Brainstorming: Weaknesses of the AI

SchpailsMan said:
In my current game, I noticed the following :

The Balseraph had built Loki (good), but kept him in their capitol all the time (bad), probably for the Gipsy Wagon-like culture effect or maybe to have a permanent Inspiration spell in their capitol. Anyway, it doesn't seem correct to lock their civ-only hero for tasks a couple of low cost units can do as well, and deprive themselves for Loki's unique skills.

Yeah we will need special ai for loki and the wagons. Rioght now the ai isn't to good with them.

The Malakim have trouble expanding, but they hardly make any effort. As of year 300, they still only have their capitol and are packing LARGE amounts of units (90% archers, with a couple of warriors and workers) inside without ever thinking of building a single settler. Their starting pos is near toundra, so I guess they had trouble in the beginning and lost land to their neighbours (they are currently surrounded by the Dovellio and another civ although I don't remember which one exactly), but still they have a perfectly good location for settling inside their own boundaries (a Gold resource, along with another one I can't remember, possibly rice or reagents).

My guess is that they still didn't get Mining so they can't exploit the gold resource yet and just dont go for it because they don't yet have a city that could use anything the Mining tech grants :crazyeye: Another possibility is that they don't want to settle right there because it's right against the Dovellio borders, which they have been at war with for the last ~100 turns (the war is not even likely to end soon : it will take centuries for the Dovellio to kill the ~12-15 archers in the Malakim capitol, and the Malakim don't look to have the tech and prod required to force their way through the land). I should also mention that I have supplied the Malakim with free sheep and horses almost since the war began since they were small and extremely far away from me while the Dovellio were already large and spreading like crazy. Anyhow, the Malakim STILL don't have a single mounted unit and won't get out of their city to grab whatever land/resources are still available to them.

At a high level the playerai is deciding that its on the losing end of a war so its putting all of its resources into defence (hence the stacks of archers). Which, although it may not be a good strategy in a specific situation is a decent strategy overall. You certainly wouldnt want to push the ai to make settlers which it was in difficult wars.

Anyway, after 300 turns they really should have got the tech (if not alrdeay) and the gold resource : it's only 4 squares away from their palace, and there are 2 perfectly settable tiles that would grant them both the gold and the other resource. I think the AI just didn't choose its priorities well enough and ended up locking itself into a full-defense strategy it has no chance of escaping from. I guess "Something" (yeah, that's not very precise) could be done to force the AI out of this behaviour... making it avoid the situation might be really hard, but doing something as simple as forcing it to produce offensive units from time to time seems necessary.

If the Malakim are more powerful than the Doviello than it would be appropriate to produce offensive troops. But if it is trying to outlast the war it is doing the correct thing (in general).
 
I globally agree with your comments regarding the Malakim, it's just that in that particular situation it feels wrong to me. That said, I'll try to have a look in the worldbuilder and the game log... maybe they already lost a couple of cities, or the Dovellio have a lot more units nearby than what I can see now (zero... but all I see is from religion-spying so it's not very accurate), and that alone would justify sitting on their rock till the end of the game.
 
Another Vanilla AI problem:

The AI is very happy to destroy improvements and roads within an enemy's borders. One problem with this is that they sometimes do it at the expense of attacking the nearby city, something that I've used to my benefit in the past (e.g., building roads and cottages on tiles that I can't even work, since it'll take two turns to pillage it).

Another problem is that those improvements can be quite useful when you've taken over a city, especially the roads which give nothing when you pillage them. Perhaps a quick check by the AI to see if they are 'confident' that victory over the city units is assured?

One final thing is that it is quite useful to pillage roads in 'wild' areas sometimes, especially those roads in tiles adjacent to your borders. Anyone can use those roads, and that typically means the enemies of said empire will be using them to zoom into the city radius. As well, in my most recent game I took the dragon hoard which was far away from my empire (hey, we dwarves need our gems!). I wanted to keep a route to the city so I selected one that went through a jungle/hill and pillaged the rest. That way any barbarians coming from that direction either had to try to oust my fortified units (not much chance of that) or try to sneak past off the road - and I had some mobile units on the hill to pick off those ones.

- Niilo
 
Noticed this last night - AI leaders with the BAR trait still dispatch units to barbarian cities, then sit them outside not doing anything until the city is taken or razed by another player.
 
AI never knew how to choose right promotions for it's mages, and it's still not quite good, so why not incerting the good(checked by people) development strategies for adepts/mages/etc? I post this because I see very often AI's Archmages/Summoners with only combat and some other unusefull with good spells promotions, it is really rare when I see AI's mage which promotions is looking ok, another thing, maybe mages and heroes dont attack when having low odds, but they still move to dangerous enemy lands without enought escort, risking to be killed. For example, Basium is always going allmost alone to the enemy lands, it's suicide even if he wont attack with low odds... AI get his mages and heroes killed very fast, while almost never killing my, I can't imagine human player doing such a things...
 
