Is Civ 6 Tactical AI the worst in the series?

[Askia Muhammad]

I don’t think the AI is “braindead” tactically. It’s not the greatest economic manager, but you can watch it pull off things like moving a group of melee ranged and siege units to a city and taking it.
The real issue IMO is that they don’t seem to keep building units during a war. People would lose a lot more on deity if this was the case.

I’m personally blown away when I see an AI Civ successfully take a city after the classical era. I like to spectate AI only games and I’d say it’s very common for AI to beat a city down and back off to pursue surrounding units instead of actually capturing the city.

I’ve probably spectated 200-300 hours of Civ with the auto play function enabled I’d say more often than not I’ve had games where 1) the AI never declares war after the classical era and 2) no cities aside from city states were captured the entire game.

I really think they need to divert some attention to making the AI capable of threatening a player. It’s nice if they can pursue a victory condition effectively (which I think has been improving) but it gets quite boring when every game turns into a turtle unless I’m the one starting the wars.

 

[JesseS]

If AI just kept spamming units like barbarian lunatics to push the advantage, combat would be (mostly) fine....At present the AI doesn't seem to focus enough on military units after declaring war.

Not really...the AI builds a TON of units. If you gain the diplomatic visibility to see their build queues or run a spectator game, you'll see that about half of their builds are military units of some sort, throughout the entire game. There are occasionally times when they don't shift production away from a wonder to units following a declaration of war, but this is more related to the "stickiness" of their build queues than failing to properly prioritize units.

The problem is that they don't coordinate the units well and lose them constantly. Also, by the time they get to units that require resources per turn, they inevitably build way more units than they can maintain, meaning that their tanks are worse than cuirassiers.

Some modest changes to the underlying rules that would greatly help the AI are:
1) reducing the production and direct gold cost of later units across the board. Maybe -15% for medieval and renaissance units, -20% for industrial, -25% for modern, -30% for atomic and later.
2) increasing the cost of upgrading units to align it with the typical 4:1 gold production ratio, instead of the current ~2:1 we have for upgrades. Then maybe nerf professional army to only reduce gold upgrade costs by ~33%.
3) reducing the penalty for lacking a strategic resource. -20 combat strength is extreme; it would be better to just stop units requiring a strategic resource from healing or being produced when lacking that resource. Maybe a movement penalty would be appropriate too.

I’ve probably spectated 200-300 hours of Civ with the auto play function enabled I’d say more often than not I’ve had games where 1) the AI never declares war after the classical era and 2) no cities aside from city states were captured the entire game.

The AI matches I've observed recently have had extensive, almost constant wars. But I have not observed them capture a walled city. And it is true that a human player can almost always avoid meaningful war declarations after the classical era.

 

[oSiyeza]

A problem here is how difficult is to take a city due to game mechanics.
Honestly the current system is a bit ridiculous in how easy is to put a single ranged unit in a walled city and being able to defeat anything thrown at you from the same era.

It is obvious that they went too far with city defense and some changes are needed. IMHO, the siege and bombard units should take far less damage from cities, range units should take a bit less damage, and only melee and cavalry units should take full damage from cities. But Im sure many other posible ways of balance exist.

Is also obvious that Fxs does not want to balance any of these systems, which makes it seem as this has been an intentional choice. Either they are ok with sacrificing any chance of competitive AI at domination as long as the player has also a hard time taking cities. Or they decided to sacrifice the domination victory entirely to turn Civ VI into more of a Sim-Builder-Management game, which would be a very weird decission since Civ VI is imo one of the weakests games on the series in empire management and diplomacy.

In any case, in the moment that I heard the AI designer say he had never seen the AI achieve domination ever, and that it was probably not posible; without him looking concerned and as if there was nothing wrong with that; was the moment I should decided to just not buy the game.

In the end I did, and now I only regret it from time time, maybe I just need to accept that good old civ days are over. Oh well...

