sirmammingtonham / alphastone

Using self-play, MCTS, and a deep neural network to create a hearthstone ai player

Caution

The account you are attempting to view is currently under review for potential violations of GitHub's Terms of Service and Community Guidelines. Some functionalities of this profile may be limited during the review process. GitHub is committed to maintaining a respectful and lawful community, and we take all reports of violations seriously. This review is part of our effort to ensure compliance with our policies. We appreciate your understanding. For more information, please refer to our Terms of Service and Community Guidelines.

alphastone's Issues

which CardDefs.xml do you use?

it seems that some cards in CardDefs.xml are not implemented in fireplace,so which vision of CardDefs.xml do you use?

the link:https://github.com/HearthSim/hsdata

how much time cost to train

how much time cost to train with random deck

questions

will overfitting ever be a problem with self-play or would an overfitting neural net always lose to the previous version?

ideas

Embedding vector for all cards with entity_id as index
Create additional policy/dqn for mulligans at start of game