AI rarely captures cities?

[Karibskih]

Hello! I'm new to Vox Populi, played 4 full games. Huge map, 18 civilizations, 26 CS, warlord, normal speed. In all 4 games, AI civilizations are often at war with each other, but it was extremely rare for one civilization to capture cities from another, and even more rarely for one civilization to totally conquer another. Almost all wars between AI ended with a warscore in range (0-4). On average, about 4-5 cities were captured per game, and in 2 out of 4 games, one of the civilizations was completely conquered.

Is there a way to increase the frequency with which the AI conquer cities on huge maps other than by disabling other victory conditions besides domination? Maybe I should try epic speed, 22 civilizations and the maximum number of CS? Or should the difficulty be raised? I also think using really advanced setup to give one of the civilizations a few military units at the start of the game, but I'm not sure if this will help.

 

[andersw]

Hello! I'm new to Vox Populi, played 4 full games. Huge map, 18 civilizations, 26 CS, warlord, normal speed. In all 4 games, AI civilizations are often at war with each other, but it was extremely rare for one civilization to capture cities from another, and even more rarely for one civilization to totally conquer another. Almost all wars between AI ended with a warscore in range (0-4). On average, about 4-5 cities were captured per game, and in 2 out of 4 games, one of the civilizations was completely conquered.

Is there a way to increase the frequency with which the AI conquer cities on huge maps other than by disabling other victory conditions besides domination? Maybe I should try epic speed, 22 civilizations and the maximum number of CS? Or should the difficulty be raised? I also think using really advanced setup to give one of the civilizations a few military units at the start of the game, but I'm not sure if this will help.

I play mostly epic pangea and I would say it depends on the patch and a bit on rng, what agressive AI's how they are positioned and their neighbours, in a recent game (I think 2 versions back) strong early game Persia got wiped before I met them.

 

[Rosete]

I think AI not taking cities are a side effect of really good war AI. The AI are so good at defending that it is hard for other AI to break through. The AI will actively target range and siege units, so a weak civ can defend against a large army with very few units. It's not that AI isn't taking cities, the defending AI just play good defense.

 

[Recursive]

I think AI not taking cities are a side effect of really good war AI. The AI are so good at defending that it is hard for other AI to break through. The AI will actively target range and siege units, so a weak civ can defend against a large army with very few units. It's not that AI isn't taking cities, the defending AI just play good defense.

This has been my observation as well in test games. That said, it's possible the AI is being too cautious. Hard to say from only a single game.

 

[CppMaster]

Hmm, but that would mean that warmongering is not really feasible and it's better to just play defensive, thus reducing variety of strategies.

 

[Recursive]

Hmm, but that would mean that warmongering is not really feasible and it's better to just play defensive, thus reducing variety of strategies.

It's possible that it's simply too easy to defend under the current balance.

 

[looorg]

I don't think I have seen this issue. I have seen indications of it but they usually get around to it. The AI takes other AI (and certainly city-states). They might need some time to do it. It's hard for them early on to get their war on. But it could also be a matter of coastal/inland cities. The AI is a lot better at capturing coastal cities. But once they get like Frigate and such there is no issue. They'll just not blindly taking everything. They will raze cities to if they can.

They just are not that good at taking early inland cities from my experience. Also the wars have gotten to be time wise weird now so they might give up to early or they might not really be that into the war that was started and not really commit to it.

In some regard I think it might be cause they don't think they can hold the city. The city has no value or they have happiness issues and can't deal with having a drain on them. But it's a bit weird when you see them properly bombard a city down to capture range and then just give up and go home or just never do that final push. I try and do this from time to time with the city-states; if they are close I'll bombard the city low and hope/wait for them to capture it and they might be unwilling or slow to do it but eventually they usually come around to the idea and do it.

(edit) It could also be how AI make war. There is a unit in almost every tile. They get replaced more or less every turn. Citadels everywhere (which they are quite bad at disabling). There just isn't very much movement. This is more or less AI trench warfare ala WWI.(I'm Russia in this game so I'm not even in the picture except as scouts giving vision so this is just AI France vs AI Persia)

 

[Fluffball]

It's possible that it's simply too easy to defend under the current balance.

I have noticed very few, if any civs taking authority in my recent games. My general impression is progress seems to be the classic warmonger AI's choice of first tree at the moment.

This may have something to do with even the classic warmongers not being overly effective at taking cities as you really notice the difference those seemingly small bonuses you get from authority really make when you don't take it.

 

[Rosete]

I want to add that it also had to do with how cities are really tough to crack these days. City attack alone can ~3 shot a siege unit, so you need to constantly rotate them in n out because the AI has gotten better at targetting.

 

[youngsteve]

I have noticed very few, if any civs taking authority in my recent games. My general impression is progress seems to be the classic warmonger AI's choice of first tree at the moment.

This may have something to do with even the classic warmongers not being overly effective at taking cities as you really notice the difference those seemingly small bonuses you get from authority really make when you don't take it.

There are usually a couple who take Authority in my games, & all end up either destroyed, vassals or tiny empire by the time Industrial age comes around. Think there is a problem with Authority. They usually start well, menacing everyone, but by Medieval ages civs band together to hunt them down. Although they have military bonuses going authority, that is not much use if the other civs are techs ahead & have superior troops, which always seems to happen.

England is always bad, even when they don't go authority, is always isolated & taken down. Mind you not helped with Elizabeth being so obnoxious.

 

[andersw]

I want to add that it also had to do with how cities are really tough to crack these days. City attack alone can ~3 shot a siege unit, so you need to constantly rotate them in n out because the AI has gotten better at targetting.

I basically havent bothered until I have +1 range on my siege for some time.

