[Vanilla] The AI still can't take walled cities

[Victoria]

I will not buy any Civ related product

Love civ VI but any related item is a money maker instead of a thing of value, I certainly agree with the above part of your comments.

 

[tekjunkie28]

Whenever the code becomes available, I'd gladly do a full rewrite of the AI code to make it more bearable. While the game is complex and large, it really isn't that hard to make something significantly better. It's clear the devs either don't have too much time dedicated to AI, have little experience with building bots, or have shot themselves in the foot too much with the architecture they use to keep the game moddable. I've written several bots for different games now, and these tend to look more competent than the civ AI does after just a week or two of work, while taking turns within 10 ms with hundreds of units and without resorting to code cheats like the Civ AI does. To be fair bots in this game are a ***** to debug because everything takes so long, so it'll probably take me a little longer than that for this game.

There's really just no good excuse for bots with deity level bonuses to be beatable by humans, even if the bots have to be kind of role-playing. I'm confident even a game without buffed AIs would be no cakewalk with just a bit more effort in unit control and building/policy choices, beating immortal should be nigh impossible without abusive cheese tactics.

I read an article not too long ago on the new AI dev working on GalCiv3 AI. He came in and made the AI substantially better win just a few weeks. I believe he stated that he doesnt like moddable AI bc it is generally much harder to make it competent for the developer. It shouldnt be a modders responsiblilty, I paid Firaxis and they should be the ones to do it right.

 

[Siesta Guru]

I read an article not too long ago on the new AI dev working on GalCiv3 AI. He came in and made the AI substantially better win just a few weeks. I believe he stated that he doesnt like moddable AI bc it is generally much harder to make it competent for the developer. It shouldnt be a modders responsiblilty, I paid Firaxis and they should be the ones to do it right.

And civ AI actually isn't very moddable either, despite their efforts. It allows some tweaks here and there, which while it can provide some minor gains, just isn't anywhere close to what you can accomplish by just coding it better to begin with.
For example they use a sort of priority system for unit moves. There's a list of actions bots can take, like "attack high priority unit", "use promotion", "attack city", "move to safety" where it always appears to just pick the highest priority action available. And modders get to shuffle the priorities around or remove certain actions (on a global basis, not in a way specific to individual units/situations). This offers some miniscule control for modders, but at a gigantic cost.
A simple priority list like that cannot possible come close to clever behavior. It completely ignores some of the most fundamental aspects you need to make proper decisions. Attacking a unit means your unit is now going to have to end the turn on some tile. This can range from perfectly fine (no enemies around) to absolutely horrendous (you just put your 6 promotion army with 10 health next to an enemy city). A priority list like that doesn't care, it just sees you can attack a unit, then does it because it's at the top of its list, goodbye army. Allowing modders to shuffle priorities around is completely nonsensical when they should've just made something that actually takes in important information and makes meaningful decisions. I mean yeah, mods can improve the game, turning "chase target" off completely made the AI more likely to take cities because it'd just get distracted all the time otherwise. But how is that at all relevant when melee units frequently completely ignore cities on 0 health because they literally don't care at all when city attack is not at the top of the list (and if you do put it on top, they will ignore all defenders/promotions/pillaging/running away).

The systems in place for 'operations', which are groups of troops with some goal, production choices logic, diplomacy, all of that is build on similar systems which are mod-wise exposed enough to be able to tell there's little else going on (otherwise my changes wouldn't do anything), but which aren't exposed enough to actually be able to do something meaningful with it.

In contrast, a game that does AI moddability right is Dota 2, it just has its own internal c++ bot code, but allows modders to turn the entire thing or sections of it off, and use your own lua code instead, complete with a decent API that allows you to write whatever you want. The bots written by the devs aren't constrained by moddability and also seem significantly more impressive, especially considering it's a harder game to code for. And modders still get the best too. You get to completely write your own thing, and people can download your bot using ingame tools and play against it. No dll hacks required.

