How is the AI?

[user330977]

Hello,

Has the AI improved much since release? I understand it was always one of the game's biggest drawbacks. Can they compete to a certain degree?

 

[chazzycat]

since release? it has been vastly improved IMO. but there are still issues. currently they aren't improving resources for example...

 

[Buktu]

currently they aren't improving resources for example...

As far as I know thats a corporations mode problem. In games without the mode the AI still improves resources as usual.

 

[Conquistador_514]

The AI can't fight itself out of a paper bag, about 10% better than after initial release but that's about it. As much as it pains me to say this, they need to make civ 7 with doom stacks like in civ 4, the devs can't be bothered to code a competent AI so they should just make a civ 7 with districts but also with doom stacks again, no more 1 unit per tile please, it was good concept that was squandered. Hopefully Human Kind is the game I always wanted civ 6 to be and we can all join the Human Kind Fanatics forum.

 

[Time Tested]

The AI can't fight itself out of a paper bag, about 10% better than after initial release but that's about it. As much as it pains me to say this, they need to make civ 7 with doom stacks like in civ 4, the devs can't be bothered to code a competent AI so they should just make a civ 7 with districts but also with doom stacks again, no more 1 unit per tile please, it was good concept that was squandered. Hopefully Human Kind is the game I always wanted civ 6 to be and we can all join the Human Kind Fanatics forum.

You're not wrong, of course.

That said, I'm always surprised at how effective barbarians are at fighting compared to civs. I feel like the tactical capability is there and I have no idea why it's not leveraged. The AI is strategically inept at war.

In my current game, 3 civs joined me in an emergency when Japan captured Valletta. The civ nearest Valletta (Zulu) did not even make an attempt to liberate Valletta. Instead, they made a push for Japan's capital with their entire army.... which involved them trying to cross two of Japan's border cities, along with 3 encampments, all with walls. Suffice to say that the army never made it to Kyoto.

 

[ThunderLizard2]

Funny. AI is garbage and Firaxis REFUSES to release dll.

 

[nauberry]

For me, it feels like the AI is very passive, when it comes to late game war. In early game, they make huge pushes and try to build large armies. In the late game, they build sometimes an decent sized army, but they don't really use it for anything useful. I have even tested it, by being in a joint war, bombing enemy cities to 1 hp, and still, my ally often refuses to capture the city for themselves.

Other than war, they seem to be decent in most district placements and can definitely win a science and a diplomatic victory. On huge maps, religious victories are rare, but It's very tedious even for a human player. Culture and domination victories I have never ever seen.

 

[aieeegrunt]

We don’t need doomstacks back unless we also have ZOC and a doomstack surrounded by ZOC dies, which is how you get realistic tactics.

Allowing one unit of each class to stack in a hex, and increasing movement by one fixes most of the UPT problems

 

[kaspergm]

That said, I'm always surprised at how effective barbarians are at fighting compared to civs. I feel like the tactical capability is there and I have no idea why it's not leveraged. The AI is strategically inept at war.

I think this is related to how they tried to give the AI civs a sense of risk assessment and caution. Basically, the barbarians will always attack you headfirst, no matter the consequences for their loss, which is what makes them such a nuisance, because when they come at you in hordes, they may not be able to defeat you, but they may well take out a couple of your units while they lose a dozen of their own.

I also think it's well documentet that this is why AI is so passive in late game: They are coded to need a huge military upper hand in order to engage in combat, and when they keep throwing their own units away in petty combats against city states and at the same time can't keep up with you in science, they are rarely going to dominate you in military score.

We don’t need doomstacks back unless we also have ZOC and a doomstack surrounded by ZOC dies, which is how you get realistic tactics.

Allowing one unit of each class to stack in a hex, and increasing movement by one fixes most of the UPT problems

I'm not sure if this is the place to discuss this, but I'm hoping for something similar to what I think Humankind has, which is grouping units while moving them, and then spreading them out when engaging in combat - not unlike games like Heroes of Might and Magic. Civ6 sort of tried this a bit with the corps and army idea, except it was a weird intermediate solution which didn't really do anything other than shift the combat strength by 10 and 20 points in late game, and resulting in you having to use a lot more production/gold/faith to produce new units (armies), where in fact units were already too expensive to produce.

 

[Gulathar]

AI is still garbo, biggest quality of life change if they manage to make it smart, or just let the modders at it.

 

[aieeegrunt]

I think this is related to how they tried to give the AI civs a sense of risk assessment and caution. Basically, the barbarians will always attack you headfirst, no matter the consequences for their loss, which is what makes them such a nuisance, because when they come at you in hordes, they may not be able to defeat you, but they may well take out a couple of your units while they lose a dozen of their own.

