AI Considerations

[Mailbox]

Also warriors are so powerful for so long that they make a more than adequate defensive unit for a very long time. Hunters are great on defense as well (although I'm pretty sure the AI doesn't realize this) and only one tech up for some civilizations.

Something I've seen lately though is a relatively new problem, the "runaway AI" an AI that wages a great number of wars in the early game and ends up becoming disgustingly powerful and having twice the score of the next highest player. This could be its own thread though but I think having larger AI defense forces would help with this as well.

 

[rocklikeafool]

I think the solution would be better quality AI def forces. Meaning the AI needs to use def-based promotions, not jus spam useless promotions.

 

[Bad Player]

I may make a weighting change so the ai undervalues military goals if it hasnt met another player (making it more likely to pursue growth). Stays about where it is (maybe a little lower) if its met other players. And boosts the military weighting a bit if aggressive ai is on or its at war.

On that point, one time I moved my initial settler around and built my capital very close to another civ's capital. That should have been perceived as a very aggressive move (there ain't room for the both of us here!) and the AI should prepare for war in that situation (capitals within a certain distance of eachother).

 

[rocklikeafool]

Yeah, but as that limits how far you can expand, no 1 really expects you to do that. Forthermore, most people jus settle the initial settler wherever it is. They wanna get research, commerce, etc. goin.

 

[onedreamer]

Yeah this is the dilema. If we push growth the AI is weak to rushes but stornger if it isnt rushed. If we push military the ai is stronger against rushes but lags in growth.

Well, like someone said warriors can still make good defenders even vs. t2 units, especially if they use bronze weapons. So the AI could prioritize Mining but not Bronze Working. There is a huge cost gap between these two, but in defense you actually gain very few, since you gain +1 strength but loose the 50% city defense bonus of warriors which is higher than this, and warriors cost much less to produce as well. It must also be said though that in certain starts Bronze Working is vital because there are just forests around.

 

[GePap]

Has anyone ever experienced a Bannor AI that chose to undergo a crusade? I find it strange that given how central this civic is to Bannor, I have never seen the AI use it.

 

[Demus]

Well, like someone said warriors can still make good defenders even vs. t2 units, especially if they use bronze weapons. So the AI could prioritize Mining but not Bronze Working. There is a huge cost gap between these two, but in defense you actually gain very few, since you gain +1 strength but loose the 50% city defense bonus of warriors which is higher than this, and warriors cost much less to produce as well. It must also be said though that in certain starts Bronze Working is vital because there are just forests around.

you still need bronze working to get the bronze weapons, even though you can see the resource with mining .

The main thing about the AI which i don't like atm. is their tendency to bee-line to specific techs, and expecially the religions (khazad and ljos mainly). They're "saved" if they have gems/gold or deer/fur/ivory nearby, but otherwise their research is rediculously low for way to long. Make sure the AI techs atleast 1 commerce-giving technology before going to their designated research target should go a long way towards changing this. Which commerce tech is chosen should depend on starting terrain.

 

[[to_xp]Gekko]

agreed, sometimes you see them doing very dumb choices like beelining techs that will take 100+ turns to finish, that doesn't look very efficient

 

[zup]

AI does not need crusade for anything. All it builds is units. I use it for situations where I focused on economy early and then some dumb ai with hordes of tier 2 declares a war on me. crusade + rally = immediate 1000% boost to military power.

Anyway. How do Calabim horsemen compete with mithril weapons demagogs?

 

[rocklikeafool]

Has anyone ever experienced a Bannor AI that chose to undergo a crusade? I find it strange that given how central this civic is to Bannor, I have never seen the AI use it.

Yes, I have several times. They tend to use it tho only if your evil and piss them off by expansion and conquest of their allies or neighbors. Make em nervous nough and they use it.

 

[Squirrelloid]

I've also noticed that the AI may build abundant spellcasters, but I've never seen it cast a spell. Seriously, what's up with that.

I think ultimately the solution is to figure out what the AI *can* know, and figuring out how to get it to respond to that in an intelligent manner. I imagine the first is actually 'quite a lot' and the latter is a matter for brainstorming likely human reactions to given known conditions, and choosing the reaction which is most often the best one.

 

[Demus]

i've actually seen the AI use spells quite a lot, but mainly unit buffs (haste, dance of blades etc.). So yes, they do need 1-3 spellcasters per stack

 

[Ekolite]

I've seen priests of leaves summon tigers every turn while besieging my town in .32. I was pretty impressed actually.

 

[MagisterCultuum]

I've fallen victim to Ring of Fire a lot.

Kael recently revealed that he found that the AI was programmed to only consider using fireballs (and other summons too I think) when attacking a city, never out in the field. I guess that mages/archmages AI probably wasn't set up so that they would be moved to the front times to take cities very often, so the summoning spells were pointless. He claims this will be fixed in the next patch so that summons will be used a lot to defend AI cities or fight out in the open too.

The AI has a hard time handling units that cannot be built but have level requirements, but the game option that lets them ignore these requirements should help them out a lot.

 

[rocklikeafool]

Hmmm...true. I jus wish we had a concrete way of knowing how 'stupid' the AI really is. It's not like we can give it a IQ test. It'd prolly be so damn confused that it'd cease to function.

 

[xienwolf]

It is all in how complicated you make the conditions for each decision. A moron AI is set for simple booleans "IF X, then Y" but a genius AI is set for recursive loops which set values and a final comparison of said values at the end. Balancing all of the values is a delicate process though which requires a player who thinks numerically to set them out. If you can decide "I want to build building X because it offers 2 Culture and costs 150 Hammers. If it cost another 25 Hammers more though I would rather build Building Y because it offers 3 Health and I am currently under my happy Cap by N citizens and only gaining M food per turn because I am losing C Food per turn to Disease" Only for every possible building attribute and gameplay situation. With all new attributes any building can have being added to the equation by the same person who initially wrote it, or another person who has studied the balance of factors closely enough ot understand the precise value of the new added field. And of course as you make the field be of a different relative importance in the game, you must go back and re-adjust this value in the AI logic process (ie - if you make Culture no longer affect City Defense, then the value of culture on a building should be drastically reduced for an AI which is not seeking a Culture Victory or engaged in a Border-Push to maintain tiles it is working)


Overall, AI is something which most modders just don't think of writting at the time they invent new things, but really ought to start with so that they can have a solid grasp on the balance factor of each new addition (if you find yourself programming the AI to ALWAYS do a certain action, it is either unbalanced, or a rare one-time only affair)

 

[rocklikeafool]

Yes, I knew all that. I was being facetious.

 

[Imuratep]

I agree to your points, as I already tried this challenge and gave the AI that took my old civilization a very, very good situation (Malakim with dozens of well developed cities and Empyrion). I switched to the Khazad and recognized that they researched nothing except Runes of Kilmorph by turn 120. I managed to rise again with some conquers, but I finally realized, why the Khazad are always so incredibly weak in normal games.

 

[Zechnophobe]

Gah, another just awful one. Switch the second time to the Bannor... and they've researched exactly 0 economic techs.

There really isn't any point in 'favoring' military techs except in extreme situations... which by there very nature come up very rarely.

It's just depressing playing the Bannor pulling in +2 science a turn because you have 4 cities, and absolutely no economy. Archers and Axeman, yes, but so does EVERYONE ELSE because they got an economy going first, and still reached the same point as fast.

It is very literally broken.

 

[Zechnophobe]

Actually, y'know what, let me put it another way. The second switch especially, the civ you become is always one that hasn't researched education.