While psychologists analyze network game-playing behavior in terms of players’ social interaction and experience, understanding user behavior is equally important to network researchers, because how users act determines how well network systems, such as online games, perform. To gain a better understanding of patterns of player interaction and their implications for game design, we analyze a 1, 356-millionpacket trace of ShenZhou Online, a mid-sized commercial MMORPG. This work is dedicated to draw out hints and implications of player interaction patterns, which is inferred from network-level traces, for online games. We find that the dispersion of players in a virtual world is heavy-tailed, which implies that static and fixed-size partitioning of game worlds is inadequate. Neighbors and teammates tend to be closer to each other in network topology. This property is an advantage, because message delivery between the hosts of interacting players can be faster than between those of unrelated players. In addition, the property can make game playing fairer, since interacting players tend to have similar latencies to their servers. We also find that participants who have a higher degree of social interaction tend to play much longer, and players who are closer in network topology tend to team up for longer periods. This suggests that game designers could increase the “stickiness” of games by encouraging, or even forcing, team playing.

1. Network Game Design:
Hints and Implications of
Player Interaction
Kuan-Ta Chen, Academia Sinica
Chin-Laung Lei, National Taiwan University
NetGames 2006

2. Observation
User behavior is a key factor of how well a network system
performs (and how should a system be designed)
Example: Virtual World Partitioning Problem
Uniform distributed Server 1 Clustered Server 2
Server 1 Server 2
Server 3 Server 3
If game players tend to be uniformlyin the gameover the game world
clustered distributed world
dynamic and adaptive partitioning of the game world sufficient. required.
static and fixed-size partitioning of the game world is would be
Kuan-Ta Chen / Hints and Implications of Player Interaction 2

3. Motivation
Drawing Design Implications
from Players’ Interaction
for Designing More Responsive &
Scalable Online Games
Kuan-Ta Chen / Hints and Implications of Player Interaction 3

4. What We’ve Done
1. Collecting game traces (packet-level)
2. Inferring user interaction from game traces
Who are interacting?
Where are the players?
How do they interact? (stay together or team up)
3. Studying the implications of user interaction on game
design
Kuan-Ta Chen / Hints and Implications of Player Interaction 4

5. Talk Outline
The question
Trace collection
Deriving user interaction
Analysis of user interaction (and its implications)
Conclusion
Kuan-Ta Chen / Hints and Implications of Player Interaction 5

6. Game Studied -- ShenZhou Online
Kuan-Ta Chen / Hints and Implications of Player Interaction 6

7. Game Trace Collection
Game Traffic
Internet
Ethernet
Gigabit
Management
interface
L3 switch Game & Database servers
Monitoring
interface
Port
forwarding
Traffic Monitor L4 switch L2 switch
trace conn. # packets (in/out/both) bytes (in/out/both)
N1 57,945 342M / 353M / 695M 4.7TB / 27.3TB / 32.0TB
N2 54,424 325M / 336M / 661M 4.7TB / 21.7TB / 26.5TB
Kuan-Ta Chen / Hints and Implications of Player Interaction 7

8. Why We Use Packet-Level Traces?
Packet-level traces are much easier to obtain
no modification to game servers is required
recording packet traces does not increase the workload
of game servers
Player behavior inferred naturally connects to network-
level factors, e.g., IP addresses and network latency
Kuan-Ta Chen / Hints and Implications of Player Interaction 8

9. Extraction of Player Interaction
We would like to know …
whether any two players are at the same place
whether any two players are teammates
For each player (game session), we have …
a client packet arrival process
a server packet arrival process
We’ve proposed an algorithm
based on the correlations between the packet arrival
processes
Kuan-Ta Chen / Hints and Implications of Player Interaction 9

10. Example: Four Neighbors
Time (minutes)
Server packet rates imply the degree of PC/NPC
activities around the avatar
Kuan-Ta Chen / Hints and Implications of Player Interaction 10

