I have a basic MCTS agent for the game of Hex (a turn based game). I want to tune the parameters of UCT (the Cp parameter) and the number of rollouts parameter.

Where do I have to begin? The problem is that the agent is smart enough to always win if it plays first against another agent. So I don't know how to do the evaluation of each pair of hyperparameters.

If anyone has any ideas let me know.


Since no one answers, I will say the approach I used. It was inspired by the successive halving algorithm. The concept is that we find good performing hyperparameters for one agent by playing manually against it. Then, we put the agent we want to tune to play against it. We assign different configurations. All of these configurations are tried for a predefined number of games per configuration. We keep the best performing half, and then go to the next round with the half we kept and repeat this process. The last configuration that remains is the best performing configuration.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.