and occupation boundaries, synchronicity defined as

in-game time overlap between different players is considered an important niche dimension.

In most MMOGs, an avatar can only join one guild, which adds to the competition among

guilds to recruit new members. From the player perspective, avatars join guilds to achieve

social, personal, and/or game goals. However, MMOG player motivations are more likely to

change over a short period of time compared to real-world member motivations, making

guild dynamics much more complex. Since surveys and interviews only capture single

moments, they are less likely to identify why players join and leave guilds and less efficient

in obtaining detailed data on guild categories frequently mentioned by players (e.g., elite,

leisure, and family guilds). In contrast, avatar demographic data can be used to investigate

niche dimensions such as organizational structure, synchronicity, and social network

2.2.1. Informal and Formal Groups

Due to the high-end content design found in recent MMOGs, players are showing a

growing tendency to form social structures and to play together to gain experience points by

completing complex missions [17, 26]. According to some researchers, it is not uncommon for

players to spend eight or more hours in one session playing with other members of their

communities [26]. Using WoW missions as an example, many are designed for small group

activities involving avatars at level 30. To complete these missions, players form temporary

groups, visit a few dungeons together, and coordinate their activities to defeat monsters. These

informal groups and casual missions often lead to larger groups consisting of both online and

offline friends, then to guilds marked by explicit rules and management structure, then to

temporary hybrid groups formed to solve the largest and most difficult missions [17]. To

support these groups, game developers are providing private communication channels for guild

Table 1 presents WoW community and/or group categories in terms of Bartle’s (2003) and

Taylor’s (2006) definitions and categories: formal or informal, temporary or permanent, flat or

Table 1. Categories of WoW communities according to definitions from Bartle (2003) and Taylor (2006).

Community Community in WoW Number of Players

formality Time- persistence

Structure Relationshi p

Pairs Players in WoW may pick up to play together and share their experience points 2~5 Informal but recorded by WoW Often temporary Flat Hard-wired or soft-wired Pickup groups

There are many missions designed for small group after avatars achieving level 30.

3~5 Informal temporary Flat Soft-wired

Friends Players can store other avatars to their friend list for contact.

Formal Permanent Flat Hard-wired

Groups for raiding

Players often follow the guild schedule to solve a sequential raiding mission. WoW raiding missions need high degree of coordination of players. Each raiding mission has its requirement on the number of participants and their levels.

5, 10, 20, 40.

Formal Temporary Flat Hard-wired

Guild and ally groups

Guild is a well-structured group in WoW with exclusive sign and name. Players manage their guilds, recruit new member, and plan to solve raiding mission by themselves.

10~480 Formal Permanent Hierarchical Hard-wired

Hybrid groups

Different guild’s members often cooperate to attend the

battleground playing in WoW.

40 Formal Temporary Hierarchical Hard-wired

2.2.2. Game Element for Forming Groups—Guild

Since the creation of Multi-User Domain games (MUDs, precursors of MMOGs), guilds

have attracted considerable attention from players, game developers, and researchers.

Originally viewed as self-emerging organizations formed among avatars belonging to different

players, guilds have evolved to become an important factor in game design following the

successful implementation of guild mechanisms in Ultima Online, EverQuest, and Lineage.

These mechanisms (which continue to be refined and improved) are now viewed as important

equipment. The high percentage of players joining guilds also attests to their essential position

in game worlds, as well as their suitability as a focus for research on MMOG development, the

behaviors of players who join guilds, and social tendencies in game playing. For long-term

users and power gamers, the importance of guilds can approach that of the game itself [14, 27].

Guild member avatars have exclusive guild names and flags, and game systems often set

aside conversation channels for intra-guild communications. Taking into account system and

design variation, guild functionality and capability tend to be MMOG-specific. For example,

guilds in Final Fantasy XI have special conversation channels, Lineage guilds have access to

treasuries for storing guild-owned equipment and gold, and WOW and EverQuest guilds are

given special roles that encourage long-term cooperation and coordination among members.

Different levels of purposefully designed tasks explicitly and implicitly guide players to

various self-organizing activities [28].

