As far as I understand it that gets evaluated only once the initial decision, casting a spell, has been taken. It also seems to be subject to weights, possibly with the hero receiving a higher weight and other factors like mana cost getting taken into account. There may be some hidden “weight” score directly attached to each spell instead, that would take data file diving to figure out though. Still, main point is that the AI shoots blindly. It doesn’t know what a spell will achieve and whether it even makes sense to use it in the current situation.
There’s another step on top of that, picking a target. My gut feeling says that when it targets opponents, part of the weight given each target is determined by mana (or fill percentage). However, I’ve never analyzed that in detail.
This feels consistent with many experiences:
- In Wednesday’s Boss Dungeon, Rubirath’s team contains Green Slime and Dark Troll – who cast often despite contributing (effectively) nothing beyond a board shuffle or Extra Turn.
- Similarly, if Detect-o-Bot catches 2 or more Troops with full mana, it tends to start looping indefinitely until its spell self-destructs, despite that it’s no Chalcedony (haha) and other Red troops on its team could use that mana to cast their own spells.
- I forget which cpu team this is specifically, but it has a pairing of Stone Giant and Rock Troll where the cpu tends to cast them in order, despite the change to Stone Giant (remove Red → destroy Brown) leaving almost no Brown Gems for Rock Troll to double.
2 Likes
Just got back from my “sabbanical” (I could still see everything just not post). This ain’t a GoW Discord server it’s the dev forum and I didn’t treat it them or y’all as respectful as I could’ve. Truly my bad. I’ve apologized to Bramble & gmorin2 and thanked him for his service - to the USA and to this game (even tho I disagree with this particular quest). Gem on everyone!!! 
1 Like