Author: Kunwoo Park, Meeyoung Cha, Haewoon Kwak, Kuan-Ta Chen Presented at International World Wide Web 2017 Conference at Perth

1. InternationalWorldWideWeb Conference 2017 2017.04.07
Achievement and Friends: Key Factors of
Player Retention Vary Across Player Levels
in Online Multiplayer Games
Kunwoo Park, Meeyoung Cha, Haewoon Kwak, and Kuan-Ta Chen
KAIST, Qatar Computing Research Institute,Academia Sinica

2. 2
$ 38.3 B
2016 global box
office revenue
$ 99.6B
2016 global game
market revenue
Ref: http://bit.ly/2mwci42, http://bit.ly/1ppyKvj

3. Player retention is a critical and long-
running quest in online game industry
Ref: http://bit.ly/2n51Ks5 3

4. Backgrounds
✤ Understanding motivations for play
✤ Player taxonomy are defined by one’s motivation: achiever, socializer, killer,
explorer [Bartle 1996]
✤ While general players are achievement-oriented, motivations can vary
across different demographics [Williams et al. 2008]
✤ What types of motivations are important for player retention?
✤ Achievement-oriented players tend to continue to play than other players
[Debeauvais et al. 2011, Bobora et al. 2011]
✤ Social influence helps predict retention better [Kawale et al. 2009]
4

5. Existing works assumed that motivations of
players are consistent across the game play
Ref: http://bit.ly/2n55p9f 5

6. Experiences across levels
6
Low level High level
Playing alone Collaborative play

7. Retaining maximum level players are more
important than general players’ retention
7

8. Why max-level players?
8
Major source of profit
Helping newcomers
to be adapted well

9. Research questions
RQ1.Across multiple phases within a game,
what are the characteristics of players who
achieve the next higher levels and get retained?
RQ2.Why do some individuals continue to play
even after having reached the max level?
9

10. Fairyland Online
✤ One of the longest serviced online multiplayer games
✤ 262,711,811 action logs of 157,812 players
10

11. Test of player retention
Phase Observing level Definition of retention
Low level 1 10-15 Achieves level 20
Low level 2 20-25 Achieves level 30
Mid level 30-35 Achieves level 40
High level 40-45 Achieves level 50
Max level 45-50 Active for 90 days
11

12. Methodology
✤ Two goals
✤ To model whether a player is retained for each level phase
✤ To understand changes of retention indicators
✤ Logistic regression
✤ Feature selection using LASSO
✤ Over sampling to make balanced set
12

13. Studied features
Achievement SocialTemporal
13

14. Retention indicators
14
✤ From low- to mid-level phases
✤ (+) Short average time per level up (Achievement)
✤ (+) The number of strong friends (Social)
✤ (-) The number of weak friends (Social)

15. Retention indicators
(+) Short time per level up
(Achievement)
15
High level phase
(+) The number of strong
friends before max level
(Social)
Max level phase

16. Trajectory over phases
Achievements are important for lower level, but social
features become the most important for maximum level!
16

17. Implications
✤ Game designers can make systems better by fulfilling
varying needs of players
✤ Player retention model could be constructed separately for
multiple phases
✤ Other user behavior studies could be done better by
considering multiple phases having common characteristics
17

18. Future works
✤ Findings from one game
✤ Missing features (e.g., churned friends)
✤ Retention of players at very early stages (e.g., level 1-10)
18

19. 19
Thank you
Meeyoung Cha
KAIST
Haewoon Kwak
QCRI
Kuan-Ta Chen
Academia Sinica
Twitter: @words_life
E-mail: kw.park@kaist.ac.kr
Paper: bit.ly/2ndqp1F