Preset AI gods - help please !

[Arralen]

Ok, folks, what are your suggestion for preset AI races?
(Those are supposed to be started as "human" and handed over to the AI in the very first turn. This will get rid of ridiculous pretender, path and fortress choices as seen so often from the AI. And it occures to me that this still applies in DOM2)

Had some good success in DOM1 with preset AI gods, as the comp used them quite well.

So I'm taking suggestions on how the races/gods should be created so that the AI can play them well.

As soons as I have compiled the complete list, I'll make the files available for download from my webspace.

A.

PS: Those races could be a good help for newbies to start playing with before making their own "constructions", what do you think?

[apoger]

In my opinion:

The computer AI does not understand the concept of building a supercombatant, so I wouldn't give it one.

The computer AI does not understand the concept of focusing towards an efficient or concentrated goal. So I would suggest avoiding pretenders that would take advantage of stuff the computer will ignore. Bless effects for example. It's pointless to give a pretender massive magic for blessings, only to have the computer make only a handful of sacred troops.

The only way to help the computer with consistency is to force it to pump up it's economy.

I suggest (for all nations);

A Wyrm Pretender with no magic. It's hard to misuse this, it's big and eats things.

A fortified city, since the computer almost never builds forts might as well maximize the one it's got.

Order+3
Production+3
Heat - whatever the nation likes
Growth+3
Luck+0
Drain-1
Dominion+5

For nations that get extra points due to temperature scale, just put it into extra dominion. If anything is left pump magic or luck scale.

Avoid Ermor [Ashen Empire/Soul Gate] as the other AI's have no way to respond to it.
Same for Pangaea [Carrion Woods].

Jotunheim should play Utgard. Pangaea should go New Era. Everyone else should be default theme.

That's my best advice for trying to help the AI considering the limitations of it's current programming. Hope it helps.

[Gandalf Parker]

There were long conversations about this before.

http://www.shrapnelgames.com/cgi-bin...=000281#000006

Id still recommend the Dom1 gods but that page seems broke...

Gods that the AI can play well arent quite the same in my opinion as gods the newbies can start out with. You might want to pick one or the other.

Also what might be easier for newbies is that instead of making god files for someone to download, make a map with the gods defined in it. That will take care of the AI and the newbie player in one shot. If you select starting locations, and define the info for each province near them, then you can include a text file giving hints for the first few turns.

[apoger]

>Gods that the AI can play well arent quite the same in my opinion as gods the newbies can start out with.

Oh, absolutely. My offerings below are for the computer not for players.

[Arralen]

Originally posted by apoger:
In my opinion:
The computer AI does not understand the concept of building a supercombatant, so I wouldn't give it one.

It can't build one, but it can shurely use a combat pretender. Seems to depend on nation/pretender/magic paths, though. I have seen a Nifelson (standard Jotunheim) and a Golden Naga (New Era Pangea) running around the map killing indies early. Than, on the other hand, it tends to put dragons into the lab, maybe because of the high research (path lvl 9)

The computer AI does not understand the concept of focusing towards an efficient or concentrated goal. So I would suggest avoiding pretenders that would take advantage of stuff the computer will ignore.

But what does the Ai take advantage of, and what not?
It does not use rainbow pretenders for site searching, as far as I know, but it may use the same mage for casting the searching spells if it is in a lab. And most likely it is there, as the high research value tells the AI to use the mage as researcher.
This is something only the developers can tell for shure methinks.

Bless effects for example. It's pointless to give a pretender massive magic for blessings, only to have the computer make only a handful of sacred troops.

But you can only make a handful of sacred troops each turn
If 30% of Jotunheims army are Woodsmen -obviously the AI likes light units anyway-, is this sufficient to justify some blessing effect that makes them actually useful (only I can think of is nature-9, by the way, which doesn't fit that well with the main theme, and may endanger the priests)

The only way to help the computer with consistency is to force it to pump up it's economy.

That's an too easy solution, and doesn't do the AI justice. I'm looking for something a little bit more elaborate. And I'm pretty shure it will work (better), as it did in DOM1.

I suggest (for all nations);
A Wyrm Pretender with no magic. It's hard to misuse this, it's big and eats things.

And it's incredibly boring and dumb. More as the AI necessarily is anyhow.
I think some Ai nations could make use of bless effects (even only if it's for the commanders only), some use combat pretenders (spellcasters) very well, and the research boost from rainbow pretenders shouldn't be underestimated.
(The AI normally gets pretends with multiple times the normal amount of magic picks, remember)

A fortified city, since the computer almost never builds forts ..

Really hope this gets fixed in the patch. Will put some forts on the map to conquer until then.

That's my best advice for trying to help the AI considering the limitations of it's current programming. Hope it helps.

I think the AI deserves a little bit more credit.

And I'm not looking for an easy solution, but custom-tailored AI races that give the maximum amount of enjoyment for the player. Something that IMHO your suggested production-maxed, all-wyrm races wouldn't be able to accomplish.