 

[salty mud]

I've played Civilization for years and I've never seen an AI even come close to a domination victory, ever, in any game. Not saying that makes it acceptable, but to imply that this is a Civ VI only issue and "the good old days are over" is nonsense. AI has never put up a fight, militarily.

 

[Drivingrevilo]

I've played Civilization for years and I've never seen an AI even come close to a domination victory, ever, in any game. Not saying that makes it acceptable, but to imply that this is a Civ VI only issue and "the good old days are over" is nonsense. AI has never put up a fight, militarily.

^was about to say exactly the same thing.

There’s some very rose-tinted memories of past Civ games here among certain posters. I’ve played the series since Civ 2 and I’ve never seen a ‘runaway domination AI’, in any of the iterations.

Maybe it’s a case of ‘play a game long enough and you’ll see everything’. I played about 300 hours of Civ V, including ~100 on Deity, and I never saw the AI initiate a massive domination snowball. But on the other hand, I’ve played over 1000 hours of Civ VI and I have seen certain AIs conquer their entire continent; Mongolia, Macedon, the Ottomans, even Victoria’s England once...

So if you have 2000 hours on Civ V, you’re bound to claim that you regularly saw ‘runaway domination AIs’.

 

[oSiyeza]

I've played Civilization for years and I've never seen an AI even come close to a domination victory, ever, in any game. Not saying that makes it acceptable, but to imply that this is a Civ VI only issue and "the good old days are over" is nonsense. AI has never put up a fight, militarily.

Oh, in previous iterations, as far as I can recall all of them, AI wining by domination was posible (but I bet most players, me included rarerly stayed to see an AI dominating their entire civ and just surrendered and started again or loaded a previous game). Regardless, one civ giving you a very hard time after conquering half the world happened often in every previous iteration I can remember.

There’s some very rose-tinted memories of past Civ games here among certain posters. I’ve played the series since Civ 2 and I’ve never seen a ‘runaway domination AI’, in any of the iterations.

I have seen AI being serious contenders for a Domination victory in every previous civ iteration from 1 to 5. That means conquering several continets and wiping out several other civs.

I can atest to civ games of I, II and IV (I did not play much of III) were I never met half the civilizations in the map, cause they were exterminated before me arriving to the continent. In V was less common, but still I often ended a game with a big war between me an an AI powerhouse that was controlling most of the World Map.

I have never seen one in 6, and as far as I know nobody has.

As far as my experience goes, in lets say 100 games (close to 600 hours of gameplay, mostly on King). An AI conquering two other AIs is the most I have seen. I think It happened once. (Though I used mods to help the AI in most of those games in VI). In previous games, that scenario happened in every other match (and they did not stop there).

Also I dont remember ever the AI taking another civ in Civ VI after the renaisance, like ever. Hope someone has.

If Im wrong I would be extremely happy to see a game where the AI achieved something more than this, and at least know that is posible. So If anyone has, I hope they post it.

 

[Sostratus]

I have never seen one in 6, and as far as I can know nobody has.

If Im wrong I would be extremely happy to see it, and at least know that is posible.

I can’t think of any GS examples, but in vanilla I’ve seen a fair few examples of one continent being over taken by a Gilgamesh type AI. Much less common than 5 though, where Napoleon or Hiawatha would inevitably go manifest destiny on their continent eventually.

At least in the Xpacs even the Shaka types seem to just stop conquering.

 

[oSiyeza]

I can’t think of any GS examples, but in vanilla I’ve seen a fair few examples of one continent being over taken by a Gilgamesh type AI. Much less common than 5 though, where Napoleon or Hiawatha would inevitably go manifest destiny on their continent eventually.

At least in the Xpacs even the Shaka types seem to just stop conquering.

The taking one continent examples I know are usually done by conquering two cities and wiping out the only other civ on the continent in ancient age, and maybe even some weak city states. And then refusing to take advantage of the situation and not keep expanding. So I think having control over a continent is not so much testament of the conquer habilities of the AI per se.

