Machine Learning Predicts Monster Levels for TTRPG Design

Designing balanced adversaries in tabletop role-playing games (TTRPGs) is a complex and time-consuming process, requiring designers to assign an "ordinal level" to monsters based on numerous numerical attributes. This research investigates how machine learning can automate and assist this task. The authors have created the first dedicated dataset for TTRPG monster-level prediction, derived from publicly available Pathfinder Second Edition data, framing the problem as tabular ordinal regression. The study compares various machine learning models, including classical regression, dedicated ordinal regression algorithms, and neural networks. The results indicate that tree-based ensemble models significantly outperform other approaches, achieving near-perfect ordinal ranking and high predictive accuracy. Explainable AI techniques confirm that the model's judgments align with human intuition and the underlying game rules, suggesting that machine learning can reliably approximate designer expertise. This demonstrates the potential for AI as an effective computer-aided tool for balancing game mechanics and streamlining TTRPG system design.