Inferring and Comparing Game Difficulty Curves using Player-vs-Level Match Data

Area of Science:

• Game Studies
• Computational Intelligence
• Human-Computer Interaction

Background:

• Previous research formalized difficulty curves using function composition within a single game's dynamic difficulty system.
• Existing frameworks were limited to specific game mechanics and lacked cross-game applicability.

Purpose of the Study:

• To develop a generalized method for inferring and comparing difficulty curves across multiple games.
• To extend the formalization of difficulty curves beyond single-game, ratings-based systems.
• To establish a standardized vocabulary for discussing difficulty curve transformations.

Main Methods:

• Inferring difficulty curves from player-vs-level win/loss outcomes in gameplay data.
• Adapting a ratings-based dynamic difficulty system for broader game application.
• Utilizing function composition to compare difficulty curves across games with fixed or dynamic level ordering.
• Presenting an adjusted method for analyzing traditional ratings-based match data.

Main Results:

• Successfully inferred and compared difficulty curves from four diverse games.
• Demonstrated the generalizability of the proposed method beyond its initial single-game context.
• Validated the use of function composition for cross-game difficulty analysis.

Conclusions:

• The developed method provides a robust framework for understanding and comparing game difficulty.
• This approach facilitates the analysis of player experience across different gaming platforms and designs.
• The findings contribute to a more formal and universal understanding of game difficulty dynamics.