Status of Civ 6 AI

[Kouvb593kdnuewnd]

I cannot understand why they have not taken Civ5 and try to work from there rather than scratch everything and be left with this mess.

Because civilization VI is not civilization V and have different rules. If you mean why they did not make civilization VI an improved civilization V, why would they?

Anyway throw in house rules that try to limit the advantages you have over the ai and the game may become a bit more enjoyable. To be fair civilization V and IV also have their weakness tied to ai.

Thanks for the update. Still sounds completely busted to me. I don't get why a large developer can't get a decent AI built. I'll continue to wait until it's fixed.

It is not going to be fixed because the problem is the game itself. Generally the simpler rules the game has the better the ai can play the game. Civilization VI is quite complicated with districts and all that not to talk about 1 unit per tile and the ai can understand all that unlike humans. Yes they can change the ai strategy to play like humans do such as spam archers and spam campuses but then the humans may just adapt to new strategies and so on.

Maybe civilization X or so will have great ai with even more complicated rules but that is a bit in the future.

 

[PhilBowles]

I don't know, I've seen a major improvement since before the patch. Playing as Nubia on Immortal, Gilgy actually managed to steamroll my captured Kumasi right after he plowed through two other city states, and although I managed to turn the tide and liberate Zanzibar, by Jerusalem my classical army was getting ripped by gilgy's brand new knights and crossbowmen. I was overzealous based on previous dealings with the AI, but there definitely seems to be improvement. And in my other game I've noticed the AI using air units a little better.

It's still not where I'd like it, but it's definitely better. As mentioned, the AI picking the best target to fire upon is still an issue.

No doubt there's an element of not practising what I preach in my analysis. Despite suggesting Archon_Wing increase the difficulty, I'm basing critiques on playing the game on Emperor (=King at best in Civ V) when I played Civ V on Emperor and Immortal. The reason for going down a difficulty is that I still don't feel I've played enough to have a good handle on the strategy side - despite having about 150 hours I think I've only completed two playthroughs due to boredom, and as I recall I narrowly lost both. The strategic AI is bad enough on Emperor that randomly clicking buttons in an unfamiliar system is enough to get you a close second place, but not quite bad enough that it's actually enough to win.

Nevertheless, it does mean I'm facing a below-par military challenge - the last development when I played a couple of days ago was Qin amassing for another unsurprise attack, prompting me to redirect my army from an attack on Kongo. Qin seems to be the largest power in my area and may almost rival my number of cities, but his military units are equivalent to or slightly below my tech level. Slightly better units actually in place - as I have horse archers and so haven't prioritised crossbows yet as I have no base archers to upgrade, and Qin has at least one crossbowmen - but no real threat.

Because civilization VI is not civilization V and have different rules. If you mean why they did not make civilization VI an improved civilization V, why would they?

Why wouldn't they? The alternative they chose was to make an inferior Civilization V.

The combat system is similar - the weakened cities in Civ VI disadvantage the AI, to be sure, but there's no reason the same code that worked for such things as target prioritisation in Civ V couldn't have been reused. By its end, Civ V had mostly fixed the 'not escorting settlers' issue as well, so it's not clear why this has resurfaced in Civ VI. Especially as there's a convenient new system to lock the escort and civilian as a stack.

 

[Archon_Wing]

Sigh. As expected, difficulty doesn't excuse the AI. Kept playing this emperor game and I actually just stopped here because honestly this feels terrible. (Though the lack of CQUI hurts) Saladin forward settles my capital and has a bunch of archers and warriors. Guess what he garrisons the city with. If he had put the archer inside he would have put up a long fight. But that's not the worst, I arrive to his main cities and this happens... really? There is no excuse for this.

He has no walls by this time and his defender comes out of the city... to do what? Besides die.

I wasn't even trying to rush if he hadn't forward settled me like a jerk. Pretty sure if I optimized my BO or picked my civ instead of randoming one, or picked the Forge pantheon, this would have been even worse. And we already had some no-name die offscreen by 2000 BC. Their build orders aka strategy are just fundamentally horrible. No wonder why they can't handle NPCs.

Now if you'll excuse me, I'm going back to Prince to build wonders.

 

[DizzKneeLand33]

Is it still that bad ?

I cannot understand why they have not taken Civ5 and try to work from there rather than scratch everything and be left with this mess.

