slo-moe wrote:I appreciate the insight, Nekoatl. Just brainstorming here: maybe it can be solved by having a way to represent the game-state -- e.g., ordered list of card name without ID, power, toughness, tapped/untapped -- along those lines. Then see whether the game-state changes; if it doesn't change, don't repeat the action. Just my thoughts....
That would work for some cases, including the one you posted, but not for all. That's why I mentioned the Myr example. Throw in something like a
Myr Propagator, and the AI has a legitimate reason to keep using the combo until it has "enough" tokens on the field. Since it shouldn't know what you'll be able to field for defenses on your next turn, every additional
Myr Propagator token increases its chances of being able to overrun you on its next turn, so if it's seeking to maximize its advantage in terms of game state, it'll never stop creating tokens.