Play-Persona: Modeling Player Behaviour in Computer Games
Play-Persona: Modeling Player Behaviour in Computer Games
Play-Persona: Modeling Player Behaviour in Computer Games
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
ARTICLE 9 - PLAYER MODELING USING SELF-ORGANIZATION<br />
IN TOMB RAIDER: UNDERWORLD<br />
IEEE Symposium on Computational Intelligence and <strong>Games</strong> 2009. Anders Drachen, Alessandro<br />
Canossa and Georgios N. Yannakakis<br />
ABSTRACT<br />
We present a study focused on construct<strong>in</strong>g models of players for the major commercial title Tomb<br />
Raider: Underworld (TRU). Emergent self-organiz<strong>in</strong>g maps are tra<strong>in</strong>ed on high-level play<strong>in</strong>g<br />
behavior data obta<strong>in</strong>ed from 1365 players that completed the TRU game. The unsupervised<br />
learn<strong>in</strong>g approach utilized reveals four types of players which are analyzed with<strong>in</strong> the context of the<br />
game. The proposed approach automates, <strong>in</strong> part, the traditional user and play test<strong>in</strong>g procedures<br />
followed <strong>in</strong> the game <strong>in</strong>dustry s<strong>in</strong>ce it can <strong>in</strong>form game developers, <strong>in</strong> detail, if the players play the<br />
game as <strong>in</strong>tended by the game design. Subsequently, player models can assist the tailor<strong>in</strong>g of<br />
game mechanics <strong>in</strong> real-time for the needs of the player type identified.<br />
Keywords: <strong>Play</strong>er model<strong>in</strong>g, unsupervised learn<strong>in</strong>g, emergent self-organiz<strong>in</strong>g maps,<br />
Tomb Raider: Underworld<br />
1. INTRODUCTION<br />
Be<strong>in</strong>g able to evaluate how people play a game is a crucial component of the user-oriented test<strong>in</strong>g<br />
process <strong>in</strong> the game development <strong>in</strong>dustry. Dur<strong>in</strong>g the development phases, games are iteratively<br />
improved and modified towards the f<strong>in</strong>al gold master version, which is published. Representatives<br />
of the target audience as well as <strong>in</strong>ternal professional testers spend hundreds of hours test<strong>in</strong>g the<br />
games and evaluat<strong>in</strong>g the quality of the gam<strong>in</strong>g experience [1]. Moreover, one of the key<br />
components of user-oriented test<strong>in</strong>g both dur<strong>in</strong>g production, as well as after game launch, is to<br />
evaluate if people play the game as <strong>in</strong>tended — and if not, to f<strong>in</strong>d out why there is a difference<br />
between the <strong>in</strong>tended and actual play<strong>in</strong>g behavior, and whether this has an impact on their play<strong>in</strong>g<br />
experience [1], [2]. Given that nonl<strong>in</strong>ear game design becomes <strong>in</strong>creas<strong>in</strong>gly popular — massively<br />
multilayer on-l<strong>in</strong>e games be<strong>in</strong>g a good example of the <strong>in</strong>creased popularity of nonl<strong>in</strong>ear sandbox-<br />
type games — the need of more reliable and detailed user-test<strong>in</strong>g is grow<strong>in</strong>g.<br />
207