Denkt, Acken wrote the mod that makes the AI an actual war machine in Civ V (and the reason I have so many hours of Civ V in), so his question has deep credibility. I think he is referring to why the tactical AI was changed and made worse, considering they could have instead chosen to build upon what was learned during the creation process of V and it's expansions. Acken's programming of the AI was superior. One case in point, I had a game where Montezuma declared war on me and I knew that ocean separated us and he wouldn't be a threat. I cannot even remember the amount of frigates and caravels he threw at me (it was over 12, and kept coming), but after he took FOUR cities with navy alone I finally regained my foothold. Tell me about your Civ VI experience like this.

It is almost like for VI they started with a whole new team, and that team decided that the other team didn't know anything about anything and that they would reinvent the wheel. Why not at least look at the superior programming of the modder community to see where improvements have been made, and incorporate those into the product (hey, they at least worked!). I am very disappointed with all of this. I can mod other things, but a poor AI is not one of them I can fix.

 

[Kouvb593kdnuewnd]

As I said civilization VI is not civilization V, yes they have similar unit system, they are still two separate games and just copy pasting the ai, even the tatical one may not work well because it may be unable to factor in the difference in rules such as the new movement rule that the whole thing may just be even more messier than now. Yes they should try to use what works but stuff may not be as simple as they seems.

 

[DizzKneeLand33]

It's not copying and pasting, but it is logic. I never said it was simple. But it should be reasonably achievable for a game company who already created a system. And if it indeed isn't achievable, then perhaps a different system should have been chosen.

But here's the real problem. We can obviously see that the game is not really getting playtested properly. How can you create a more effective AI if you don't know how the AI *should* play in the first place?

 

[Kouvb593kdnuewnd]

We can obviously see that the game is not really getting playtested properly.

Im not sure it is playtesting that is the problem, more likely is that the developers limited budget do not allow for much ai improvements that may get completely obsolete with the next patch/expansion. Modders don't have the same problem.

then perhaps a different system should have been chosen

Yes it may be a good idea to return stacks.

 

[antimony]

Somewhat off-topic, but related to ArchonWing's post, it seems to me that the AI will give up their cities more easily in a peace deal now. I remember games in the last patch where it would be hard to just get them to agree to cede the ones that you already occupy (not that ceding or not matters this much), now I can routinely get an extra city which is not even under siege.

 

[Archon_Wing]

Somewhat off-topic, but related to ArchonWing's post, it seems to me that the AI will give up their cities more easily in a peace deal now. I remember games in the last patch where it would be hard to just get them to agree to cede the ones that you already occupy (not that ceding or not matters this much), now I can routinely get an extra city which is not even under siege.

Can be good or bad, depending on the situation. It's a good idea if they are getting destroyed and want to survive, it's not good if they could just pay the attacker off and thus lose even more things in the process. Perhaps peace treaties could last a bit longer too. I'm not even sure how long they last-- 10 turns? I'm not too fond of the "hey my units have healed! Gonna attack again!"

It's certainly a difficult thing to balance. I would imagine your typical human player would consider giving up a non-junk city unacceptable, so it's hard for me to think of what would be the right spot.

But I think the greater problem is they don't seem to value their cities properly. You would expect them to give up worse cities first.

 

[Kouvb593kdnuewnd]

Just remove city trading from the game. Probably better to have the ai fight to the death than just giving up any chance to win anyway.

 

[Archon_Wing]

Well, it's a huge QoL thing though. It isn't really fun to drag outcomes when the late game is already really dragged out. And warmongering penalities are big in those cases.

Of course, my Civ 4 bias makes me want to bring back capitulation/vassals but with bigger penalties.

 

[Kouvb593kdnuewnd]

It isn't really fun to drag outcomes when the late game is already really dragged out.

But that is probably more to faulty game design than faulty ai. The later game is pretty poorly done in civ VI. They add stuff such as the neighbourhood and mechanized agriculture but they have not really made these stuff interesting or worth it. Why have they not made the later game about industrlization, like they could have made the industrial zone work like the neighbourhood and buildable in multiple copies so you get massive cities with huge industrial capacity instead of what we currently have is just about building as many cities as possible. Maybe they will do that in an expansion. Atleast civilization IV you is encouraged to change your economy but that is something they did not do in V and VI.