 

[Lobox]

That's why I am not buying this DLC. I do not want new features if the AI is so brain stupid. Civ V AI was fixed by modders. The AI in the mod, even though not perfect, behaves good enough to enjoy the game and gives a lot more challenge etc.

Offtop

I also regret buying the Collector's Edition of Civ VI. It was so poor. Just compare it with other games collectors editions. Such a big disapointment. I personally think it was not worth it. Simple as that. I thought the album was big enough, but it was so small and pretty short. It contains little text and images. The coins are made of really low quality material, probably plastic etc. The box is so terrible. There is no box, actually. It is some sort of very thin cover that can get damaged just by looking at it. Paying so much for almost nothing. I feel deceived.

Offtop end.

I will not buy any Civ related product untill the AI is at least as good as Civ V Gazebo CPP Vox Populi Mod.

Compared to Blizzard´s Collector's Edition Civ 6 Collector's Edition is a bad joke. I regret I bought it and when the salesclerk gave me the box my first thought was that it was not worth the money.

 

[darko82]

Love civ VI but any related item is a money maker instead of a thing of value, I certainly agree with the above part of your comments.

Yeah, the thing is: as long as most people accept the current state of things, it will basically carry over to the next Civ game, which worries me a lot. We should end this so that both sides are really happy. And objective and constructive criticism is always good.

 

[Disgustipated]

I just wanted to share this picture as proof of sorts. Of course it is very imbalanced since Netherlands is very strong and Scotland is very weak at this point (Netherlands took their Southern city on the UK island earlier in the game), island civs just don't do well. But they at least had walls up, I'm not sure what level of walls, certainly not city defenses with a city strength of only 66. I thought it was noteworthy because of how late in the game this is. The Dutch just wanted those cities. She took Robert out of the game. He was my scientific ally, but that's okay, I got my use out of him as you can see I'm about finished with the science victory at this point.

 

[liv]

They have plenty of battering rams but none of them used in the war
Poor Scotland. I hope they were happy at least.

 

[Leyrann]

I just wanted to share this picture as proof of sorts. Of course it is very imbalanced since Netherlands is very strong and Scotland is very weak at this point (Netherlands took their Southern city on the UK island earlier in the game), island civs just don't do well. But they at least had walls up, I'm not sure what level of walls, certainly not city defenses with a city strength of only 66. I thought it was noteworthy because of how late in the game this is. The Dutch just wanted those cities. She took Robert out of the game. He was my scientific ally, but that's okay, I got my use out of him as you can see I'm about finished with the science victory at this point.

Renaissance walls, as it's a fortification strength of 150. You can also see from how ornamental they are on the map.

 

[ThunderLizard2]

Yeah, the thing is: as long as most people accept the current state of things, it will basically carry over to the next Civ game, which worries me a lot.

Easy solution - stop buying game or DLCs until they fix AI or make moddable.

 

[Jackanape]

I just dumped on a game where the AI danced around my cities for 10 turns--able to take it at any time after the third--not taking the city. This was an AI civ and a city state. Both could have crushed the cities they chose to take.

I felt foolish for allowing myself to get I to that situation, but after that time passed, felt I was cheating by continuing. If I am unconquerable on Emperor, where is the challenge?

For what it's worth, I hadn't seen this since the expansion, but it still stinks...

 

[kb27787]

I just dumped on a game where the AI danced around my cities for 10 turns--able to take it at any time after the third--not taking the city. This was an AI civ and a city state. Both could have crushed the cities they chose to take.

I felt foolish for allowing myself to get I to that situation, but after that time passed, felt I was cheating by continuing. If I am unconquerable on Emperor, where is the challenge?

For what it's worth, I hadn't seen this since the expansion, but it still stinks...

Play deity instead of emperor then? They certainly did not hesitate to kamikaze in all their starting warriors into my capital until the damage is too much for the 20 hp regen per turn to absorb on t 10ish...

 

[MarigoldRan]

Yeah but after that early rush even on Diety, you're unconquerable.

