A Go variant where players see only a subset of the board — partial observability turns a perfect-information game into something closer to poker. The project explores how reinforcement-learning agents adapt to hidden-state Go, and includes a playable interface for humans to challenge the trained agent. TODO: link a demo session, document the observation model, drop in the win-rate plot.
InvisibleGo
activeBlind / partially-observable Go variant with an AI opponent.
gamerl
⊟
visualization tbd
drop a hero image at
public/projects/invisible-go.png and set image: "/projects/invisible-go.png" on this project in content/projects.ts