Show HN: I trained a chess engine to play like humans

Category: ai-ml

Tags: chess, neural-network, transformer

Score: 7.0/10 (Innovation: 6, Technical: 8, Documentation: 8, Utility: 6)

ChessMimic is a neural-network-based chess engine that mimics human play across rating ranges, incorporating clock usage and blunder patterns. It uses a tiny transformer architecture (9M parameters) for CPU inference and achieves impressive move prediction accuracy compared to academic baselines like Maia-2. The project is interesting for its practical approach to making chess AI feel human-like, with a well-structured codebase and live demo.

Target audience: AI/ML researchers, chess enthusiasts, backend devs

Repository: https://github.com/thomasj02/1e4_ai · C++ · NOASSERTION

View on Hacker News