Some comments on the AI

[graywarden]

tharg,

you have hit one of those rare random games where peace reigns. i have never seen it happen for that long a period of time, but i have seen a number of games where you expand at a nice leisuraly pace. dont get use to it... haha..

 

[[to_xp]Gekko]

the AI still has issues grasping the concept of "take Orthus' axe after you defeat him" ... the axe has been sitting next to the Bannor capitol for a good 15/20 turns now. would a savegame be of any help?

 

[tharg]

tharg,

you have hit one of those rare random games where peace reigns. i have never seen it happen for that long a period of time, but i have seen a number of games where you expand at a nice leisuraly pace. dont get use to it... haha..

I have seen the Sheaim capturing cities from other civs, however I have yet to meet them. (Ok I got curious and looked in the worldbuilder )

AI doesn't understand always war gameoption much.

Ah! Are you saying the AI only attacks when it would have declared war anyway and doesn’t counter attack?

I also had influence driven war setting set, although I don't have any idea what this means?

 

[Seven05]

Influence Driven War is an awesome Civ4 mod component that changes your influence/culture with combat. It also causes some slowdowns during huge wars that go away if you completely eliminate your oponent so all of their influence is removed entirely from the map

 

[Sephi]

Ah! Are you saying the AI only attacks when it would have declared war anyway and doesn’t counter attack?

war only influences the decisions of city attack stacks (stacks of doom)
If it doesn't counter attack it either has only city defense units or it considers its attack odds too low

 

[Econian]

Those sneaky Illians

Last night the Illians cast Stasis and declared war on me in the same turn (around turn 120). They had about 3X my military strength. They came with about 12 axemen and 2 frost maidens. I had about 20 axes and 2 horsemen so I was able to beat them off losing about 10 axes. Their axemen all had about 3 promotions which was way less than 7-8 promotions I saw under WM7.42. Has something changed with the occasional promotions mod under WM 7.44?

This invasion was much more manageable because the Illian axes weren't super promoted like the Balseraphs were in my last game. I was impressed that the Illians knew how to use Stasis so well with a war declaration. If they had brought more of their home forces, I wouldn't have survived the assault. They should have been able to crush me, but fortunately didn't bring enough to do the job.

 

[[to_xp]Gekko]

another possible issue I've noticed is the Illians should prioritize building the temple of the hand a lot more when the city is on tundra terrain... +1 food from every tile in the BFC is huge and should be built ASAP.

 

[Sephi]

Those sneaky Illians

Last night the Illians cast Stasis and declared war on me in the same turn (around turn 120). They had about 3X my military strength. They came with about 12 axemen and 2 frost maidens. I had about 20 axes and 2 horsemen so I was able to beat them off losing about 10 axes. Their axemen all had about 3 promotions which was way less than 7-8 promotions I saw under WM7.42. Has something changed with the occasional promotions mod under WM 7.44?

only a minor change in the xp calculation, nothing big. Maybe they had less barbs to train to get lots of promotions.

 

[[to_xp]Gekko]

Balseraphs built a city right next to a goblin fort, only to get it razed a couple turns after due to not enough defense there.

 

[Sephi]

combat is random, so these things happen once in a while.

 

[[to_xp]Gekko]

I really have no idea why the Grigori settled there considering they started between the warriors and the scouts

 

[Skyre Noktis]

The AI just sucks at picking city sites, despite having a long and complicated function to work out which ones are supposedly the best. Might be a good idea to just rip the whole thing out and start again if we can get down a solid set of criteria for rating them.

 

[Zharkov]

Playing Wild Mana 7.44 I was quite impressed by some things the improved Wild Mana AI pulled off (taking into account that the civ4 AI is fundamentally flawed):

The AI sends cavalry to pillage and thus destroys their enemies' economy. And it counters tries to send fast units to do the same in their territory quite often (in one game, though, Bannor managed to pillage every Sheaim improvement with one chariot). I have even seen suicide commandos trying to pillage strat. rescources!

In one game the Elohim used their world spell very efficient: The nominal superior Sheaim conquered one Elohim city, the Elohim took it back and cast Sanctuary. Then again, in the next game, the now nominal superior Elohim lost one city after another to the Sheaim and only used their world spell when their last city was threatened (that is what I call fundamentally flawed: The AI only takes simple nominal values [e.g., total point value of the armies] into account and ignores evidence - like totally f****** *p - although there are algorithms possible to even translate this into nominal values [see Stardock; I hope firaxis will not forget to implement an AI into civ5... and I really hope I will live long enough to see pdp implemented in game AI, well, hope dies last]).

However there things that might be modifiable:

The AI has a tendency to send its best units (e.g., Heroes) out for unnecessary suicide missions, let some wander arround alone without apparent goal, thus making them easy to pick off, or send some out before they are fully healed.

While the infamous stack of doom might not be the most newest trick in the book, it is quite effective. The AI, however, has a tendency to show up with such a stack in front of my cities, and I think I am doomed in fact, but then it splitts up, sending heroes & cavalry into my hinterland, letting its catapults stay where they are, and its infantry retreats.

 

[[to_xp]Gekko]

still on the topic of Illians not building the temple of the hand, it just occurred to me why they didn't have a single blizzard after a long time in the game last time I've seen them: blizzards spawn on snow tiles and they had none around, just tundra.

 

[Sephi]

I gave them a high value for building a pagan temple in their capital now.

 

[[to_xp]Gekko]

awesome already in 7.51 or scheduled for 7.52?

 

[Sephi]

The AI just sucks at picking city sites, despite having a long and complicated function to work out which ones are supposedly the best. Might be a good idea to just rip the whole thing out and start again if we can get down a solid set of criteria for rating them.

I have created an alternative AI_foundvalue function that can be toggeled on/off via globaldefinesALT.

first tricky thing is how the AI should value bonuses? Sometimes it is important to value bonuses which cannot be improved yet.

7.52 Gekko

 

[Seven05]

first tricky thing is how the AI should value bonuses? Sometimes it is important to value bonuses which cannot be improved yet.

It gets even worse with FfH because of all of the different civics, techs, civ bonuses and whatnot that affect yields and commerce types.

For placing their starting plots I used to evaluate the yields based on early game techs and improvements they allowed and used the same to value bonuses with a lower value for 'invisible' bonuses. But, it just didn't work right for FfH, it was ok for most civs, good for a few and really, really crappy for others.

In the end I just went with static values per civ for each of the terrain and feature types and a flat value for bonus types. If you don't mind hardcoding values in for a set list of info types that's probably the best way to go. Especially considering how the yield changes for bonuses are kind of all over the place due to the extra effects they provide.

 

[[to_xp]Gekko]

the AI should expand more carefully when using raging barbs, they often get newly found cities destroyed.

 

[[to_xp]Gekko]

AI REALLY needs to be thaught how to defend from spiders, it's waaay too easy to kill all their workers with no retaliation.