My name is Thomas Debeauvais. I'm a PhD student in the Informatics department of UC Irvine, in the group of Crista Lopes. Expected graduation date: June 2016.
My research focuses on player behavior in games, particularly real-money transactions, skill, content consumption, and retention. I apply statistical methods such as regressions or clustering algorithms on datasets with millions of player data. I can also use qualitative methods such as participant observation and semi-structured interviews.
In the mobile game Jelly Splash, three mechanics limit player progression: level difficulty, friend gates, and life regeneration. We found that the difficulty and gate mechanics increase revenues, but also cause churn. Published at FDG 2015.
In this paper, we looked at patterns of play, skill, and progression in a racing game. We also predicted when a player is ready to permanently increase the game's difficulty with precision and recall reaching up to 90%. Published at FDG 2014.
We correlate demographic variables with in-game activity and churn. For 100 World of Warcraft players, 10 drop out every month, but 5 come back to play again. Published at FDG 2014.
2012. PhD advancement paper about retention and buying gold in World of Warcraft.
We looked at demographic, social, and game features related to buying virtual gold in World of Warcraft. Achievement-oriented men with full-time jobs and little time on their hands were more likely to buy gold. Published at FDG 2012.
This paper looked at demographic, social, and game features related to player commitment in World of Warcraft. A quarter of players keep paying $13/month for 6 months or more without even playing the game. Published at FDG 2011.
Players of Ragnarok Online prefer private servers because they provide more customization, tighter communities, and less repetitive gameplay. Published at FDG 2010.
Technical Report from 2013. Performance analysis of a multiplayer jigsaw game server.
Preliminary numbers and microbenchmarks for a REST-based game server. Published at IGIC 2012.
RCAT is a scalable RESTful back-end architecture aimed at supporting thousands of concurrent and interacting users. Published at Netgames 2011.
Map-Reduce scales up parameter tuning, but can harm accuracy. I used Weka, Amazon EC2, and random forests. Published at LDMTA 2011.