Intelligent Agents Lab


Creating Social Systems

Travian Network Datasets

The aim of this project is to study the structure of different types of networks that emerge in massively multi-player online games. Travian is a popular browser-based real-time strategy game with more than 5 million players. Players form alliances of up to 60 members under a leader or a leadership team. Each game cycle lasts a fixed period (a few months).

This network dataset was designed for studying multiplex network problems, in particular community detection since the ground truth for alliance membership is available. It was extracted from data from a study of virtual organizations conducted by researchers across several institutions, including the University of Arkansas. Data was collected from a 3x server (with a 3.5 month game cycle) based in Germany. These networks are from a 30 day game period.

  • Attacks: This network contains information about raids that occurred between players. The main goal of a raid in Travian is to loot resources.
    csv   graphml
  • Messages: Travian has an in-game messaging system (IGM) for player communication. The network contains information about these messages. Based on our analysis, around 70% of messages exchange occur between users in the same alliance, making it a reasonable benchmark for community detection algorithms. Along with the dataset itself, the community structure is also provided.
    csv   graphml  
  • Trades: Represents trades between players (both players must possess a marketplace). Similar to messages, the trades network is appropriate for testing community detection methods. The community structure for the trades network is also available.
    csv   graphml  
  • Community Structure: We treat the alliance membership information as the community structure.
    community structure
  • README: describes the structure of the datasets

Citation

If you want to use the network data please cite one of the following papers:

  • "Conflict and Communication in Massively-Multiplayer Online Games" Alireza Hajibagheri, Kiran Lakkaraju, Gita Sukthankar, Rolf T. Wigand, and Nitin Agrawal. Proceedings of the International Conference on Social Computing, Behavioral-Cultural Modeling, and Prediction. 2015. (pdf)
  • "Using Massively Multiplayer Online Game Data to Analyze the Dynamics of Social Interactions" Alireza Hajibagheri, Gita Sukthankar, Kiran Lakkaraju, Hamidreza Alvari, Rolf T Wigand, Nitin Agarwal. Social Interaction in Virtual Worlds. 2017. (pdf)

Code

We have developed a Python toolkit for multiplex link prediction on this dataset. The algorithms are described in these papers:
  • "A Holistic Approach for Link Prediction in Multiplex Networks" Alireza Hajibagheri, Gita Sukthankar, Kiran Lakkaraju. Proceedings of the International Conference on Social Informatics. 2016. pp. 55-70 (pdf)
  • "Extracting Information from Negative Interactions in Muliplex Networks using Mutual Information" Alireza Hajibagheri, Gita Sukthankar, Kiran Lakkaraju. Proceedings of the International Conference on Social Computing, Behavioral-Cultural Modeling, and Prediction. 2017. (pdf)

Other Publications

Other aspects of the Travian data have been used by the University of Arkansas team and their colleagues to study virtual teams and organizations. Please see the following publications for more information:


Current Contributors

  • University of Arkansas: Nitin Agarwal, Rolf T. Wigand
  • University of Central Florida: Alireza Hajibagheri, Gita Sukthankar
  • Sandia National Labs: Kiran Lakkaraju

Sponsors

  • University of Arkansas: NSF-0838231
  • University of Central Florida: NSF IIS-0845159