Abstract

Dynamic Difficulty Adjustment (DDA) is a technique in game design used to adapt challenge in real-time based on player performance with the goal of maintaining engagement. Existing DDA implementations are predominantly tied to specific game engines and rely on invisible static scaling. This paper presents a modular engine-independent DDA framework for boss battles that produces observable behavior adaptations. The framework integrates live telemetry of combat with an external Python-based adaptive controller. A dominance ratio difficulty model with exponential smoothing drives one-directional phase transitions that trigger behavioral changes without modifying the core game engine. The framework was evaluated through an exploratory study with 30 participants from varying skill levels completing 166 boss fight sessions in Skyrim. Post-test analysis combined self-reported Likert-scale survey data with quantitative K-means clustering on seven telemetry features yielding four unique profiles: Reckless, Struggling, Defensive and Dominant. Survey results indicated strong self-reported player engagement (96.7%), high perceived fairness (72.4%) and near-unanimous behavioral awareness (93.3%). An individual trace analysis illustrated that the DDA made contextually appropriate decisions across the observed sessions. These results demonstrate that modular, behavioral and transparent DDA are practically achievable and positively received with implications for adaptive game design beyond any single engine or title.

IPC Classification

Keywords