And something else: I noticed that AI when declares war looking only on diplomatic relations between our states and on our military count, so I made an expiriment, I started game and from the first turn of it built my capital in the very north of the map, surrounded with ice, and with only 1 tile to access to it(used the editor), built 7-10 immortals in this city and gived AI teams much time to grow, after some of this time theyr teams started to declare war on me, my only city was comletely unusefull, they couldn't take it so easyly because of immortals, but they declared war anyway..

I thing that before starting a war AI should realize first what it possibly can gain by it, second is what is his odds to win(not talking about numer of troops), by WIN i mean capture at least 1 usefull city or if enemy is tecnicaly more advanced and will win the game someday, it should realize how big the chance to break enemy empire
 
Recently, I was the Malakim, and Morgoth was my neighbor. I converted him to the Order, but for some reason he was still annoyed with me...

The diplomacy screen showed that he should have a net +5 relation with me, but this doesn't seem to be the case. Is there something wrong with the way that religions/alignment effect diplomacy?
 
I posted this in the "bugs" thread, but the more i think on it the more i believe it's an issue with the AI

Spoiler :
Farms:

These are only ever built on the periphery of my territory, and i cant tell if this is because of the allowance of the Kuriotates to access the third tier of city radious. But the only place farms get built (by automated workers) is in that third radii. All other territory is built up with cottages, workshops (if appropriate) and windmills/etc. Farms are always neglected. Mines are also neglected, and i have a lot of naked hills until i can build windmills. Normally i would believe that my workers are merely on "priority" however, i've noticed that they will "finish" working and go hide in my cities instead of putting something up on my naked hills OR changing landscape into farms (even if i dont yet have education) but DO have agriculture.

I'm not sure if the AI is behaving differently because of the 3rd teir mentality, but to test this theory, i had two cities built pretty far away from each other, say 7 or 8 squares. The two immediate rings around these cities were filled with cottages, the 3rd ring of each was filled with farms. I then placed a city exactly inbetween them. After everything had been built......this was in order to take advantage of all those farms built, my workers then IMMEDIATELY destroyed the farms and put up cottages. This anti-farm mentality my workers have, make it virtually impossible to build farms if i EVER want any automatic workers. This also means that i'll never really use Aristocrasy......as I have cottages, which produce MORE gold anyway, plus i dont have any farms in city radii anyway, so it doesnt matter.

Now I've played the Boy-kings empire enough to apprecaite the awesome capabilities of that culture, but for other civs it seems like farms are still necessary (or at least should be acceptable to workers). All my strategies have had to alter to "im not getting much food, but ill have more gold." My cities grow slower, consequently, and this is FINE......but it takes away one whole PURPOSE to an entire improvement-type.

This may just be an AI bug, but i suspect it has to do with the ai thinking about towns being near the city, and farms being on the edge. I am thinking that the AI doesnt differentiate between the Kuriotates and everyone else, so it puts "farms" on the periphery (3rd tier) of every city, regardless of faction. All "inner" squres get town. This is a great idea, except that not everyone can access that 3rd teir, and farms should be on the 2nd. NOt sure how to fix this.


-Qes
 
That is a problem of the tech tree. The ai selects which improvements it wants to have built even if it cannot built them so it keeps some tiles unimproved until it has the planned improvement available.

The farms are built next to cities to spread irrigation.
unfortunateley the AI has most of the time improvements available that are better than farms (in its view) and plannes those inside the cities. Right now i also know why (did not think about it until wrinting this...) The reason is the seafaring tech. When planning improvements the AI considers all available techs. With the -1 food for seafaring farms are no valid improvement to be built inside city radii for the AI.

It will be solved in 0.15.
 
Chalid said:
That is a problem of the tech tree. The ai selects which improvements it wants to have built even if it cannot built them so it keeps some tiles unimproved until it has the planned improvement available.

The farms are built next to cities to spread irrigation.
unfortunateley the AI has most of the time improvements available that are better than farms (in its view) and plannes those inside the cities. Right now i also know why (did not think about it until wrinting this...) The reason is the seafaring tech. When planning improvements the AI considers all available techs. With the -1 food for seafaring farms are no valid improvement to be built inside city radii for the AI.

It will be solved in 0.15.

I'm glad it'll be solved, how (because im curious) do you plan on fixing it to keep the tech tree generally as is (or if your not thats interesting too), while also keeping the AI on being prioritzational?
-Qes
 
In turn 353 (normal game speed!!) garimm gyr still defends all of his cities with warriors. He was completle defensless and crushed after a few turns. The Problen: he was ahead tech wise and had a far bigger empire than me.
 
