Due to a steady increase in popularity, player demands for video game content are growing to an extent at which consistency and novelty in challenges are hard to attain. Problems in balancing and error-coping accumulate. To tackle these challenges, we introduce deep player behavior models, applying machine learning techniques to individual, atomic decisionmaking strategies. We discuss their potential application in personalized challenges, autonomous game testing, human agent substitution, and online crime detection. Results from a pilot study that was carried out with the massively multiplayer online role-playing game Lineage II depict a benchmark between hidden markov models, decision trees, and deep learning. Data analysis and individual reports indicate that deep learning can be employed to provide adequate models of individual player behavior with high accuracy for predicting skill-use and a high correlation in recreating strategies from previously recorded data. CCS Concepts•Information systems → Massively multiplayer online games; •Human-centered computing → User models; •Computing methodologies → Machine learning approaches;

No abstract

Many online games suffer when players drop off due to lost connections or quitting prematurely, which leads to match terminations or game-play imbalances. While rule-based outcome evaluations or substitutions with bots are frequently used to mitigate such disruptions, these techniques are often perceived as unsatisfactory. Deep learning methods have successfully been used in deep player behavior modelling (DPBM) to produce non-player characters or bots which show more complex behavior patterns than those modelled using traditional AI techniques. Motivated by these findings, we present an investigation of the player-perceived awareness, believability and representativeness, when substituting disconnected players with DPBM agents in an online-multiplayer action game. Both quantitative and qualitative outcomes indicate that DPBM agents perform similarly to human players and that players were unable to detect substitutions. In contrast, players were able to detect substitution with agents driven by more traditional heuristics.

Balancing games and producing content that remains interesting and challenging is a major cost factor in the design and maintenance of games. Dynamic difficulty adjustment (DDA) can successfully tune challenge levels to player abilities, but when implemented with classic heuristic parameter tuning (HPT) often turns out to be very noticeable, e.g. as "rubberbanding". Deep learning techniques can be employed for deep player behavior modeling (DPBM), enabling more complex adaptivity, but effects over frequent and longer-lasting game engagements, as well as comparisons to HPT have not been empirically investigated. We present a situated study of the effects of DDA via DPBM as compared to HPT on intrinsic motivation, perceived challenge and player motivation in a real-world MMORPG. The results indicate that DPBM can lead to significant improvements in intrinsic motivation and players prefer game experience episodes featuring DPBM over experience episodes with classic difficulty management. CCS Concepts •Human-centered computing → User models; •Computing methodologies → Neural networks; •Applied computing → Computer games;

No abstract