 

[Victoria]

I've played Civilization for years and I've never seen an AI even come close to a domination victory,

bang on the answer along with @AriochIV saying they are incomparable bar a growth out of V.
The issue with walls is they have buffed them to defend against players who are advanced in science so they become tough unless advanced. Then there is just the pure processing power we do not even realise we are doing that the AI cannot do within the time limits. Then there is the singular nature of the AI aims which we are not subject to (no, I will not switch to building units as my wonder has not finished and has a higher priority) and the fact the AI has no memory, gives you all it’s gold before a war or mass buys expensive units then deletes units as it runs out of gold.

I do see some intelligence, it is far better than it was and I do lose units unless very cautious. All promising but no cigar, like every other iteration. I do not really mind so much, it’s just those carpets of catapults that get me down personally.

 

[kaspergm]

Main reasons why the AI is bad in Civ 6 are (already mentioned above):

1. Not cycling wounded units with fresh ones. It's either kill or be killed for the AI's front.
2. Walls and City defense are OP against the AI

As discussed in another thread recently, taking walled cities simply is too difficult imo. That goes for the AI as well as for the human, although humans will manage if able to pull ahead in technology. Walled cities should either have a massive defense and hit points or have a potent ranged attack, but not have both, which is currently the case.

 

[8housesofelixir]

Walled cities should either have a massive defense and hit points or have a potent ranged attack, but not have both, which is currently the case.

Realistically speaking walled cities did have a massive defense advantage against cavalries (see how Mongols struggled with Japanese/Hungarian walled defenses), but city ranged attack is something that didn't exist before the age of gunpowder. The in-depth defense of a city with fire support from the flanks - represented by the encampment ranged attack in the game - is also a very "modern" idea emerged in 18 century and only populated after the arrival of explosive shells.

In this sense, city/encampment ranged attack should be unlocked at a much later tech.

 

[Oberinspektor Derrick]

I liked the civ 5 take on city defenses. A city ranged attack should be baseline as it helps fend off early (especially ancient era) attacks and barns, but the ridiculous staying power of walls should go. There is too much defensive power coming from the jump from an unwalled city to a city with ancient walls, much greater than investing in an encampment, a fort or a new unit (things that should also matter). Civ 5 at least encouraged building forts and using units to prevent a surround on a city, but in civ 6 I feel like I often only need walls early and I'm fine (I never bother with encampments for instance, nor forts). Same for the AI, I don't like seeing them have crossbowmen when I'm attacking but I'll usually keep on pushing, but if they have walls I'll often just stop my early war right there and sue for peace for the time being.

 

[kryat]

Not sure I can agree with that. A while ago, I invaded the US (Earth TSL) with a pair of aircraft carriers fully loaded with bombers. They sent out fighters which played havoc with my bombers, taking a few out. I learned that fighters can also take out bombers still on board a carrier the hard way.

Of course, the reason why I sent two carriers full of bombers with no fighter escort in the first place was because I'd never come across AIs that have built fighters and used them to defend themselves, so certainly not deserving a high score, but slightly higher than 0 I think.

I’m glad someone is seeing them use air. I’ve never seen the AI use an air unit.

 

[AriochIV]

I’m glad someone is seeing them use air. I’ve never seen the AI use an air unit.

I've seen it three times now, and the last time it was actually used to some effect. I was at war with Brazil, and he had two or three Fighters, and used them rather effectively to kill two or three of my full Armies (Musketmen and Anti-Tank). I actually had to build AA units, after which he took some damage and the air strikes stopped.