Initially, the decision to join a guild and the process of doing so were considered the most

important criteria for player categorization. As guilds and data collection techniques evolved,

researchers broadened their investigations to include player experiences, the sharing of those

experiences, searches for friendship, cooperation for performing missions, and the formation of

virtual social norms [14, 21]. Guild structure and activity are now viewed as representative of

player needs, especially since guilds now tend to satisfy the social needs of players in virtual

worlds and serve as centers for interaction in the physical world.

Due to the growing complexity of MMOGs, newbies are increasingly choosing to play the

same games and to join the same guilds as their friends in the physical world; this allows them

to rely on each other for help in elevating skill levels. Veteran players report asking friends and

acquaintances in the physical world to join their guilds as a means of reducing the risks

involved in dealing with strangers. Such actions are the result of different perceptions

regarding risk and trust for different guild functions [29]—for instance, socializing, seeking

help to solve problems or to build skills and experience, and organizing unified attacks and

defenses. A simple example is the exchange of tips, gifts, or favors between players with

distinctly different skill levels or experience. Those who act as the givers in these scenarios are

often generous, placing less value than the receivers on their gifts or favors [30]. Givers often

have enough time to evaluate the capabilities of the receivers, thus reducing the sense of risk.

However, perceived risks involved in exchanging equipment and other resources increase

rapidly when players form large groups to attack cities or camps; in these situations, exchanges

among players are more likely to be controlled by a trust mechanism. Game guilds consisting

of real-world friends help reduce the sense of risk [28].

spans (and profits), they have strong motivation to incorporate guild-based activities into their

products. However, companies also know that different players have different motivations for

playing, and therefore they must address those motivations while providing various incentives

associated with guilds. For example, guild members must belong to the same faction, and

certain virtual classes and occupations (e.g., priests) are considered more valuable to guilds

than others.

Bartle’s [22] proposed player categories include achievers, explorers, socializers, and

killers, each with characteristics and powers considered essential to establishing successful

guilds. The expanding variety in guild functions has significantly increased the complexity of

game social interaction mechanisms—for instance, guild management tasks attract players

with specific interests in organizing. Bartle [3] has identified two player types—politicians and

planners—who are more likely to participate in establishing and building guilds as a means of

achieving their personal game goals. Game mechanisms such as information sharing, learning,

cooperation, and dragon killing points (DKPs) also encourage communication and cooperation

within guilds in order to set goals, establish trust and reputation, and show responsibility [14,

31, 32]. Still, variety in individual player motivation and purpose translates into multiple

reasons for creating or participating in guilds as well as distinct differences in guild style and

game-focused guilds generally consist of players with little interest in spending time chatting

with other players.

2.2.3. Game Content for Promoting Group Play—Raiding Missions

“Raids” (also referred to as “raid missions” and “guild missions”) are high-end activities

suitable for groups of 20 to 40 avatars. Raids often consist of a series of missions requiring up

to one week to complete. The serial structure of these missions means that players must

collaborate more than once for planning and execution purposes. WoW designers have also

incorporated raiding missions suitable for groups consisting of 40 avatars whose levels go as

high as 60 or 25 avatars whose levels go as high as 70. One game feature that encourages

guild membership is the fact that players can obtain rare equipment and treasure only by

participating in raid missions. However, a successful raid produces only one copy of a coveted

piece of equipment or treasure, therefore a guild must perform the same raid mission many

times in order to make sure that all members have their own copies. This design feature

requires a sense of trust [31]; the perception of being a valued member of a team may explain

why mission execution is a very popular aspect of WoW’s high-end content. Furthermore,

different raids require different combinations of avatar abilities; inappropriate combinations

usually result in failure.

“large-scale cooperative problem-solving endeavors [1].” One of their more interesting

observations is the willingness of individual players to collaborate according to their avatar

capabilities. Thus, WoW avatars that serve as “tanks” during raids must be willing to withstand

attacks from powerful monsters; “healers” accept responsibility for restoring the lives of other

avatars; and the sole purpose of “DDers” is to inflict harm on monsters. To manage these tasks,

guild members organize themselves into hierarchies that include such positions as “raid

leaders” skilled at planning and execution and “DKP” managers who reward guild members in

accordance with their level of participation in a mission.