In this sense, your posting wasn't all that helpful, sadly.

A.

[Gandalf Parker]

Originally posted by Arralen:
Originally posted by apoger:
In my opinion:
The computer AI does not understand the concept of building a supercombatant, so I wouldn't give it one.

It can't build one, but it can shurely use a combat pretender. Seems to depend on nation/pretender/magic paths, though. I have seen a Nifelson (standard Jotunheim) and a Golden Naga (New Era Pangea) running around the map killing indies early. Than, on the other hand, it tends to put dragons into the lab, maybe because of the high research (path lvl 9).

The early concept of dragons (as far as I can guess) was that their use was in their surprise. What appears to be one of the other pretenders suddenly goes dragon on you. I think thats why AI's use dragons that way.

The computer AI does not understand the concept of focusing towards an efficient or concentrated goal. So I would suggest avoiding pretenders that would take advantage of stuff the computer will ignore.

But what does the Ai take advantage of, and what not?
It does not use rainbow pretenders for site searching, as far as I know, but it may use the same mage for casting the searching spells if it is in a lab. And most likely it is there, as the high research value tells the AI to use the mage as researcher. This is something only the developers can tell for shure methinks.

What we have learned from the other AI threads here is that there is mostly one AI. So whether the AI has a love of spells such as "call of the winds" and "seeking arrow" might not be dependent on the nation as much as on having someone in the castle who can cast it, and the gems for it. So if we learn what the AI does and build around it we might be able to make intelligent use of it. If those spells fit the AI a person is building on their map then we can add them. If we gave Ermor the theme that lets them generate an army automatically in any province they take, gave them a dragon with lots of air magic, and added a site to their home that generated air gems.... That would create an Ermor that can break out like a disease all over the map fairly early. With mods you could even give them Call of Winds as a starting spell.

Bless effects for example. It's pointless to give a pretender massive magic for blessings, only to have the computer make only a handful of sacred troops.

But you can only make a handful of sacred troops each turn
If 30% of Jotunheims army are Woodsmen -obviously the AI likes light units anyway-, is this sufficient to justify some blessing effect that makes them actually useful (only I can think of is nature-9, by the way, which doesn't fit that well with the main theme, and may endanger the priests)[/quote] Regen and berserk? Thats playing into their strengths, not their weaknesses. In theme yes I guess but what about adding fire to their weapons? speed boosts?
particularly reinvigoration. Hmmm can they get the bull that feeds troops when it goes forth?

The only way to help the computer with consistency is to force it to pump up it's economy.

That's an too easy solution, and doesn't do the AI justice. I'm looking for something a little bit more elaborate. And I'm pretty shure it will work (better), as it did in DOM1.

I suggest (for all nations);
A Wyrm Pretender with no magic. It's hard to misuse this, it's big and eats things.

And it's incredibly boring and dumb. More as the AI necessarily is anyhow.
I think some Ai nations could make use of bless effects (even only if it's for the commanders only), some use combat pretenders (spellcasters) very well, and the research boost from rainbow pretenders shouldn't be underestimated.
(The AI normally gets pretends with multiple times the normal amount of magic picks, remember)

Rainbow and bless gods can go together. Some testing might show but Im thinking that the pretender choice might be what the code looks at for stay-at-home or go-to-war. Titans and large gods tend to leave home even if they seem like excellent rainbows? Human-like small gods tend to stay closer? Which might explain the dragons. I wonder if once he changes form due to an attack, he goes out? Experimenting with a .map could tell you since you can assign him to an AI as starting out in either form. (based on the number of mob)

A fortified city, since the computer almost never builds forts ..

Really hope this gets fixed in the patch. Will put some forts on the map to conquer until then.

Please lets not degnerate this into another "ai is broke" thread. Lets make intelligent use of it. If a nation would benefit from lots of forward castles then give him a wandering god with enugh earth magic to maybe find firbolgs or other castle sites. If more labs then make sure they can find those sites. more temples, etc. If none of those then some nations might be better off with a much stronger home defense.

That's my best advice for trying to help the AI considering the limitations of it's current programming. Hope it helps.

I think the AI deserves a little bit more credit.

And I'm not looking for an easy solution, but custom-tailored AI races that give the maximum amount of enjoyment for the player.
A.

Sounds good but then I think we have defined this as being different than pre-set gods for the newbies. This will be an excellent line of discussion toward a difficult .map

One I had in mind was something with an island surrounded by all the water, then a large continent. maybe up in the corner of a non-wrap map. Or a land area connected by a choke-point with an ocean on each side. Another way might be a polar continent surrounded by ocean on a wrap-around (antarctica?).

Specifically built gods and boosts on Ermor, Rlyeh, and Atlantis built around the AIs good and bad points. With all 3 of those starting so tight in an area, and allied so they wont attack each other. You would have to get thru the water nations (or use late-game transport tactics) while Ermor is forced to just sit and grow huge. Now THAT would be a challenge.