The AI Pretender design is insane!

[PyroStock]

Twan said:
Earth just look better for most nations, and has some synergy with AI strength (huge armies of chaff needing some buffs or some solid summons to complete).

In a recent battle, AI Argatha had 3 earth mages chain-cast curse of stones without considering the magic resistance of everyone involved. The pretender cannot be everywhere to buff every army either. Nevertheless, Earth is hard for the AI to mess up.

Fire has no good summons out of uniques and may cause the AI pretender to chain cast fire from afar when gems are available (an annoying spell but not really cost efficient). May be good for nations with good sacred anyway

The AI isn't efficient with any gems in general. I was more impressed with the astral and nature paths under the AI.

(as well water is good for land AI nations with good sacred).

A water9 AI land nation could cast Tethis' Blessing send it's troops into the water & then override Tethis' Blessing with another global killing their troops. I've seen it happen. The AI hurting itself with Wrath of the Sea could be possible too since the AI has no problem casting Utterdark & Burden of Time to kill itself. But if the map has no/little water then it's not so bad.

Nature >8 would make sacred berserks, so it's bad for nations with holy mages. Nature 8 may be good (except when the AI poison its troops with bog beasts).

Bogbeasts require a water path so as long as you don't mix water & nature on the same unit you're fine. Nature8 allows for all nature globals and the regeneration is nice on sacreds with higher life. Mass Protection, Relief and Mass Regeneration can be nice army buffs too. The benefits from the nature globals I mentioned earlier alone would far outweigh the rare tarrasque IF the AI ever even it makes one which I've yet to see. I have seen Niefelheim with Gift of Health.

I don't trust the AI for using astral well, I fear to see AI pretenders feebleminded by some spells they may cast, or an immobile pretender teleporting somewhere and then never building a lab to return home (never seen this happen anyway, but with the AI I always imagine the worst case).

I've never seen the AI use teleport or astral gateway ever. Anti-Magic and Will of Fates are nice mass army buffs too. Unfortunately, I don't think the AI does Master Enslave or Arcane Domination. SoAP is a big help for the AI finding magic sites. The biggest danger is it going feebleminded, which is why I suggested an immobile with a chalice. It doesn't need to be the pretender.

If you don't experiment you'll never know for sure.

[Twan]

In fact I haven't experimented a lot boosting only pretenders as there is a better way to boost AIs : giving them some provinces with
-recruitable powerful units (troglodytes, trolls, knights on a province with high ressources)
-immobiles mages and a lab or appropriate site (ie enter to summon bane + immobile d1 mage)
-known gems prod sites (in provinces owned or nearby provinces)

Then you may let AI design the pretender, even if the choice is strange it will be hard.

(but I've tried one time to duel an AI just boosted by an immortal E10 cyclops, and I think it's one of the best boost if you just change pretender -but you may add nature and fire of course; give it relics etc-)

[hnchrist3]

Guys, lighten up.

Just programming a system not to lose at tic-tac-toe is hard, and that is a closed, super-simple, well-defined system.

Here we're talking about a super-complex system. The mere fact that it [AI] can do anything even remotely intelligent is incredible.

Don't criticize it.
Point out difficulties, surely, and any gaping holes, and if that isn't enough, join an MP game.
That'll teach you.

[Agrajag]

hnchrist3 said:
Just programming a system not to lose at tic-tac-toe is hard, and that is a closed, super-simple, well-defined system.

Not really, since Tic Tac Toe has a known undefeatable strategy.
But yes, even slightly more complex games are very hard to write good AIs for.

[Morkilus]

Even Checkers and Connect Four have been "solved". That means there is no way for an AI programmed correctly to lose. These are much simpler games, though.

[Loren]

hnchrist3 said:
Guys, lighten up.

Just programming a system not to lose at tic-tac-toe is hard, and that is a closed, super-simple, well-defined system.

Here we're talking about a super-complex system. The mere fact that it [AI] can do anything even remotely intelligent is incredible.

Don't criticize it.
Point out difficulties, surely, and any gaping holes, and if that isn't enough, join an MP game.
That'll teach you.

I consider the pretender I ran into to be an example of a gaping hole.