It had been so long since I had built an Anti-Air Gun, I had forgotten what they looked like.

edit: Also, and I don't think this is an AI issue, but there is some real weirdness when it comes to air units in Aerodrome districts that are under attack from ground unit. Sometimes a ground unit can move into the Aerodrome hex and just sit there, and the air units in it continue to operate. Other times the air unit actually goes through a combat animation with the attacking ground unit, yet stays operational in the Aerodrome with the unit still sitting on it. One time (though this was some time ago pre-NFP, and probably pre-GS) I took the city controlling the Aerodrome and the air unit in it was STILL there, though I don't think it actually did anything, for several dozen turns. It seems like there still aren't clear rules in the coding for what happens when an Aerodrome is attacked an occupied.

 

[oSiyeza]

Realistically speaking walled cities did have a massive defense advantage against cavalries (see how Mongols struggled with Japanese/Hungarian walled defenses), but city ranged attack is something that didn't exist before the age of gunpowder. The in-depth defense of a city with fire support from the flanks - represented by the encampment ranged attack in the game - is also a very "modern" idea emerged in 18 century and only populated after the arrival of explosive shells.

In this sense, city/encampment ranged attack should be unlocked at a much later tech.

Ranged attacks and defense is ok. But the ranged city attack is too powerful against range units and bombard and siege units. Cities should not be more dangerous to catapults than catapults to cities.

 

[Linklite]

Ranged attacks and defense is ok. But the ranged city attack is too powerful against range units and bombard and siege units. Cities should not be more dangerous to catapults than catapults to cities.

In terms of humans rather than AI, I don't really agree. If all that's needed is one catapult to take out my walled city, why waste all that production on walls? It would also make Domination ridiculously easy; the only reason it takes me as long as it does is because these are two key points in the tech tree where my siege weaponry get outclassed by the walls and I have to pause to allow my siege tech to advance (the period just before my cars upgrade.to bombards, and then to a lesser degree just as my bombarda upgrade to artillery). Making siege weaponry individually outclass walls would just negate the point of them.

Maybe the AI could do with a boost or something, but for humans, Domination shouldn't be made any easier.

 

[kaspergm]

Realistically speaking walled cities did have a massive defense advantage against cavalries (see how Mongols struggled with Japanese/Hungarian walled defenses), but city ranged attack is something that didn't exist before the age of gunpowder. The in-depth defense of a city with fire support from the flanks - represented by the encampment ranged attack in the game - is also a very "modern" idea emerged in 18 century and only populated after the arrival of explosive shells.

In this sense, city/encampment ranged attack should be unlocked at a much later tech.

Ranged attacks and defense is ok. But the ranged city attack is too powerful against range units and bombard and siege units. Cities should not be more dangerous to catapults than catapults to cities.

I'm not sure I find that the idea of a city having a ranged attack without a military unit stationed in it makes much sense. Certainly not one that does as much damage as a contemporary military unit, if not more. I reckon that you can have civilian citizens man some defenses, but without proper trained soldiers, that would not have been hugely effective, certainly not at striking units at large distances (i.e. range 2 attacks).

I think a much better solution, both in terms of gameplay and in terms of realism, would be to have walls provide defensive bonuses as well as providing higher ground for a ranged unit stationed in the city, but having a city without a ranged unit stationed in it provide only a weak range-1 attack (i.e. representing whatever even an untrained man can throw against attacking units immediately below the walls).

Also one of my favorite mods from Civ5 was called Collateral Damage, which made any attack against a city also provide some damage to a garrisoned unit. So that if your city were under heavy assault, you might need to heal the garrisoned unit in order for it not to die.

 

[suedenim]

What annoys me with civ 6 (which relates to the AI waging war) is not necessarily that it's always bad at warfare in general, but that walls and city combat strength scaling in particular punish the civ 6 AI way too much.
As long as a city has walls, the AI doesn't seem capable of handling a well defended city.
The AI is pretty good at rushing down a city (it has enough units and will generally mass them close before DOW), but once walls (and a decently scaling city combat strength from a high tech unit) comes online, the AI can waste nearly its entire army through mindless suiciding right into the walls.
The end result is that I tend to never see any massive runaway domination civs like in civ 5, as the AI generally only manages to take a city or two (mostly those who have no walls, or where the target was far behind in unit tech, causing low city strength) before their advance stops.
Worse still, it can quickly lose the occupied city to loyalty pressure, negating any gains made and as such making a runaway domination AI very unlikely.

