A Non-intrusive Approach for 2D Platform Game Design Analysis Based on Provenance Data Extracted from Game Streaming

The usage of provenance data drastically increases the potential for game data mining since it is able to record causes, effects and relationships of events and objects during a game session. However, it commonly requires modifications in the game engine in order to collect such provenance data. The modifications in the game engine may be unviable in commercial (and not open source) systems. In this paper, we propose a novel and non-intrusive approach for collecting provenance data in digital games. Our proposal collects provenance data using image processing mechanisms and pre-defined image patterns, thus avoiding accessing and modifying the source code of the game. Using our approach, we are able to generate, analyze and visualize game design features based on the gameplay flow using provenance data. Furthermore, we evaluated our proposal with a well known commercial 2D game, called "Super Mario World".