Why Self-Play Works for Go but Struggles with LLMs
i_dg23 · x · 2026-08-01
The author explores why self-play excels in two-player zero-sum (2p0s) games like Go, Poker, and StarCraft, but is much harder to apply in real-world domains.
The core reason is that self-play converges to a minimax equilibrium in 2p0s games, which is an unbeatable strategy in expectation. However, this isn't always optimal: against weak opponents, playing minimax might not maximize payoffs (e.g., failing to extract maximum profit from bad poker players). Because the real world is messy and non-zero-sum, self-play hasn't seen many successes in LLMs yet.
More from Research
- AI Failed All 20 Patent Claims Yet Insisted Its Reasoning Was Better — hashiromer · 2026-08-01
- Open-Sourced AI Agent Experiment Repo: Bug Hunting and Model Benchmarks — PawelHuryn · 2026-08-01
- New Method Pre-routes MoE Layers to Optimize I/O for Edge Streaming — dai_app · 2026-08-01
- PixelGPT Trained Locally on MacBook Air M3 in 10 Minutes Using Synthetic Data — zzznah · 2026-08-01
- Evaluating RAG Retrieval Without Ground Truth: Methods and Data Leak Analysis — dima806_dima · 2026-08-01
- OpenAI's Internal Model Solves 10 Major Open Math and CS Problems — alphacolony21 · 2026-08-01