This is an issue I have with walls in general though, as once the game hits about the medieval era, even a human advance is very significantly slowed through walls and high city strength.
This lasts usually until the industrial age and flight, where I as a human player have balloons and later bombers to start blitzing again.
As a human player, I know full well how impregnable some cities are when all you have are a few catapults/bombards and a handful of units, and taking a well defended city can often take a lot of setup (moving in multiple catapults at a time for instance).
Sometimes I realize that I just have to give up on attacking the target alltogether because I won't break the city given my current tech level, causing a stalemate.
The AI doesn't realize this and keeps wasting his army, even though its handling of units is otherwise ok.
Anyhow, since the AI is neither adept at choosing to attack undefended cities (and target cities generally having walls after a point) nor using balloons/bombers later on, it hurts the AI disproportionally.

Personally I think the AI would be much more of a threat if walls were toned down significantly.
I want to see a runaway domination AI, but as of now I can't really recall the last time I saw one in civ 6.
The only runaway civs I usually see are peaceful science/culture runaway AIs, which is becoming boring.

As you say, Walls and Encampments are tricky even for a good human player to deal with. If you're not paying attention, it's easy to lose units in a crossfire, or to lose your siege engines through carelessness.

Which leads me to another point, that the AI is rather good when it comes to defensive tactics. It, wisely, does tend to focus on those siege engines (though you can sometimes lure it into attacking an easier-to-kill but irrelevant Crossbowman or somesuch instead of your Bombards.)

But the thing that I find most aggravating is that the AI won't attack an enemy city that's just sitting there defenseless. In the late game, I often reach a point where I don't particularly want to take an enemy's cities, but I do roleplay a good ally and try to help my allies to progress by killing units and knocking down defenses. But even when the ally has units next to the city and it's knocked down to no defense, it usually just ignores that gift and wanders off to do something irrelevant.

 

[Victoria]

I have been deep diving on Barbs lately. Here is a good example that they obey rules and also in a way that would be considered poor AI.
Barb spawns T1 to NW hex of camp as normal but does not move.
T2 it looks at each visible tile and choses one to move to. It does this in a single movement rather than a hex at a time. In this case below it reaches the target hex after only using 2 MP so re-asseses and moves to another tile (smaller yellow arrow)
On T3 it will move to a known hex in a single move even though it is not visible. (I see other examples on T3 of double moves to unexplored tiles.

When you move a scout you should always move it one tile at a time because otherwise you can get caught and mauled/killed in a ZOC. But the AI does not do this, you see a scout in game plough into your unit ZOC as long as your unit was not visible when the move started.
I just thought it was interesting because
a. The AI assesses only what it sees (I see this with other civs weights in the AI_planning.csv logs also)
b. It could have been written better for cost of more CPU cycles which is not a lot and coding wise just a repeat of the previous assessment.

Of course we do not always move 1 tile at a time but hey, that is our stupid choice.

EDIT:and for the real geeky amongst you ...

Spoiler :

The log AI_planning.csv plots out all the possible movements before randomizing where the scout will explore But the pink x squares for some reason it did not include.

 

[MrRadar]

Ranged attacks and defense is ok. But the ranged city attack is too powerful against range units and bombard and siege units. Cities should not be more dangerous to catapults than catapults to cities.

I think ranged city attack vs siege units is ok, it may also represent daring sorties. Bring in more forces to cycle and expect some casualties.

What is truly baffling though that a besieged city can still work the tiles occupied by enemy troops. Civ 6 is the first game in this franchise to allow that and to this day I am not sure if this was done by design (on what reasons?) or by some giant oversight which goes on too long already. I mean, before you could at least starve down the besieged city a bit. Now, even if you pillage, it still gets some production and food.