Guarantees for Self-Play in Multiplayer Games via Polymatrix Decomposability

Self-play is a technique for machine learning in multi-agent systems where a learning algorithm learns by interacting with copies of itself.