Show HN: I had a Catan bot play itself 20k times to measure what wins

Category: ai-ml

Tags: game-ai, simulation, monte-carlo, catan, strategy-analysis

Score: 6.5/10 (Innovation: 6, Technical: 7, Documentation: 8, Utility: 5)

catan-advisor is a Catan-playing bot that simulates tens of thousands of self-play games to empirically determine what strategies win, with a browser-based advisor. It combines a rules engine, Monte Carlo lookahead, and rigorous head-to-head experiments to test heuristics, revealing counterintuitive findings like the importance of patience and the dominance of awards.

Target audience: game AI enthusiasts, Catan players, simulation researchers

Repository: https://github.com/martha889/catan-advisor · JavaScript · MIT · 1 stars

View on Hacker News