Title:

Stronger Practical Game-Playing AI for Duelyst II

Duelyst II is an online collectible card game (CCG) that features a 9x5 grid board, making it a cross between the popular CCG Hearthstone and chess. It is a partially-observable stochastic game (POSG) with a large branching factor and the ability to take several actions in a time-limited turn, making it a challenging domain for AI. On top of that, successor generation is very slow, so the best practical AI for the game currently is one-step lookahead using a hand-crafted static evaluator (which I developed). I introduce novel methods of action abstraction and behavior cloning for Duelyst II and show that two of my new agents exceed the performance of the existing AI. I also investigate why incorporating learning does not lead to a higher winrate even though it helps to predict some human player actions.

Name

Comment*

Comments are viewable only by submitter

Captcha*