"Why Civ 6's AI sucks"

[Agent327]

Provocative title, I know. But this will be a serious discussion.

Civ 5 introduced a key change, that was perpetuated in Civ 6. That change was the switch from an isometric map to a hexagrammatic map.Now why is that important? Simply put, the new map design takes away moves per tile. Instead of 9 possible moves, we now have 7. Now this may suggest the game has become more simple, but in effect the opposite is true. With regards to the AI at least.

But Civ 5 introduced another change, which was perpetuated in Civ 6: the solution to the Stack of Doom in the most simple (and effective) way. No more than 1 military unit per tile. Again, this may seem to simplify the game, but again the opposite is true.

Why is this so? It is the intuitive thought to think the game has now become more simple rather than the opposite. But what we're discussing is the AI, not the human side of things. The reduction in possible movement combined with the removal of stacking means the AI should have been greatly improved to deal with thsi greater challenge. And it is here, that both Civ 5 and Civ 6 have been lacking.

The game as it is now poses a bigger challenge to the AI. and yet this was not accompanied by a hugely improved AI - neither with Civ 5 and not with Civ 6. (Although I would argue that the AI has improved in 6 over 5.)

I would say that the key problem is solving the Stack of Doom issue without realizing a better AI in the process. Perhaps this issue was not acknowledged by the developers, for whatever reason..It's impossible to tell, but it is an oversight.

But it only becomes a key issue when simultaneously replacing an isometric map design with a hexagrammatic one. Having less possible moves with no stacking ability poses, rather obviously, a bigger problem for the AI than for any player.

And now this becomes a major oversight, rather than a possible one. Because prior to these game changing new features should have come the development of an AI, at least up to Civ 4 standard, but essentially better. I think it's not an overstement to say that such an AI doesn't exist in either Civ 5 or Civ 6.

So, in essence, this is as good as it gets. Because with two expansions released and no major updates to the AI, I think it would be wishful thinking to expect the AI issue to be fixed in a patch or two.

One can only hope that Civ 7 will be better than the past two. and that will require some major changes. For the better this time.

 

[Bibor]

The AI cannot be fixed, not by much at least. All you can do is tweaks.

The game mechanics are simply not compatible with an AI. (That's not DeepMind, anyway). The list why this is so is endless: mountains, fractional movement, unit upgrades, ranged units.
The civilization franchise is stuck in its own loop by refusing to move away from these things, as if these are features that would somehow define the franchise. On that note, units could move just fine over mountains as recently as Civ4: Col.

Someone, somewhere in Firaxis insists that Civilization franchise is a board game transported to the digital world. But it's not a good board game and it's not a good computer strategy game. Both have their known limitations and the Civ franchise refuses to acknowledge them (and doesn't provide an innovative solution either).

 

[UWHabs]

1upt is a big change to the AI. While in the past, basically they could hide AI flaws by simply having them pile in more troops, and a simple AI of "keep all your troops together" was enough to prevent mass issues. The current model makes troops more valuable, so even if they get more of them, the tougher movement makes strategy hard with them. We saw that early on in civ 5 and 6 where the AI wouldn't know what to do, so they'd be attacking a city and then go embark their catapults. They're at least past that point now, now the problems are more about getting a proper combined warfare going, setting up your troops, and figuring out when to suicide and when not to. Too often they just plow into you and lose all their troops, or on the flipside, they're like 1 turn away from potentially capturing a city and decide to stop the attack.

There's also the flaws in that because the game is now more complicated in that you need to plan districts and place them, that's tougher on the AI as well. It's very easy for them to get too far down one path where they build holy sites everywhere but have no campuses, or neglect commerce hubs entirely, or something like that. They also often have some very strange district placement that I can't explain, putting harbors in the middle of nowhere where at the very least they should just place it next to the city for the base +2 bonus.

Now, can these fixes make it in in just a patch? I don't know. Personally, I do think 1upt is much better than stacks of doom, but perhaps the better fix for that is to ease the stacking rules more. Even a simple change where if siege units were treated more like support units, then it may not be as challenging to adjust the AI to bring a balance of siege and regular troops. You could potentially treat archers and ranged units the same way too, where they get their own layer, and again, that could be a change that would boost the AI as well, since they would have less times that troops block each other.