I also think it's well documentet that this is why AI is so passive in late game: They are coded to need a huge military upper hand in order to engage in combat, and when they keep throwing their own units away in petty combats against city states and at the same time can't keep up with you in science, they are rarely going to dominate you in military score.


I'm not sure if this is the place to discuss this, but I'm hoping for something similar to what I think Humankind has, which is grouping units while moving them, and then spreading them out when engaging in combat - not unlike games like Heroes of Might and Magic. Civ6 sort of tried this a bit with the corps and army idea, except it was a weird intermediate solution which didn't really do anything other than shift the combat strength by 10 and 20 points in late game, and resulting in you having to use a lot more production/gold/faith to produce new units (armies), where in fact units were already too expensive to produce.

Is this the mechanic where you move a stack on the map, and then when combat is initiated you “zoom in” to a tactical map where the stack spreads out and you move and fight units individually?

I can see this being interesting, but it’s a pretty big change for Civ

 

[Gulathar]

Just make unit stacking possible without stacks of doom, you could limit per type of unit like how it works now with support and civilian units, otherwise you could add an attrition mechanic like in Paradox Interactive games so that one tile can only sustain a certain amount of units before they lose health.

 

[kaspergm]

Is this the mechanic where you move a stack on the map, and then when combat is initiated you “zoom in” to a tactical map where the stack spreads out and you move and fight units individually?

I can see this being interesting, but it’s a pretty big change for Civ

Yes, something like that. I'm not sure exactly how Humankind handles it, but I think it does something a bit similar from what I saw in some of the youtube videos. Obviously doing something like that would be a major change for the Civ franchise, and I'm not myself certain it's the way I want it to go - I can see a lot of advantages, and some disadvantages. One thing that's certain is that if something like HoMaM battle mode is introduced, it needs to have a limited number of battle-turns for each game-turn to allow to bring in support units.

 

[ThunderLizard2]

Just make unit stacking possible without stacks of doom, you could limit per type of unit like how it works now with support and civilian units, otherwise you could add an attrition mechanic like in Paradox Interactive games so that one tile can only sustain a certain amount of units before they lose health.

I'm been saying this for years. 1 UPT was a mistake.

Yes, something like that. I'm not sure exactly how Humankind handles it, but I think it does something a bit similar from what I saw in some of the youtube videos. Obviously doing something like that would be a major change for the Civ franchise, and I'm not myself certain it's the way I want it to go - I can see a lot of advantages, and some disadvantages. One thing that's certain is that if something like HoMaM battle mode is introduced, it needs to have a limited number of battle-turns for each game-turn to allow to bring in support units.

It's looks like HumanKind might find the right balance on this.

 

[oSiyeza]

I'm been saying this for years. 1 UPT was a mistake.
It's looks like HumanKind might find the right balance on this.

1UPT brings complex combat tactics to the game, instead of simplistic aproaches of piling numbers, which is the last thing Civ needs more of. So I cant disagree more.

Also the problem with AI is not 1UPT, the problem is simply that Fxs did not care to code a good AI. When you implement a thousand game systems, and do not care about making an AI able to handle them, you end with a bad AI and a bad game.

Honesty, wishing for a Civ game where the mechanics require so little tactical thinking that a bad AI can handle them is not what I want for the franchise. Hearing people advocating for this "solution" is something that does not enter my head.

We should ask for a good game, not for a game that is simple enough that would work even if its made poorly.

 

[lordhaw]

Generally, from what I've seen it's a lot better than it used to be. Though in my current game I've got very large world wars going on since the classical era between all parties that just happened and have been going on through two additional eras. However, the best that I've seen them do is the Zulus taking out a city state that had walls. But it held out for years against the Zulus and Scythians until they Zulus finally wore it down. The Civ to Civ wars have been lots of back and forth but no cities have exchanged hands. They seem a lot more aggressive than they used to be though, but they can't follow through on their wars too much. I have of course seen them take cities in other games but they just can't seem to use their forces effectively for the most part. I have seen barbarians and Free Cities do things better than the average AI Civ for some reason.

 

[konokoni]

I've been playing since release. It's better than it used to be. The degree to which it is better is debatable, but all aspects (diplomacy, tactics, empire management) have improved noticeably.

Can they compete to a certain degree?

Yes. King provides a decent challenge if you are new to VI but not to Civ. Emperor after you've learned the ropes and are working towards playing optimally. However, you can eventually outgrow Emperor and then you're into Immortal and Deity, both of which give the AI massive bonuses that aren't necessarily fun to overcome. You spend a large chunk of the game on the back foot before you finally catch up and surpass the AI. Once you've done that, though, they stand no hope of catching up. The fun at that point is in learning to play progressively more optimally and in achieving victory in fewer and fewer turns.

