Understanding Game Balance with Quantitative Methods

Abstract

Game balancing is the fine-tuning phase in which a functioning game is adjusted to be deep, fair, and interesting. Balancing is difficult and time-consuming, as designers must repeatedly tweak parameters and run lengthy playtests to evaluate the effects of these changes. Only recently has computer science played a role in balancing, through quantitative balance analysis. Such methods take two forms: analytics for repositories of real gameplay, and the study of simulated players. In this work I rectify a deficiency of prior work: largely ignoring the players themselves. I argue that variety among players is the main source of depth in many games, and that analysis should be contextualized by the behavioral properties of players. Concretely, I present a formalization of diverse forms of game balance. This formulation, called `restricted play', reveals the connection between balancing concerns, by effectively reducing them to the fairness of games with restricted players. Using restricted play as a foundation, I contribute four novel methods of quantitative balance analysis. I first show how game balance be estimated without players, using simulated agents under algorithmic restrictions. I then present a set of guidelines for using domain-specific models to guide data exploration, with a case study of my design work on a major competitive video game. I extend my work on this game with novel data visualization techniques, which overcome limitations of existing work by decomposing data in terms of player skill. I finally present an advanced formulation of fairness in games - the first to take into account a game's metagame, or player community. These contributions are supported by a detailed exploration of common understandings of game balance, a survey of prior work in quantitative balance analysis, a discussion of the social benefit of this work, and a vision of future games that quantitative balance analysis might one day make possible.