 

[looorg]

I basically havent bothered until I have +1 range on my siege for some time.

Isn't that usually when the player wins? As soon as you have the extra range it's basically game over for the AI. Then it's just a matter of time.

But in general this is when the human superior mind and logistics kick in. Have some war-roads up and start quickly rotating units in and out of range and attack. Eventual win. While the AI sort of runs around and doesn't really know what to do or how to react to this.

 

[azum4roll]

The AI can't react to something they can't see. They don't seem to have deduction skills to figure out "there was a ranged unit shooting from that tile, so most likely it's still on that tile unless it has logistics".

AI vs AI war has to be very one-sided for them to start taking cities. Or they start a surprise war and take an undefended city.

 

[Amask]

I think it's a combination of two things.

1. AI is way too passive. They have a dozen units hanging back around their own city instead of sending everything they got in the area to flood the enemy.
Even when an AI is actively attacked by multiple neighbors they still rarely lose any cities. I don't care how brilliant of a general you are. There's no way you should ever be able to hold out against an army triple the size, barring extremely advantageous terrain. I've been playing on 5 billion worlds (more flat land, fewer hills), and even that doesn't really change things much.
2. Cities have absurdly high HP.

 

[Tekamthi]

I've noticed this as well -- not sure if this would help, but if other UU's could be found, maybe siege towers and battering rams could be made available to all?

 

[GoreBush]

First i think cities is extremely powerful stationary units itself. Second is one unit on tile system and inability to use big forces to break enemy on narrow piece of frontline. Things go slightly better in lategame with long range artillery and planes appearance. But overall much weaker player can defend himself against strong enemy almost infinite in a limited areas. Would be nice to see somewhat like armies of stacked few units like in Civ 6.

 

[notaspambot]

The biggest issue I've seen on this topic is that the AI simply can't deal with both sieging a city and fighting reinforcements off at the same time, iirc. They're pretty effective at taking undefended cities, or even human-owned ones, but struggles to deal with the swarms of units that other AI can put out.

Personally, I've been taking to slap a +75% city attack modifier on the AI in my games, which has made AI-AI wars much more volatile.

 

[Fluffball]

Trying to analyse the differences between my own gameplay and the AI gameplay and thinking what would be easy for the AI to understand. In bottleneck/chokepoint city cracking situations particulalry the main differences which the AI could be improved with are;

Use of citadels...The AI has gotten very good at the use of citadels when at peace with the player but i have never seen an AI place a citadel while at war with me and often one of the key points in me breaking through in a long siege in particular but also generally in breaking though to a city is the key placement of a citadel.

Forts...A subsection of citadels really, i have seen some amazing things the AI has done with forts when it comes to making canals, the most memorable being a deep land locked city in an area with a lot of lakes where the AI had built 4 forts to chain these lakes together and allow it access to the sea but i rarely seen the AI use forts for offense or even defense apart from occasionally city states. Another key to breaking a line is usually well placed forts which allow my ranged units in particular to hit important targets in relative safety when playing offensively but also allow my melee units to contrubute much more on offense as they no longer leave the fort when attacking and abviously on defense.

Placement of roads on the front line...Standard practice for me on any potentiall front whether offensive or defensive is to spam roads around the front, especially on slow moving/obstructive tiles so i can cycle units in and out easily. Never noticed the AI specifically bring roads to a front line for the purpose of attack or even defense. In the past i always had a rule of thumb of 1 worker per city but in more recent times i have a rule of thumb of 1 worker per city and at least two additional workers dedicated to building military roads and forts.

For the most part difficult sieges are usualy broken when i steal some territory with a citadel, build some adjoining forts and put roads under them all so i can move in my seige units and fire them on the same turn which means they can be used before they take heavy damage at least and need to be cycled out for a replacement or because their in a fort/citadel they can probably take 2-3 hits before they need to be moved to safety instead of moving in to range and at least one seige unit being nearly destroyed before it even gets to fire a shot.

If these tactics were used by the AI certain civs in particular would be much more devastating, thinking off-hand examples being Rome with legions that can build roads and forts and Portugal which gains a lots Great Generals via trade routes alone.

From a complexity point of view even using these features in fairly random fashion should improve the AI's ability to attack on average.


An additional difference which is a bit tricky as i think this has been deliberately nerfed to stop people being stuck in wars is the fact that i will usually try to sort my happiness so i can just stay at war as it often takes a long time for the attrition to dig in with me having highly trained units against increasingly quickly conscripted, lowly promoted units if not simply the AI starting to actually run out of units where as AI vs AI wars (seemingly) end quite quickly, presumably becaue the attacker is making little progress.

Not sure how this could be fixed without us going back to the unendable wars issue although we could probably start with using warscore and happiness. If the AI has ok happiness and has a higher warscore than the opponent they will seek to persist the war

 

[Karibskih]

The biggest issue I've seen on this topic is that the AI simply can't deal with both sieging a city and fighting reinforcements off at the same time, iirc. They're pretty effective at taking undefended cities, or even human-owned ones, but struggles to deal with the swarms of units that other AI can put out.

Personally, I've been taking to slap a +75% city attack modifier on the AI in my games, which has made AI-AI wars much more volatile.

Could you explain which file needs to be modified to do this?

 

[notaspambot]

Could you explain which file needs to be modified to do this?

My recommendation would be to make a promotion, and then have it apply to the AI through a dummy building via lua. If you don't want want to mess with lua, you can also try to make a cheap building that refunds its cost and have it apply that way, although you would then have to deal with it cluttering up your UI. Relevant base mod files you can reference include PromotionChanges.xml and NewBuildings.xml.