A Game Analytics Model to Identify Player Profiles in Singleplayer Games

Abstract

Data Science is a multidisciplinary area related to systems, methods, and processes to extract knowledge from a high volume of data. In this context, we use the term game analytics to designate the science of online analysis and metrics of games. Research works in this area have been focusing on the use of player behavior data to increase revenue and avoid users leaving the game too early. To help this behavior analysis, we created a classification method based on Richard Bartle's player types model, mixed with the definition of Casual and Hardcore players, resulting in eight archetypes: Casual and Hardcore Killers; Casual and Hardcore Achievers; Casual and Hardcore Socializers; and Casual and Hardcore Explorers. We used the first four types from this new model in a singleplayer shoot'em up game, which gathers players' behavior attributes during each match. The right profile is chosen using K-means and Decision Tree algorithms, based on data from previous gameplay sessions. This whole method was tested using two new questionnaires to match the player's profile evaluation with the game's final profile, revealing accuracy between 75% and 80%.