 

[[to_xp]Gekko]

that's why you use the smoother difficulty mod

 

[Trav'ling Canuck]

Play deity instead of emperor then? They certainly did not hesitate to kamikaze in all their starting warriors into my capital until the damage is too much for the 20 hp regen per turn to absorb on t 10ish...

The behaviour Jackanape described would occur on Deity, too. The AI's logic for city attacks doesn't differ by difficulty level. Yes, the AI will happily take out your cities in the early game. City States seem happy to take out your cities at any time (anecdotal, they're on the same city attack logic as Major AIs, I believe, other than they will raze your city after capturing it; it just seems like they're more laser focussed). Later game, major civs sometimes fail to prioritize capturing vulnerable city centers.

The other thing they're still doing is weirdly timed peace offers. Poundmaker launched a war on me in my last game that I ignored. Then he showed up on the coast of one of my cities with three Ironsides, a Frigate, and a few Submarines. A few turns later and the city is out of Walls and Health and I'm organizing a re-capture force to take it back from him, when he offers me peace and a bunch of his gold. This is likely difficult to fix from a coding perspective, but he'd be in a stronger position to take my city and then offer it back as part of a peace deal that same turn.

 

[Duuk]

We've had the discussion a few times, and I'm about 90% convinced the problem isn't the AI's war/military logic.

It's the AI's warmonger avoidance logic. It doesn't want to take the cities because the warmonger penalty is too high. In the early game when warmonger penalties are low, the AI rampages all over the map conquering like crazy. Once warmonger penalties kick in they refuse to take cities or declare war any more.

To test it, someone should run a mod that completely removes warmonger penalties from the game (literally set all warmonger to 0) and run a game and see if the AI can win a domination victory on any difficulty level. I'd bet they get closer.

 

[Jackanape]

Yeah, I've gone the Deity route, but I honestly prefer a longer game, and in Deity it's just a bloody race to the finish...I've found Emperor gives the best balance.

I wish we could see a tactically intimidating AI, but for the most part I'm happy with it. This last game though really frosted me...my fault for playing peacefully, no doubt...LOL

 

[Trav'ling Canuck]

I wish we could see a tactically intimidating AI, but for the most part I'm happy with it. This last game though really frosted me...my fault for playing peacefully, no doubt...LOL

I'm finding the game refreshingly fun playing peacefully and trying to get by with whatever space I can carve out around my neighbours to fit cities into. It would be better still if the AI's military was as real a threat in practice as it looks on paper.

Right now I'm trying to win a Deity religious victory with Spain. Spending the early turns focussed on getting a religion really put me way behind, and I've got the most hill-less starting territory I've ever seen, with almost no forests to boot, mostly just plains and desert. By the time I was ready to expand, spending the Ancient era getting a religion meant I had a Classical Dark Age surrounded by Golden Age neighbours, further cramping where I could put new cities. This should be interesting.

 

[bonafide11]

Play deity instead of emperor then? They certainly did not hesitate to kamikaze in all their starting warriors into my capital until the damage is too much for the 20 hp regen per turn to absorb on t 10ish...

Yeah but after that early rush even on Diety, you're unconquerable.

Yep, even on Deity, if you survive their early rush, which is almost inevitable, they don't seem to pose much of a challenge as far as taking cities goes. Warriors are the AI's most scary offensive units. It's kind of sad really. I'm not going to say the Civ IV was great, but if you left your cities poorly defended at any point in the game, the AI could easily take it.

 

[Karmah]

A better test would be:
Start in the ancient or classical era
Give the AI ~20 turns to get some infrastructure built
Give it an army via the firetuner (I'm thinking 4 swords, 2 archers, 2 catapults and a horseman or so).
Wait until it declares war on you.
*Then* count how long it takes to capture a undefended city with walls.

Then this will test only how good the AI is at taking wall cities. At the moment you're also testing how much the AI wants to wage war, and it's pretty much impossible to differentiate the two phenomenon.

This !I also agree that the test is flawed. This setting sounds way better. As well as the metric (count only once he declares).