Comparative Analysis of the Left 4 Dead AI Director System and Modern DDA Architectures: A Systematic Literature Review

Dynamic Difficulty Adjustment (DDA) keeps players in a state of flow, yet the literature remains divided between classic rule-based architectures, such as the Left 4 Dead AI Director, and modern predictive models, with no consensus on the optimal architecture. This study maps DDA research trends and compares the trade-offs of both paradigms in terms of player experience. A qualitative-comparative systematic literature review combined with VOSviewer bibliometric analysis was applied to 10 purposively selected core studies, and the extracted data were compiled into a comparative matrix. The network visualization shows two separated clusters, algorithmic and psychological. The overlay visualization indicates a 2024–2025 shift toward engagement-oriented frameworks, and the density visualization reveals an empty region between the clusters. Classic methods are computationally efficient but prone to rubber-banding, whereas modern models are more precise but demand computation, data, or sensors that are susceptible to noise. No architecture is superior in every respect. The choice of a DDA architecture should match the available resources and the targeted player experience, and integrating the psychology of uncertainty into algorithms constitutes the main research gap.