About AutoGo

AutoGo is an open-source Go-playing neural network trained from self-play. Its implementation is based on DeepMind's AlphaGo, with ideas drawn from AlphaGo Zero and KataGo. You can play 9x9 and 19x19 versions, and use it to teach your own Go skills using Teacher Mode. For a walkthrough of how the system was built, see this tutorial.

Final scoring uses Chinese rules (area scoring with dead-stone removal).