I know it's a very challenging game to get an AI for. As mentioned there are so many moving pieces and strategy components. Another big example is that one of their big philosophies this year is "play the map", where strategies change heavily depending on your map location. While that potentially can help the AI in that if they start in a mountain terrain with lots of +3 campus locations, logically they should build more campuses there, it also has the problem in now you have to design an AI that reacts to not just the players and situation around them, but also the map as well. Sure, if I'm building a campus better to use the +3 than a +1, that really shouldn't be hard for the AI. But when do you say, "okay, I don't need another +3 campus, this city needs a commerce hub first", or the "okay, by the time this city gets to pop 7 I want these 3 districts, so they'll go in these 3 places" whereas another city you don't plan that 3rd district because you know it's going to be 100 turns for it to grow that big?

 

[Kwami]

I think we've had about 300 topics concerning 1UPT and the AI. I'm not sure what this one adds.

 

[Aristos]

And here is the fallacy again: that 1UPT is to blame for mediocre AI.

And then some get triggered by the constant need to repeat the counter to this fallacy: two fans made a strong AI with 1UPT, the same engine of civ 5 vanilla, and their spare time. Thus, it can be done; ergo, 1UPT is not to blame.

I realize the repetition can become tiresome (believe me, it is tiresome for me to write it), but then, the repetition of the fallacy has also to stop, simply because there is a shining, living example (Vox) that shows that a good AI can be achieved under the same conditions of civ 5 and civ 6.

No more 1UPT fallacy, no more counter repetition. I am all for it.

 

[Horizons]

Do you remember Civ2? You could pile up units but the whole stack would be deleted if the best defender lost a battle (unless in a fort or a city). In practice the AI would send huge carpets of doom towards you even though it could stack if it wanted to.

I tried playing Civ2 again recently and found it very difficult. I've seen people play Civ2 as if it is child's play though. I suppose it's like with chess, I would be destroyed by even a moderately competent player because it's something I don't invest time or effort in. (I want to get paid for investing real time and effort, if possible ... )

And here is the fallacy again: that 1UPT is to blame for mediocre AI.

And then some get triggered by the constant need to repeat the counter to this fallacy: two fans made a strong AI with 1UPT, the same engine of civ 5 vanilla, and their spare time. Thus, it can be done; ergo, 1UPT is not to blame.

I realize the repetition can become tiresome (believe me, it is tiresome for me to write it), but then, the repetition of the fallacy has also to stop, simply because there is a shining, living example (Vox) that shows that a good AI can be achieved under the same conditions of civ 5 and civ 6.

No more 1UPT fallacy, no more counter repetition. I am all for it.

I believe Firaxis is more interested in investing resources on 'headline goodies' to get people to buy expansions, DLC than in making core improvements that only a relative minority of diehard fans will appreciate (especially since they will end up buying the game anyway).

 

[PhilBowles]

And here is the fallacy again: that 1UPT is to blame for mediocre AI.

Indeed. Civ VI AI remains worse than (unmodded) Civ V AI at almost everything that requires 1UPT, despite the systems being identical.

In fact, I revisited the forums to query an issue in my first post-patch game: Civ VI AI seems to have got worse, after slowly getting better over time to that point:

- Early in the game I came across a Mapuche settler on its own - there was a warrior following behind it, but not stacked with it. This hadn't previously happened with any frequency in Civ VI.

- I had a very passive neighbour in Lautaro, which might have been a good sign AI-wise - he wasn't close to me so it seemed to mean the AI rushing four units ahead unsupported issue was fixed. But when he did finally attack, he sent two catapults at me ahead of any other forces (and, incidentally, into my archers). His attack from then on was piecemeal, a unit or two at a time.

- By the time I dealt with his last units and moved on his cities it was the start of the Medieval Era. Lautaro had no walls, and nor in fact did any other city I'd come across by then (city-states included, though despite that they've stopped dying off in droves).

Has anyone else noticed the AI performing worse post-patch than it did before?

 

[Aristos]

I believe Firaxis is more interested in investing resources on 'headline goodies' to get people to buy expansions, DLC than in making core improvements that only a relative minority of diehard fans will appreciate (especially since they will end up buying the game anyway).