Methods for executing missions can be used to categorize guild type, since intra-guild

communication and mission and member management are considered important factors in an

individual’s decision to join or leave a guild. Researchers of games such as EverQuest have

described guild mission play as “work-like play.”[1, 33] Whether mission execution is

considered work or play, it has become a very popular aspect of WoW’s high-end content. Over

one-third (35%) of all Taiwanese WoW guilds that I collected data on were raid guilds. This

phenomenon shows that WoW game design has impacted the virtual world game community; in

turn, it is inspiring researchers to place special analytical emphasis on how high-end activities

influence player society.

researchers to date have used a combination of goals, motives, and size to categorize them [17,

34]. However, it is very difficult to infer guild organizing behaviors that affect player

communities from data on initial player motives, since they include such complex behaviors

and factors as stability, equilibrium, dismissal, breakdown, and conflict over participating in

missions and sharing rewards—issues that often develop based on interactions between player

and game environments. Another barrier to understanding these phenomena is the

above-mentioned difficulty of gathering in-game data to record the dynamics of guild

evolution, type, collective behavior, and movement between guilds.

2.3. WoW (World of Warcraft) and Client-Designed User Interface

2.3.1. WoW Success

Between its November 2004 release and January 2007 (the most recent data available),

WoW has attracted more than 8 million subscribers in North America, Europe, Korea, China,

Australia, and Taiwan. The previous record of 3.5 million subscribers was held by Lineage. It is

currently believed that one-half of the world’s MMORPG subscribers are WoW players [8]. In

North America, 220 game servers are used by 2 million subscribers—roughly ten times the

200,000 Taiwanese players who have used that country’s 46 WoW servers since its release date

[35]. Players and analysts generally agree that WoW has created completely new styles of play

same manner as its popular predecessors, Lineage and EverQuest.

2.3.2. WoW Game World

At the beginning of play, WoW players are allowed to choose a specific server to play on.

The three options for server type are player-versus-environment (PvE), player-versus-player

(PvP), and role-playing (RP). Players who choose the most popular servers (PvE) cannot kill

other player-controlled avatars, but players who choose PvP servers can participate in player

killing (PK) combat. The RP server type is for players who prefer fantasy storylines. Players

also select one of two faction types (alliance or horde) that differ greatly in terms of appearance,

capabilities, and storylines. Based on faction choice, players must choose from four “races”

(e.g., human members of an alliance or orc members of a horde). The final player decision

concerns avatar class, with examples being warlock, warrior, and priest. These four choices

affect player decisions and processes in joining guilds. Note that since the first two decisions

(server type and faction) are not commonly found in MMOGs, they serve as examples of how

game design can influence player communities.

Blizzard Entertainment’s response to the server choice issue has been to provide PvP and

PvE game situations simultaneously on separate servers. Once an avatar is placed on one server,

behavior in advance, or PK must be scheduled for a specific time in a specific arena.

Conversation, trading, group formation, giving assistance, and all other interactions between

avatars from different factions are forbidden. As a result, many view PvE servers as supporting

more leisurely play compared to competitive and violent PvP environments. The two settings

are strong reflections of conflicting player styles and preferences.

Another strong example of how design influences player interaction and game

environment is choice of avatar race. Players have eight choices divided equally between two

factions: night elves, gnomes, humans, and dwarves belong to the alliance faction, and orcs,

trolls, the undead, and taurens belong to the horde faction. Older fantasy MMOGs (e.g., The

Ring and Dragonlance) gave similar choices to players, but not as many. Experienced players

are well aware of how their selections affect play. Most players know that alliance avatars tend

to be brighter, more civilized, and attractive and that horde avatars tend to look evil, savage,

and cold. Previous studies have shown that avatar appearance and ability are important factors

that influence player choice [36, 37], therefore design differences among WoW factions and

3. E

NVIRONMENTS AND APPROACHES

3.1. New Trends for Peer Designers to Investigate Digital Worlds

Five characteristics of online virtual worlds mentioned in section 1.2 limit research

methods for investigating them. Computer monitor recording and network packet filtering are

two alternative methods for data collection; the shortcomings of each are discussed in section

1.2. In this section I will look at two current web 2.0 service trends: client-designed user

interfaces and open user-created data trends, both serving as examples of how providers

welcome user participation in certain aspects of service design for purposes of sharing and

personalization. Combined, the two trends allow services to meet the needs of different

populations and to integrate online information access. For example, Google Desktop API

allows users to read e-mail, check on the weather, and monitor sports scores and stock prices on

a single browser screen.

