In another thread, @Lyya stated:
This made me start to wonder about other aspects of building a strong defense team. What sort of abilities should be avoided on a defense team? What abilities does the AI actually use well?
1 Like
My guess is that the AI would use skull damage quite well? Sirrian posted a summary of the AI logic recently, I’ll see if I can track down that link…
EDIT: Here it is:
So adding to my above guess at skull damage, abilities that fire off of 4 and 5 gem matches would probably work ok with the AI as well.
3 Likes
In my opinion, the best defense team skills are the AoE skills (can’t screw up a target if there is no target). Color converters are extremely hit-or-miss.
Traits, being passive, are mostly okay. In particular, I think the “____ on 4- or 5-gem matches” traits work well.
3 Likes
I forgot all about that list. Thanks for reminding me about it.
Random gem spawners also work well, considering an actual player can’t control them much better.
Why do you think Goblins (Goblin Shaman) and Worms (Early-Mid Game) work so well for the AI?
Legendaries that randomly spawn gems can also be troublesome when AI-played, Jarl, Celestia, Venoxia, etc.
4 Likes
I learned that the hard way. I added Valkyrie to a defense team (that previously had a decent win rate) and found out it was a big mistake. After that, the team lost most matches.
1 Like
I agree, though I’ll say that gem spawners can be played more effectively by a human. If the board is mostly devoid of the spawn color, I’ll hold off until it is more conducive of a spawned chain. This is especially true of Sheggra, who post-nerf is played quite poorly by the AI, very often handing a board of skulls to the invader.
1 Like
I assume a troop that targets the weakest enemy could work out well in defense too.
Likely so. At the very least you know what it’d do when cast 
lol True enough.
Is this referring to cards like Valkyrie and Alchemist that are any → one, or any of the converters, including the one → ones like Templar etc?
Both. The AI doesn’t choose good times to use transformation spells in either case. First off, the AI will tend to fire off spells as soon as they become ready, regardless of whether the board works for the transformation. Even when there is an any → one color that would extend the AI’s turn, the actual color the AI picks seems to be random.
2 Likes
So taking the ideas from above and just random musings about how the AI works, this is a basic team I came up with:
http://ashtender.com/gems/teams/6019,6022,6202,6100
Vampire’s Banner (Red/Purple)
Herdmaster
Miststalker
Marilith
Templar
Now Templar is a mana converter, but he’s more in there for the increase in armour, and he also gives the Defender trait to further increase everyone’s armour. The extra Green can help to fuel Marilith, but the loss of browns hurts Miststalker a bit…
Miststalker is aiming for the weakest ones, trying to use that factor of AI logic (ie you know basically what it’s going to target)
Marilith goes with the other factor in the same vein, that being AoE damage doesn’t need to be targeted.
Herdmaster… I was trying to think of a yellow user, and he showed up in the suggestions. He also has Leader, Impervious and his spell cleanses all allies, on top of exploding a number of yellow gems. It’s not Marilith’s level of explosion, but might be a little bit useful for filling up some mana.
What are people’s thoughts?
Also how do people feel that the AI uses Explosion? (Both targetted and random)
herdmaster will probably do the same thing gorgotha does… just spam it’s spell not bothering to use lower troop’s spells…
Does the AI only spam the spells of exploders that are in the first slot?
Not always. If the AI has Gorgotha or Carnex yes very much so; but if the AI have Herdmaster loaded and Sheggra or a Bone Dragon loaded, the dragons will cast first and then Herdmaster.
1 Like
does this mean AI considers higher base rarity a higher priority for cast?
That is a great observation. I wish i had an answer for you.
I’m going to have to pay more attention to that now.
1 Like