Or are we? I for sure finally learned my lesson. Funny thing is that I waited long enough to get Civ 5 at a huge discount, and when it was finished (and VP was already in development). I forgot my lesson when civ 6 came out, assuming that the "savior" of civ 5 would not fall into the same hole that his predecessor fell in. I was wrong.

Civ 7? Sorry, pass. I will wait for the entire set of reactions, especially here and especially from the long term veterans, and then I will consider.

In the meanwhile, I am doing all I can to "convince" other players in the strategy arena to "invade" this territory; it's here for the picking (looking at you PDX ).

Do I have a chance to "convince" someone in PDX to at least consider invading civ territory? Maybe a tiny little chance, but I will try nonetheless... perhaps that chance will slightly increase if more people suggest to them the idea, I don't know.

I will do my part though. This territory needs to be invaded, FXS needs to face real threats of losing the market, or else nothing will change (for the better).

 

[Kwami]

I can't imagine why you think Paradox would do any better. The AI in their games is at least as bad (especially the more recent games like Stellaris) and the games are at least as buggy, too. Plus, the prices that they charge for all of the content are just ridiculous.

 

[Aristos]

I can't imagine why you think Paradox would do any better. The AI in their games is at least as bad (especially the more recent games like Stellaris) and the games are at least as buggy, too. Plus, the prices that they charge for all of the content are just ridiculous.

Critic of the prices and DLC policies granted; you are right in that. But the AI in their games, looking at the entire version history, is in average much better than any of FXS' versions. That is a fact. You are right if using the latest Stellaris AI as a counter example, but that version is not representative of what it was before, and came together with huge changes to the mechanics which they obviously failed to train/code the AI on how to handle.

The huge difference is, they will train that AI and bring it up to the level it was before version 2.2. FXS will never bring their AI up to minimum standards; that has been already proven, and is a fact.

 

[Agent327]

So much fact, so much irrelevance...

And here is the fallacy again: that 1UPT is to blame for mediocre AI.

No more 1UPT fallacy, no more counter repetition. I am all for it.

As am I. As should be obvious from the OP. Which you, apparently, didn't read, because tiresome?

The OP doesn't 'blame' 1UPT for mediocre AI. That's not even the issue, nor the primary concern here.

 

[glider1]

Put it this way. There will be no excuse in the future for the next civ series to have the same or worse AI than Civ6. From now on there is only one valid direction, an AI that gets better after every patch, expansion and series. 1UPT is here to stay and so they should be able to build up the quality of the AI there is no excuses any more. We know that 1UPT can be made to be a good enough AI (VP mod). So Firaxis are definitely in the spot light on AI development and they can no longer hide away from it by adding "features".

Criterion for a good AI:

• It is competitive for VC which scales with difficulty (as it is in VP mod).
• It plays a tactical game that is sufficient to look as if a prince level human player made the moves. The tactical moves are coordinated and make sense. The tactics don't have to be brilliant, just sufficiently convincing and aesthetically pleasing to watch (as it is in VP mod).
• The strategic AI must be able to conduct multiple fronts simultaneously both in defense and offense because this has already been proven successful in VP mod.
• The AI must not declare repetitive wars just for the sake of limiting better civs because that is tedious even if it is optimal strategy (a problem in VP mod).

 

[Kwami]

Critic of the prices and DLC policies granted; you are right in that. But the AI in their games, looking at the entire version history, is in average much better than any of FXS' versions. That is a fact. You are right if using the latest Stellaris AI as a counter example, but that version is not representative of what it was before, and came together with huge changes to the mechanics which they obviously failed to train/code the AI on how to handle.

The huge difference is, they will train that AI and bring it up to the level it was before version 2.2. FXS will never bring their AI up to minimum standards; that has been already proven, and is a fact.

I'm not convinced. Have you seen the AI in EUIV, for instance? Vassals that breaks sieges a day before they're finished, AI stacks that just stand around doing nothing, etc. The AI is horrible with attrition, too. So much so that the AI has to be immune to naval attrition because otherwise they'd lose every ship! Largely, I suspect that the AI folks just can't keep up with all of the new half-baked features that the other developers keep adding to the games every 3-4 months. But also, it's probably hard to get everything working right with so many players and such a large map. Regardless of why, though, Paradox AI just isn't good. Would they do better with a smaller 4X game like Civ? Maybe, but I'm not optimistic.