3.1.1. Open-Data Trends

With the development of the World Wide Web, the amount of information available online

is growing exponentially. Digital tools are making it easier to record activities and exchange

information than their counterparts in the physical world. Without digital tools, player

kinds of online information still require cooperation from service providers—for example,

non-text-based services (e.g., MMOG avatar information and online maps) and author

information placed in the deepest levels of webpage hierarchies. Virtual world researchers also

find it necessary to cooperate with game companies to obtain network traffic statistics and

subscriber information to analyze MMOG connection quality and gamer populations (

Figure 3).

Open data provided by service providers have the following properties: (a) third-party

users can only access data through an API; (b) interested users often need to apply for

authentication keys to access data; (c) some private data fields are restricted to prevent

improper access; and (d) open data have fixed fields and meanings that make analysis easier.

Using Google series service APIs as an example, Gdata (“Google data API protocol”) allows

users to query syndication format data using an authentication mechanism to access

user-created data. Other Google services support the Gdata protocol for developing APIs:

Google Apps, blogger, calendar, web album, YouTube, and social network API. Other examples

include e-commerce or citation service websites such as the non-commercial flickr service API

(http://www.flickr.com/services/). Many third party applications have been developed based on

this API, including flickrball3 (which gives access to a six-degrees-of-separation experiment

3

using flickr picture tags) and retrievr4 (which allows users to draw sketches or upload images

to find similar photos on flickr). MMOGs also provide accessible data for client-side

developers; a good example is Blizzard Entertainments’ decision to give WoW users access to

basic avatar demographic and auction information. The central point of this dissertation is that

this feature can be modified and used to investigate in-game player activities. Warhammer

Online is also adopting this feature, and other games and game designers are expected to follow

suit.

Figure 3. Relationships and interactions among service providers, web services (virtual worlds or MMOGs), and researchers.

3.1.2. Client-designed User Interface

Most UIs used with network services or PC software are fixed, single-user interfaces that

Service Provider Virtual World Subscriber Researcher Cooperate survey, interview, video recorder,

?

Monitor recorder

provide functions defined by the service provider; users cannot alter them or add new

components. As designers put greater emphasis on UI usability, a growing number are giving

users the ability to design their own components for attachment to main service applications. A

popular means of delivering this feature is providing a main engine and several default

components called add-ons (Figure 4 and Figure 5). APIs for UI design differ from the data

APIs mentioned above—for example, Yahoo Widgets API (http://widgets.yahoo.com/tools/)

and Google Gadget API (http://code.google.com/apis/gadgets/). Users can also add third-party

components. Common add-on functionalities include writing memos, reading E-mail and RSS

news, and checking personal calendars.

Based on this trend, several add-ons are automatically downloaded and attached on the

main screen the first time that players connect to WoW; they can later download additional

add-ons written in WoW UI API from third-party websites. UI design is an important issue for

game companies, who are acknowledging (a) player desires to quickly become familiar with

game controls and content, and (b) the shortcomings of single fixed UIs to meet the various

needs of complex MMOG game content (e.g., completing initial missions, joining guilds,

participating in raids, and executing combat). Compared with the original WoW UI shown in

Figure 6, the third-party product in Figure 7 is more suitable for group missions. The WoW UI

is created using the Lua script language for functionalities and xml for appearance. Since 2004,

strategy games, role-playing games, and MMOGs: Supreme Commander, Ragnarok Online,

World of Warcraft, and Warhammer Online.

Figure 4. Yahoo! Widgets. Users can design their own add-ons or insert add-ons made by third-party service providers and designers.

Figure 5. iGoogle. Users have the power to add gadgets to the iGoogle panel. Since it is a

10

Calendar

My ebay

Clock New York Times

Weather

11

Gmail

To-Do List

RSS Feeding

Figure 6. An original player-controlled UI provided by WoW.

