Talks - Esther Alter: Procedurally Generated Monsters! A complete example of Python game development

Python can be an excellent choice for 2D game development, offering convenience and rapid prototyping capabilities despite performance limitations

PyGameCE (Community Edition) provides pixel-perfect rendering and excellent documentation, making it ideal for 2D games - though it’s slower than engines like Unity

Procedural generation was extensively used to create unique content including:

• Monster sprites and stats
• World maps and tiles
• Battle animations and moves
• NPC characters and dialogue
• Audio effects

Code organization was critical for success:

• Main game loop uses a renderer class system
• Commands abstract rendering operations
• NumPy handles image processing before PyGameCE rendering
• Test scripts can target specific renderers as entry points

Development constraints and insights:

• Game took 9 months from first commit to Steam release
• Total development cost was $100 (Steam fee)
• Python’s limitations forced creative solutions
• Clean code organization made debugging easier
• Cross-platform deployment was challenging

The “10,000 bowls of oatmeal problem” - procedural generation can create endless variations but maintaining interest is challenging

Avoided using AI/ML in favor of simpler procedural generation techniques

Key libraries:

• PyGameCE for core game functions
• NumPy for efficient image manipulation
• PIL for image processing
• H5Py for data handling

Game testing was simplified by Python’s ability to create quick entry points to any game state

Cross-platform deployment remains Python’s biggest weakness for game development, requiring separate compilation for each OS