Orbitope/coldopen

Predicting player skill from early gameplay: RL agent ladders as labelled training data for the cold start problem