Figure 7. Modified UI created by third-party source. Client-designed UIs can be found and downloaded from player-sharing community websites.

3.2. Web 2.0 Trends in Game Community

Web 2.0 characteristics found in digital games include player involvement in game design,

shared playing logs, and client-designed UIs. Some early examples are MUDs that allowed

players to design rooms and room functions (Bartle, 2003). Today, in addition to MMOGs, •Status

•Map

•Skills & tools •Chatting

Channel

•Status •Map

•Skills & tools •Chatting

Real-Time Strategy (RTS) and First-Person Shooter (FPS) games also give users tools for

designing their own maps, missions, and combat scenarios. Player-designed games that utilize

combinations of map editors or APIs with main game programs are called mods; they are

generally available for downloading from game-related websites. One of the most famous

mods is Counter Strike, which uses the FPS game Half Life’s engine and structure. Another

example is Warcraft III World Editor, which is similar to StarCraft’s level editor but allows

users to create and edit their own maps and scenarios. Many Asian players have designed

Warcraft III mods using different storylines (e.g., Chinese or Japanese historical legends) and

different game play features (e.g., multi-player tower defenses and two-team combat). A list of

the best-known player-designed mods is presented as Figure 8.

Figure 8. Screenshots of Warcraft III mods including two types of gameplay: tower defense and two-team combat.

Blizzard Entertainment is the most successful example of a MMOG company encouraging its

players to create, share, and download player-designed user interfaces. The three main UI

functions are (a) assisting with in-game control, (b) creating hints, maps, and libraries for

solving missions, and (c) collecting in-game information on avatar status. Whereas the most

basic function of user interfaces is to provide the outward appearances of game or website

services, my objective emphasizes the third function. Using WoW’s “AltasLoot” function as an

example, its main use is to provide data on the probability of “loot dropping”5 when a monster

is killed (Figure 9). Dropping probabilities are calculated based on information collected from

peer users—a function that I believe can be used to collect data on in-game avatar status.

Figure 9. Screenshot of a WoW player-designed UI called AltasLoot, which provides

probability of loot dropping.

5

When a player kills a monster, it drops “loot” and gold as rewards. However, loot for two monsters of the same type aren’t always identical, therefore AltasLoot has been designed to calculate the probability of loot being dropped from individual monsters. The AltasLoot UI only records loot dropping events for the player who attaches it as an add-on.

3.3. Investigating WoW Using the Client-designed UI Feature

WoW is attracting considerable research attention, similar to its popular predecessor,

EverQuest. EverQuest researchers felt that they had no choice but to spend long periods

playing the game they were studying, browsing discussion boards, and participating in or

observing game design workgroups to build their knowledge of game situations, backgrounds,

and cultures. It is now generally accepted that time-slice approaches are insufficient for

investigating progressive guild development and life cycles (i.e., creation, development,

splitting, and disbanding); such topics require data collected over time periods ranging from

several weeks to six months. For the most part, these time-consuming methods have been

passed over in favor of online/offline surveys and interviews aimed at understanding player

motives and behaviors [8, 22, 34]. Another major challenge for researchers is the

above-mentioned need to gain support from game companies for data collection.

In order to attract and retain players, many companies are creating mechanisms that allow

players to modify avatar appearances, characteristics, skills, and abilities to organize and join

guilds. As mentioned earlier, Blizzard Entertainment has made one of the most dramatic moves

in this regard by giving WoW players the ability to create and modify their personal interfaces

by means of an API. This feature allows users to collect progressive data on grouping,

changes in avatar participation in specific guilds over time and to use guilds as game agents for

the purpose of analyzing their distribution, differences, and life cycles. Specifically, I have

designed a WoW add-on that I call census++ to automatically record data on time, camp

(alliance or horde), avatar name, guild membership, level, race, class, and zone. A log sample is

shown as Table 2. Data that are returned from individual queries reflect information for all

avatars on one server. The time interval between successive queries depends on the number of

avatars on the server; average data recording frequency is once per 15 minutes.

Table 2. Example of WoW avatar behavior data collected using the game’s player-designed user interface.

