Early game AI

[beorn]

I have had several intersting games in a row deteriorate into a blowout because so many AI civs were wiped out, presumably by the barbarians, rather early on. We're talking by turn 125.

I have to assume that this could be helped by AI coding. I almost never lose a city to the barbarians in the early going, and I really don't do anything very fancy. Plus I always play a couple notches higher than noble, so I assume the AI civs are losing to civs despite having a pretty nice bonus against barbarians.

All I have to do is
1) strongly favor hill locations
2) fortify preferably 3 warriors in each of my early cities
3) don't unfortify them to go do something else
4) if on a hill, use the first promotion on hill defense

I may lose the occasional improvement, but I cannot remember the last time I lost a city, while using this strategy on my first 2-3 cities, even when facing a very incessant stream of barbs.

If this isn't possible, I really think toning down the barb numbers would improve gameplay, because as it is, the barbs are giving me promotions galore while wiping out so many civs there is no need to fight anyone other than the barbs.

 

[xalien]

Yeah, stupid AI's build 1 warrior in a city and then spam scouts to explore the world. I constantly have to open WB and put some archers in their cities to give them a chance.

 

[Guale]

The AI's problem has always been that it favors expansion over defending its current cities.

 

[xienwolf]

First thing that the AI ever check for building anything is if they have enough workers (unless they consider themselves to be in danger, which means they are at war and losing, or an enemy is VERY nearby).

 

[Bahahotep44]

I, too, have resorted to the WB-archer fortification strategy to prevent too much early AI destruction, especially on huge maps. (I'd play on gigantic maps if I could; I just can't help myself.) I just wait at least until Orthus spawns before placing them so that I have the pleasure of early exploration into the unknown.

 

[pchem]

I agree. The early-game AI is the second most glaring deficiency in the game AI. The worst part of the AI is anything having to do with spells or unique abilities.

Improving AI to make getting 2-3 warriors with Hill defense or anti-melee promotions in each city as a top priority would do wonders for helping the AI survive early barb waves, especially in raging barb worlds. Then, when available, an archer with city defense.

The #1 thing that could be done to improve the gameplay experience of FFH2 would be to improve some really basic things about the AI.

 

[Kjara]

Have to agree with the early AI's love of scouts. Its so amusing(in a depressing sort of way) to see a stack thats 4-6 scouts and one warrior come and attack you when you do an early dow to grab a free worker or somesuch.

Can somehow lower the weight on scouts or raise the weight on warriors to strongly suggest to them that they should have a larger ratio of warriors to scouts at the start? If they had a lot more warriors, that would at least be a step in the right direction.

 

[Fafnir13]

3) don't unfortify them to go do something else

I find that if I take a little initiative (when I have suitable back-up and favorable odds of course), the extra xp from attacking can really help get some super units up and running quickly. Plus I'm often using a small stack or two to protect my workers (running them away whenever something shows up is a sure fire way to stagnate) and my valuable improvements. I don't know if the AI could really make good use of this info though.

Honestly, they just need to build more warriors then they are. I saw the Luchiurp recently in a game where Orthus had taken their capital. They had Barnaxus with a couple escorts hanging around there while their only other city was being guarded by two scouts. Luckily for them, Orthus died attacking some Calabim units in the area (another pair of scouts).

If the AI's scout fetish could just be squashed I think they would start doing a lot better.

Edit: well that got ninja'd

 

[[to_xp]Gekko]

yep, I've noticed it as well. all those scouts instead of good ol' warriors often spell their premature death

 

[Nimbus]

Honestly, they just need to build more warriors then they are. I saw the Luchiurp recently in a game where Orthus had taken their capital. They had Barnaxus with a couple escorts hanging around there while their only other city was being guarded by two scouts. Luckily for them, Orthus died attacking some Calabim units in the area (another pair of scouts).

