Closed Jamesflynn1 closed 1 year ago
MCCFR produces an Epsilon Exploitable Opponent.
Run for differing number of iterations or vary parameters for different grades of opponents.
Use OpenSpiel MCCFR, requires conversion between AveragePolicy and TabularPolicy object.
Requires a wrapper to run and store the policy and configure parameters (if any). Get this working asap.
For EV experiments, might require research.