Date Time Avatar ID Guild Level Race Class Zone

2007/9/17 13:25:23 冰心冷語 Aza Aza 70 血精靈 法師 影月谷 2007/9/17 13:25:23 Debuff Aza Aza 45 血精靈 法師 菲拉斯 2007/9/17 13:25:23 忘月 Aza Aza 61 不死族 牧師 撒塔斯城 2007/9/17 13:25:23 Miyuki Banshee 25 牛頭人 德魯伊 提里斯法林地 2007/9/17 13:25:23 Tubaobao Banshee 42 食人妖 獵人 荊棘谷 2007/9/17 13:25:23 死亡漣漪 Banshee 53 不死族 盜賊 阿塔哈卡神廟

WoW currently operates 32 servers in Taiwan. There are two server types, player vs. player

(PvP) and player vs. environment (PvE); avatars on one server cannot interact with avatars on

another. Players using a PvE server cannot fight with player-controlled avatars in opposing

camps unless both players agree ahead of time. These restrictions allow servers to be viewed as

unique societies. Note also that individual avatars can only join one guild, although guilds can

recruit new members and try to convince players to switch guilds. These exclusive server and

Figure 10. Each hierarchy level requires a separate analysis of recorded log data. For

example, at server-level researchers can compare differences in guild growth and decline in

PvP and PvE servers, and in-game community guild-level researchers can categorize guild

types according to player level and size and observe changes in those parameters.

Since census++ runs as a background program in WoW and records player status over time,

interested researchers can perform both cross-sectional and longitudinal analyses. Examples of

cross-section topics include (a) size and level distribution of guilds on one server, (b) member

balance between two camps, (c) avatar level distribution, and (d) guild member survey.

Examples of longitudinal topics include (a) changes in guild level and size, (b) avatar

movement between guilds, (c) efficiency in avatar level upgrading, and (d) consistency and

reliability in guild mission participation.

Figure 10. Structure and hierarchy comparisons between physical and virtual worlds.

World Societies Communities People MMOG Servers Guilds Avatars Players

4. R

ESULTS

4.1. Data Coverage

Different datasets were established for specific topics and WoW update events. Data on

period, frequency, server type, and total numbers of avatars and guilds in each dataset are listed

Table 3. The first set, consisting of data from 62 Taiwanese servers between January and April

of 2006, were used to analyze WoW guild dynamics. Since in-game guilds change slowly,

collecting frequency was adjusted to once every 6 hours. Since Taiwan WoW service began in

November of 2005, the dataset is considered young enough to capture early guild dynamics.

Three datasets from American and Taiwanese servers created between the end of 2006 and

early 2007 were used to determine differences between the two player cultures in terms of

behaviors aimed at game achievement. A shorter collection frequency allowed for monitoring

all changes in avatar level upgrading and guild activities. The fifth dataset was established to

address the topic of why players leave their original guilds and join other ones. This topic

Table 3. Datasets. Different dataset recording frequencies, periods, and scales were used for different purposes.

File Collection Time Frequency Server Type Total Avatars Total Guilds Taiwan 1/6~4/6(2006) 6 hours All 62 Taiwanese servers 641,805 285 per server US 9/1~9/25(2006) 12mins 1 PvP + 1PvE servers 40,038+50,703 1437+1456 Taiwan 12/1~12/25(2006) 15mins 1 PvP server 8,394 156 US 1/3~1/18(2007) 18mins 1 PvP server 41,097 1312 Taiwan 9/15~12/15(2007) 40mins Selected 5 servers, 2 PvE+3

PvP 98248

4.2. Dynamics and Evolution of MMOG Guilds and Other Organizations

4.2.1. Introduction

Online game guilds have a hierarchical leadership structure that allows players to act as

unified groups to solve joint missions. Whereas guilds used to be viewed as informal and

unplanned organizations, some recently released online games have incorporated guild

formation into their structure, adding mechanisms for establishing guilds and designing goals

and missions that require coordinated actions by members of well-organized teams [3, 38, 39].

Since online players frequently change their personal gaming goals, they often leave active

guilds to join others. Accordingly, online guilds are now experiencing cyclical lives that entail

creation, development, member suspension, splitting, merging, and disbanding, thereby

becoming perhaps the most representative social organization in online game societies.