11. Example: Four Teammates
Time (minutes)
Client packet rates imply the degree of game play
activities of the avatar
Kuan-Ta Chen / Hints and Implications of Player Interaction 11

12. Talk Progress
The question
Trace collection
Deriving user interaction
Analysis of user interaction (and its implications)
player dispersion
network proximity
social interaction
Conclusion
Kuan-Ta Chen / Hints and Implications of Player Interaction 12

13. Dispersion of Players
The dispersion of players in the game world:
well modeled by Zipf distributions
30% of players gather in the top 1% of places
Implications:
static and fixed-size partitioning of the game world
might be insufficient
dynamic and adaptive partitioning algorithms should
be used
Kuan-Ta Chen / Hints and Implications of Player Interaction 13

14. Peer-to-Peer Games
Reducing server load more scalable
Faster response time
Audio/video communications
Client-server architecture Peer-to-peer architecture
Kuan-Ta Chen / Hints and Implications of Player Interaction 14

15. Overlay Construction
Game World View Network Topology View
the character played using the host
2
1 1
2
?
2
? 1
?
2 2
?
1 ? 2 1
2
How to construct overlay networks?
Goal: to optimize the overall transmission latency
i.e., how to pass information between the peer nodes?
Kuan-Ta Chen / Hints and Implications of Player Interaction 15

16. Overlay Construction Alternatives
Design 1: 2
1 1
2
by network distance 1
2 2 2
(optimize network
latency) 1 2 1
2
Which design leads to more efficient Network Topology View
Design 2: overlays for online games?
by virtual world distance 1
2
1
2
(connect frequently 1
2 2 2
contact nodes)
1 2 1
2
Network Topology View
Kuan-Ta Chen / Hints and Implications of Player Interaction 16

17. Similarity between The Two Approaches?
Network Topology View
1 2
2 1
two approaches
Game World View 1
2 2 lead to different
2 overlays
1 2 1 2
Network Topology View
1 2
1 2
two approaches
2 lead to similar
1 2
1 overlays
1 either approach
1 2 2 is OK
Kuan-Ta Chen / Hints and Implications of Player Interaction 17

18. Correspondence between
Network Distance and Virtual World Distance
Observation irrelevant players
neighbors
Players who are neighbors or teammates tend to be
teammates
closer to each other in network topology
Evidence 1: Players who have interaction tend to have
closer IP addresses
Evidence 2: Players who have interaction tend to have
shorter network latencies (between them)
Kuan-Ta Chen / Hints and Implications of Player Interaction 18

19. Implications of Network Proximity
For client-server architecture
improves the fairness of game playing, as interacting
players tend to have similar latencies to their servers
For peer-to-peer architecture
message delivery between the hosts of interacting
players is faster
opportunities for optimizing network latency between
interacting players
Kuan-Ta Chen / Hints and Implications of Player Interaction 19

20. Effect of Network/Physical Distance
Observation (for a group of players):
network distance team play time
Explanations and implications:
real-life relationship carries over into the game;
real-life interaction plays a key role in game play
enriching in-game communication encourages
players to be more involved in team play.
diff/8 same/16 same/24 same/32
Kuan-Ta Chen / Hints and Implications of Player Interaction 20

21. Effect of Social Interaction
Observation:
degree of social interaction game play time
team size team play time
Explanations and implications:
due to the enjoyment derived from interaction and
social bonds
a game could be made stickier by encouraging the
formation of large groups
Team size
Kuan-Ta Chen / Hints and Implications of Player Interaction 21

22. Conclusion
Packet-level traces
easier to obtain
feasible to extract user interaction
Findings summarized
partitioning of the virtual world should be dynamic
network proximity of interacting game players
games could be made more sticky by supporting in-
game communication and encouraging team playing
Kuan-Ta Chen / Hints and Implications of Player Interaction 22

23. Thank You!
Kuan-Ta Chen
http://www.iis.sinica.edu.tw/~ktchen
NetGames 2006