We present a novel framework to procedurally generate executable cards for Hearthstone-like digital card games, along with an approach for evaluating card strengths using an evolved match environment. Here we introduce Chaos Cards, a digital card game in the style of Hearthstone, but designed to support the procedural generation of cards, including their diverse effects, via a grammatical model. To understand the potential performance of procedurally generated cards in actual games, we integrate a simulation-based approach to evaluate card strengths, and train a neural network model for fast card strength prediction. Because the strength of a card is most meaningful when considered in the context of the pool of competitive decks (know as the meta) it plays in and against, we propose an evolutionary evaluation approach which simultaneously evaluates card strength and refines the environment in which cards are tested. We showcase some example cards generated by our framework, along with their strength evaluations. Additionally, we conduct tests between evaluations from meta game environments and random game environments to show the importance of the environment in evaluating card strengths. Lastly, we show our neural network is able to learn the strength of important cards in a meta environments with largely positive correlation.
|Original language||English (US)|
|Title of host publication||Proceedings of the 16th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, AIIDE 2020|
|Editors||Levi Lelis, David Thue|
|Number of pages||7|
|State||Published - 2020|
|Event||16th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, AIIDE 2020 - Virtual, Online|
Duration: Oct 19 2020 → Oct 23 2020
|Name||Proceedings of the 16th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, AIIDE 2020|
|Conference||16th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, AIIDE 2020|
|Period||10/19/20 → 10/23/20|
Bibliographical notePublisher Copyright:
Copyright © 2020, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.