Chandrasekhar said:
Recently, I was the Malakim, and Morgoth was my neighbor. I converted him to the Order, but for some reason he was still annoyed with me...

The diplomacy screen showed that he should have a net +5 relation with me, but this doesn't seem to be the case. Is there something wrong with the way that religions/alignment effect diplomacy?
Net diplomacy modifier does not absolutely determine AI mood. I noticed this when I installed the advanced foreign advisor mod for vanilla civ, which showed everyone's net diplo modifier in relation to everyone else in the game. I found it interesting how some would be Friendly at +5 yet just pleased at +8 for someone else.

- Niilo
 
Aletr said:
another thing, maybe mages and heroes dont attack when having low odds, but they still move to dangerous enemy lands without enought escort, risking to be killed. For example, Basium is always going allmost alone to the enemy lands, it's suicide even if he wont attack with low odds... AI get his mages and heroes killed very fast, while almost never killing my, I can't imagine human player doing such a things...
I agree with this problem. AI heroes don't concern me because they'll eventually be left alone.

A similar problem is the protection of injured regular units, especially when they just won a battle (and XP). A prime example of this in my last game was when a lowly Lizard Ranger (if I remember correctly) managed to kill my first Order hero (Valin?) - I was so pissed the computer chose to have him defend my city instead of the freakin' Heavy Crossbowman, but that's an unecessary aside. Anyway, I couldn't resist and I checked the AI unit's XP total with the World Editor. It had gotten ~50 XP from the battle. Yet the stack of AI units next to it didn't bother to protect their new 'hero' and I killed the sucker before he could gain his 5-6 levels.

- Niilo
 
Yes each AI Stack operates completely independend of the other stacks thats really a shame and thats teh reason why you often see lone units wandering around. I have a concept in mind to change that but after implementing it it will need very very much testing.
 
Is there a mechanic to change the AI costs of specific buildings? I know (obviously) that the AI at higher levels has 'discounts' on required hammers, but if the AI archery range was 1/2 cost maybe it would make some instead of defending cities with Warriors and Hunters for so long.

I know in the end this is just more 'cheating' to give the AI balance that many Civ players dislike, but I think it's a reasonable way to produce a better AI.
 
Mahatmajon said:
Is there a mechanic to change the AI costs of specific buildings? I know (obviously) that the AI at higher levels has 'discounts' on required hammers, but if the AI archery range was 1/2 cost maybe it would make some instead of defending cities with Warriors and Hunters for so long.

I know in the end this is just more 'cheating' to give the AI balance that many Civ players dislike, but I think it's a reasonable way to produce a better AI.

The consequence of AI cheating, other than feeling cheap, is that it is the AI's insistances that create the issue.

For example. If the ai, instead of building an archery range, decides to build 5 or six warriors. If you gave them a cheap archery range, would have 3-4 archers.

Now if im facing them, and i upgrade, then i may have 1 or 2 archers facging their 5-6 warriors. If they cheat, then they may have those 3-4 archers faceing my 1-2 archers. This is directly a problem, since i no longer have the technological edge. As it is, i've not really run into a computer player having mostly warriors. In fact, often enough they've more and higherly advanced units than i do. But I maintain a very small army, usually. Anyway, If there is a "warrior" surplus, on the part of the AI, it would be my guess that there are other considerations at hand. For example, its productive cities are building wonders, and buildings instead of units (which may be able to be upgraded versions), and its periphery cities, that cannot produce as much are in charge of units, which means warriors, since the cost of buildling archery ranges is a little high for non-productive cities.

I'm not really sure how this issue can be remedied without creating an inate discomfort in the mouths of players facing those enemies. As it wouldnt "balance" it, but instead balance it in the opposite direction.
-Qes
 
Personally im like the AI sometimes, and don't want to bother building things like Archery ranges, because they do nothing except allow building a unit i may not build many of there (usually i have one city with an archery range and send troops to places).
This would be fixed if all unit-allowing buildings gave some bonuses (like Hunters Lodge, which has +1 health for deer and +1 happiness for furs, which is likely why the AI's build them).
 
Sureshot said:
Personally im like the AI sometimes, and don't want to bother building things like Archery ranges, because they do nothing except allow building a unit i may not build many of there (usually i have one city with an archery range and send troops to places).
This would be fixed if all unit-allowing buildings gave some bonuses (like Hunters Lodge, which has +1 health for deer and +1 happiness for furs, which is likely why the AI's build them).

Yeah, I wouldn't mind that at all. I guess the only problem is trying to figure out how to make it believable. What would allow both archery units, and give a city bonus at the same time?
 
Yeah, but it would still be a military bonus, so I don't know if the AI would build it. Do they often build palisades and such? The +1 happiness for military state might work, but it might also be too specific.
 
Back
Top Bottom