Yeah, I know, I must be crazy. I really wanted to come up with a solo variant for this no-luck team-vs-team hand-management dudes-on-a-map multi-use-cards brawler.
I already learned this game with The Solo System
to be able to teach it to others. So I figured it must be possible to build a challenging, non-adaptive AI that I could practice strategies against. Arty, the game's designer, already attempted this and in my hubris, I followed.
Obviously, there have to be some rules to govern what cards the AI plays, what action it selects and how it performs that action. Also, the AI will need some numerical advantage over the player (because of being dumb and not planning ahead) and I decided that the AI should not earn coins and instead automatically level up.
I already have a few of these systems in place and played three games with the AI. For simplicity's sake, Wasp (me) and Arien (AI) were battling each other. The crux of the AI is determining *how* it wants to perform each of its actions. E.g. for a typical move action, there are possibly dozens of valid locations and at least 5 that could make sense. How can you narrow the selection down with easy to perform heuristics, so that a "tie-breaker" card can quickly resolve any ambiguity between the "best" options.
Honestly, I am still unsure if I want to pursue this endeavor any further. It's hard to figure out a system that will narrow down AI decisions efficiently (to make the overhead small) and effectively (to make the AI interesting).
If anyone is interested in this, let me know.
no solo mode