I'm not trying to defend Firaxis here, though. They can both be pretty bad.

But do we really need another thread about AI? I'm sure that there are a few others around here somewhere...

 

[ashendashin]

But do we really need another thread about AI? I'm sure that there are a few others around here somewhere...

Gotta agree right now it gets kinda annoying when a hot topic keeps splitting up.

 

[Aristos]

I'm not convinced. Have you seen the AI in EUIV, for instance? Vassals that breaks sieges a day before they're finished, AI stacks that just stand around doing nothing, etc. The AI is horrible with attrition, too. So much so that the AI has to be immune to naval attrition because otherwise they'd lose every ship! Largely, I suspect that the AI folks just can't keep up with all of the new half-baked features that the other developers keep adding to the games every 3-4 months. But also, it's probably hard to get everything working right with so many players and such a large map. Regardless of why, though, Paradox AI just isn't good. Would they do better with a smaller 4X game like Civ? Maybe, but I'm not optimistic.

I'm not trying to defend Firaxis here, though. They can both be pretty bad.

But do we really need another thread about AI? I'm sure that there are a few others around here somewhere...

Glad you mention EU4. I just finished (literally, 2 hours ago) a TEU --> PRU game that was a blast (thanks civ 6 for pushing me back into EU4). I decided I wanted to make the Baltic into a prussian Unsersee (pun intended). I already did the TEU-->GER multiple times in the past, so I wanted to try something different for Prussia (the TEU start now is not as easy anymore).

I did not see any of the problems you mention with the AI, none of them. Granted, it's not a genius in absolute terms, but compared to civ 6 it is indeed. Sensible diplomacy, sensible tactics both at land and sea, sensible DoWs (or lack thereof), sensible alliances... all was there, hence the blast I had. I am also a veteran of EU since the first, so the "noob" factor isn't there either.

I truly had a blast, in big part because the AI made me feel I was really there, in the court of the Teutonic Order, and then in the Prussian Army State, plotting to make the Baltic Mare Nostrum, but facing a cunning AI at all times. I have not felt that yet in civ 6, sorry.

 

[Horizons]

! Largely, I suspect that the AI folks just can't keep up with all of the new half-baked features that the other developers keep adding to the games every 3-4 months.

Haha, good old Agile and 'minimally viable product'.

Firaxis certainly have a long history of releasing a 'minimally viable product' and then iteratively improving it. Heck, they even make it worse every now and then in order to get people to pay for the process to start over again.

 

[Victoria]

The AI may be better if we were happy to wait 10 minutes between turns but we are not.
The AI May be better is more people were happy losing but they are not.
The AI May be better is we were happy to pay £500 for the game, but we are not.

I remember a poll thread along the lines of what do you want next, more toys or fix what is broken... and the winning vote was more toys. And this was people in this forum.

Firaxis is a commercial institution, not a religious one.
These threads have dogma and to me it is not the popular choice but they are allowed their say
The sad thing is we would like it to be better so it’s not like we disagree

Title - “Why civ VI AI Sucks”
Opening argument - “The key problem is solving the stack of doom issue”
Then - “The OP does not blame 1UPT for the mediocre AI”

Yep, sounds like religion,
I am with you for a better AI but I somehow think civ VII will not be better.

 

[Horizons]

Can I suggest a couple of obvious improvements?

1) Don't send a vulnerable unit, on its own, right up to a fully fortified city (such as sending a rocket artillery army right up to a modern-age city that can one-shot it in the next turn)

2) When facing an existential threat, such as a full-on invasion, don't send units out to destroy distant barb camps

3) As above, any units that were sent out previously should be brought back to face the existential threat (being invaded and about to have your capital taken for example)

4) Don't let your embarked units sit around and be attacked by enemy ships

I don't have to pay £500 for that do I? If so, I can find more entertaining ways to have fun

 

[acluewithout]

I think we've had about 300 topics concerning 1UPT and the AI. I'm not sure what this one adds.

Yeah.

There was that one really cool AI thread where a bunch of people discussed the pros / cons on decision trees and @Victoria talked about the AI's memory.

The rest of these threads have been a drag.

 

[Horizons]

I think you need to pin those threads and forbid any further discussions, because it is a nightmare to find these old discussions for those who aren't posting a lot.