We’ve recently seen an emergence of strong artificial intelligence (AI) for difficult board games such as Go and Poker. There is yet to be a superhuman Magic: The Gathering (MTG) player, but I believe this only to be a matter of time.

Once we have such a player, there will be some particularly interesting consequences. Strong gameplaying skills will be able to guide deck-building.

We should be able to sic a strong MTG player on a card pool, allow it to play thousands or millions of games against itself and “crack the meta” (build the best decks from the available cards) on its own. Monitoring these AI players would greatly benefit the game designers and balancers.

Just imagine a world where each set is more or less perfectly balanced and no bans are necessary… But, unfortunately, this technology would be beneficial for card price speculators as well. In this article, I briefly describe the basics of game AI and consider a recent method for decision-making in MTG.

I describe some experiments to test this method and discuss exciting future work. The experiments are implemented in a simple Python module, OpenMTG, which I hope will be able to serve as a basis for AI MTG players. Read more from hackernoon.com…

thumbnail courtesy of hackernoon.com