If the AI's scout fetish could just be squashed I think they would start doing a lot better.

maybe try going into civ4unitinfos and try adjusting the <iAIWeight> for warriors up from 0 to 10 and see if that helps the AI to produce more warriors than it is now.

 

[[to_xp]Gekko]

I don't think that would work nicely. IIRC warriors never obsolete, so raising their AIweight would probably make the AI prefer them to other more valuable units

 

[Black Attila]

I don't think that would work nicely. IIRC warriors never obsolete, so raising their AIweight would probably make the AI prefer them to other more valuable units

Could someone explain why this would not work nicely. I might be willing to try this and in looking at unit AIWeights
i found that most but not all heroes have a weight of 200.
Adepts have a 100, Mages a 200 and Archmages a 300
Disciples (fellowship/kilmorph/oo) have a minus 20 while the other disciples are a weight of 0
priests have a weight of 8 and high priests a weight of 16
druids a weight of 16
Freaks a weight of 100
and finally hawks a weight of minus 999
all other units have an aiweight of 0

so if increasing warriors from 0 to 10 is thought to create a horde of warriors and no other units why are we not seeing a horde of adepts and feaks by the ai at the expense of other units? ditto fro priests mid-game as 8 is definately higher than 0. what numbers are needed to create a small tendency as oppossed to a number needed to make them prioritize a unit?

I do not fully understand this and am hoping for some help.

 

[xienwolf]

AIWeight values adjust what the AI thinks of the unit in addition to the stats of the unit. If the unit is worth more than what the stats say (ie - can cast spells), then the weight is meant to account for that since the AI isn't set to see it yet.

 

[tharg]

Have to agree with the early AI's love of scouts. Its so amusing(in a depressing sort of way) to see a stack thats 4-6 scouts and one warrior come and attack you when you do an early dow to grab a free worker or somesuch.

Can somehow lower the weight on scouts or raise the weight on warriors to strongly suggest to them that they should have a larger ratio of warriors to scouts at the start? If they had a lot more warriors, that would at least be a step in the right direction.

A quick dirty fix would be to make it imposable for the AI buld scouts. Should imporve the AI's survivablity a lot.

 

[Monkeyfinger]

Hey, I don't build scouts, why should the AI?

A situation where the AI built a lot of warriors when they have the ability to make more advanced melee units would actually help them in an unintuitive, glitchy way that I really would hate to see. See, on every difficulty level - even settler - they pay 25% of what a human does to upgrade their units. Mass warriors + the ability to build champions = mass champions at the hammer cost of warriors + a little gold.

Before you bring up the AI lategame warrior swarms that were once common, those happen when the AI ignores the metal line (which they do less often these days.) If they can upgrade, they will.

 

[Kjara]

Why not rather than making warriors go from 0 to 10, move the scouts(plus all unique scouts of course) from 0 to -10 or whatever value is needed.

 

[snarko]

I think a better solution would be to remove UNITAI_ATTACK and UNITAI_CITY_DEFENSE from scouts. Scouts are made for exploring and that's it, for all other purposes in the early game you use warriors.

 

[loffenx]

Well scouts have their uses, especially for the svartalfar and the like, seems a bit drastic to put the AI propensity to build them at zero.

 

[beorn]

This remains my gretest problem with the game. Since .40 came out, I have started several interesting beginnings to games at emperor on huge plus maps with 15 or so civs, and, every time, I am down to 7 or so rivals by turn 150.

In my mind, this undermines the basic game. Wars are wise and needed when the world crowds up, but turtling is pretty much unbeatable when the world is underpopulated.

 

[PotatoOverdose]

...I have started several interesting beginnings to games at emperor on huge plus maps with 15 or so civs, and, every time, I am down to 7 or so rivals by turn 150...

Are you sure that some of those 15 rivals weren't killed by your other rivals?

A quick fix to the problem of barbarians would be to give the AI's a bigger bonus against barbarians like the experience bonus is effectively a bonus against the player on immortal and deity. It works pretty well in that application so it might work against barbarians as well.