BinG: A Framework for Dynamic Game Balancing using Provenance

Among different reasons that can lead a player to stop playing a game, frustration due to challenges that do not fit to their skills may be one of the most critical. Besides that, the players’ skills improve along the time, and the previously selected difficulty level may become inappropriate due to the player’s improvement. This can result on decreasing the motivation for the player retention, as they could get bored because of the easy challenges or frustrated due to the harsh difficulty. In this paper, we propose a new approach based on gathered provenance data for dynamically tuning the game’s challenge according to the current player skills. To do so, we developed BinG, a framework responsible for collecting and processing data provenance, allowing for the development of different balancing models to be used externally by the game. BinG uses the concept of logical programming to deliver facts and rules observed during a game session, allowing querying over the database to understand what happened. Additionally, we conducted a study using a game developed in-house and a dynamic balancing model customized to that game through BinG. This study was performed with five volunteers, who played the game using the default balancing and our dynamic balancing. Through this experiment, we showed a performance discrepancy reduction of almost 50% for the most skilled player in relation to the less skillful player when using dynamic balancing.