Ai-Assisted Analysis Of Player Strategy Across Level Progressions In A Puzzle Game

Presenting levels commensurate with players' current understanding of game mechanics and level design is a significant challenge in designing games. Often game designers create levels by hand intending for the levels to increase in difficulty over the the course of the game while relying on their intuition or extensive user feedback, reiteration, and testing. Instead, this study starts from a number of procedurally generated levels originally generated by parameters expected to encourage a good difficulty progression and then presented to players during playtests. A number of AI-bots with different characteristics were then designed to assess the difficulty of each level. These findings are then compared with player data. Our findings show that bots encapsulating idealized player strategies can help us create a richer model of level difficulty that then reveals useful information about player struggles and learning across level progressions.