QUO VADIS - Game Development & Business conference

Always-online games are rapidly transforming the video game industry. Video games entail complex and interesting problems, and the actions performed by their players generate extremely rich datasets that constitute an ideal playground to understand human behavior.

Data-driven game development has extraordinary potential both to provide players with an optimized game experience and to increase user engagement. Machine learning methods can be used to predict individual player behavior, revealing information such as the moment and game level at which any given player will leave the game, when they will make their next purchase and which virtual item they are likely to select, in which in-game events they would like to take part or which playing style they would find most enjoyable to compete with. This allows developers and game planners to take customized preventive actions aimed at maximizing player engagement and lifetime.

I will review the main artificial intelligence techniques that will revolutionize the mobile game industry, focusing on deep learning and ensemble learning methods that are crucial to provide a customized player experience through personalized game events, marketing campaigns, and in-game item/action recommendations. I will then discuss how the ability of these methods to deal with really big datasets (petabytes of data) and easily adapt to different kinds of games and players allows for their use in an operational environment.