I agree with the discussions surrounding the AI's continued tactical weakness. I'm not certain removing 1UPT is the magic bullet required to fix the issue. I think doomstacks mainly allow the AI to leverage it's production bonuses on harder difficulties to mindlessly throw waves of units at you and hope to win by attrition. That may be more challenging than watching the AI struggle to shuffle its units around, but I don't think you can call that sophisticated AI tactical execution. Having said that, I'm not opposed to the return of stacks; I was OK with them then and I'd be OK with them returning if it would actually help the AI compete militarily. Or just for the fun of it, stacks were entertaining in their own way. I have to admit that the tactical challenge of unit movement and positioning is fun to me as well, it was definitely worth trying in V and VI.

Honestly, I find the AI's military management far less disappointing than the AI's empire management. Cities in strange places, cities with districts in terrible places (when better options are clearly available), cities with resources unimproved far past when they should have been, cities in the line of fire without walls even after war has been ongoing for many turns, the list goes on and on. Giving the AI ludicrous bonuses on harder difficulties really doesn't fix this problem, it just papers over their inability to compete throughout the course of a game. It also makes capturing AI cities less rewarding than it should be, since they have a decent chance of being painfully sub-optimal. Even if Firaxis would just take the time to have the AI optimally place districts it would help immensely. Look how well Korea does in regards to science, just because the AI is forced to place +4 adjacency Campuses. Now imagine every AI is able to assess the three tile radius of its cities, note the optimal district locations, buy tiles to place them optimally if required, prioritize districts that will get particularly good bonuses (weighted by that Civ's victory biases), clump dams / aqueducts and wonders to create adjacency options for IZ's and TS's respectively, etc. It would be a real game-changer and would help to keep the AI competitive into the modern era, injecting some fun back into the late game.

I'm not sure if the things preventing this are coding resources or computing resources (or both). Either could be fixed (hire more / better AI devs, offload AI to the cloud) but so far we've just seen improvements around the edges. Good, worthwhile improvements, but they don't address the heart of the problem. After four years, I think this is (mostly) the AI we're going to have going forward. But who knows what Firaxis has up their sleeve.

 

[Conquistador_514]

I think that the AI should be coded to always keep a ranged unit as a garrison in their frontier cities, just doing that alone would keep their military score consistently higher throughout the game. They also should never declare war in the mid-late game unless they have a very high military score, and they should prioritize destroying Barbarians over declaring a sense less war. Once that is done then who knows maybe they could just buff up the tactical AI a bit more and then maybe even allow for the AI to plan wider strategic war.. like imagine the AI being able to do cross continent efficient D-day type invasions, especially for important strategic resources, It's a dream I know but who knows, other games have had really bad AI the first few years of release and then been amazing Strategy games, like Rome 2 Total war,etc.

 

[kaspergm]

Honestly, I find the AI's military management far less disappointing than the AI's empire management.

I agree very much with this. The tactical AI still isn't brilliant, but since the last couple of patches I've seen the AI throw multiple sieges on city states where they will bring a good number of melee or mounted units and will actually deploy and use siege units. I've seen several city states taken down by a combination of a handful of Longswordsmen or Musketmen and two Trebuchets (I use the Steel and Thunder mod that adds these units). I've also seen AI use combinations of land and sea units to put a city state under siege.

That said, the AI still has some severe shortcomings. They can overpower city states by out-teching them, but they generally fail against major civs on their same tech level. To be fair, this is as much a question of walls being too strong (imo.) as it is a question of the AI being bad. They also seem to fail at regrouping and falling back when their initial assault is dismantled. These are situations where we still see the famed single unit rush, or even rushes using civilian units. Again, this is as much an issue with the AI as it's a question of game design: The fact that producing new military units take at least a handful of turns, and often significantly longer than that if you don't have a production-focused city, leaves the AI extremely vulnerable once you have cut their gold reserves and supply from them.

Wrt. district placement, there has also been improvements in this area. I actually see some decent district placements fairly often. They probably still don't understand to buy tiles to get optimal placement, but the nonsense placement we saw in the early days seems to be gone. I think the issue is not as much that the AI places their districts badly as it is that they generally build too few districts and/or the wrong ones (yes they now build many campuses, which actually helps them keep up in science, but they sure do love their encampments and holy sites still, not to mention entertainment complexes and renaissance walls (which may actually not be a bad priority to help them survive)). Again, I think their tendency to wasting military units futilely ends up acting as a major production sink that draws away production from more constructive elements like districts and buildings.

 

[ThunderLizard2]

Still doesn't use air units and forget naval entirely. We need dll released!