Developing an engaging Metroidvania game through procedural content generation

Abstract

Procedural content generation (PCG) is widely used in modern game development to enhance replayability and reduce development overhead. However, applying PCG to the Metroidvania genre poses a unique challenge: how can algorithmically generated content preserve the handcrafted feel, spatial logic, and sense of progression that define the genre? This project explores a modular PCG system designed to generate individual Metroidvania rooms with structural coherence, traversal complexity, and gameplay diversity. Using the Godot Engine, a generation pipeline was implemented based on configurable primitives, zone segmentation, anchor-based pathfinding, and player-aware movement constraints. A custom interestingness scoring function was developed to evaluate each room based on factors such as anchor coverage, goal distribution, difficulty, and the variety of movement and ability-based interactions. The system was evaluated both quantitatively, via a Monte Carlo simulation analysing over 1,200 generated rooms, and qualitatively, through a visual assessment of selected room screenshots. Results from statistical modelling revealed that room width, door count, and difficulty level significantly influenced interestingness scores. The findings demonstrate that meaningful structure can emerge from procedural methods when guided by a well-defined evaluation framework, offering new avenues for scalable and expressive Metroidvania design.