Civilization IV and V encouraged you to build great cities. Civilization VI encourage you to build as many cities as possible because each city allow for multiple copies of the same districts and there is not much reason to grow your population because it is so expensive for so little gain.

 

[Archon_Wing]

But that is probably more to faulty game design than faulty ai. The later game is pretty poorly done in civ VI. They add stuff such as the neighbourhood and mechanized agriculture but they have not really made these stuff interesting or worth it. Why have they not made the later game about industrlization, like they could have made the industrial zone work like the neighbourhood and buildable in multiple copies so you get massive cities with huge industrial capacity instead of what we currently have is just about building as many cities as possible. Maybe they will do that in an expansion. Atleast civilization IV you is encouraged to change your economy but that is something they did not do in V and VI.

Civilization IV and V encouraged you to build great cities. Civilization VI encourage you to build as many cities as possible because each city allow for multiple copies of the same districts and there is not much reason to grow your population because it is so expensive for so little gain.

Well, growing your population allows for many more districts until you get bogged down by amenities. The recent changes to reduce district costs will help. Late game also has stacking power plants which is kind of cool.

And I certainly prefer more cities for that great empire feel. Though I suppose it's not just about size, like what Trajan and Cleopatra think. IV supported both tall and wide, V was mostly tall, and VI is.... mostly wide atm

 

[Kouvb593kdnuewnd]

The problem is that the game encourage you to pack cities as close as possible which is called ICS. Yes the game don't really encourage ICS because at some point settlers get way to expensive but before that point it is pretty much ICS mainly because 2 cities mean 2 districts and 2 cities have more housing capacity than 1 city and neighbourhoods are really not far from being an expensive waste of resources. Population growth also get way more expensive the bigger a city population is which don't make much sense then there is really no exponential gain from larger population in a city. Housing limit and amenity limit should be enough to stop enormous population, they don't need that ridiculous pop cost at all.

Civilization IV also encouraged a large empire but without necessarily enourage you to build as many cities as possible which is a significant difference.

Late game also has stacking power plants which is kind of cool.

Did they not remove that in one of the patches? Anway district area of effect buildings is also one of those mechanics that encouraged and still encourage to pack cities.

Power would be kind of cool to have as a late game resource which could add amenties and be used for factories. But really the late game can be so much more developed and differentiated from the early game. That is one reason why Civilization IV is still fun because each part of the game is different which keep even the late game fun while V and VI it is often just tedious as nothing really new happens here.

Both previous games have had expansions to work on the later game so I assume the same will be done with VI.

However I think changes like these are a reason why the ai is not developed much in patches because a major change to the game is likely to need major changes to the ai.

 

[Archon_Wing]

Did they not remove that in one of the patches? Anway district area of effect buildings is also one of those mechanics that encouraged and still encourage to pack cities.

Oh, wait they did remove that. But the power plants are a good idea in general.

You were actually encouraged to pack cities in IV, at least early on, due to distance maintenance-- you paid a ton more for cities far away from your capital but almost none for cities near it-- and also the happiness cap means that spreading out too far would mean most tiles would never get worked. Overlapping cities was the fastest way to grow cottages too. In fact, ICS is a great way to deal with things if the surrounding land is bad or if you don't have much land period. Also remember that national wonders required a minimum of cities, so contriving an extra city or two was worth it. It is to the point where I can immediately point out a new player's troubles by telling them that their cities are too far apart.

Granted, you can't spam settlers like you can in VI though nor do you get penalized that heavily for overexpansion.

 

[PhilBowles]

Denkt, Acken wrote the mod that makes the AI an actual war machine in Civ V (and the reason I have so many hours of Civ V in), so his question has deep credibility. I think he is referring to why the tactical AI was changed and made worse, considering they could have instead chosen to build upon what was learned during the creation process of V and it's expansions. Acken's programming of the AI was superior. One case in point, I had a game where Montezuma declared war on me and I knew that ocean separated us and he wouldn't be a threat. I cannot even remember the amount of frigates and caravels he threw at me (it was over 12, and kept coming), but after he took FOUR cities with navy alone I finally regained my foothold. Tell me about your Civ VI experience like this.

It is almost like for VI they started with a whole new team, and that team decided that the other team didn't know anything about anything and that they would reinvent the wheel. Why not at least look at the superior programming of the modder community to see where improvements have been made, and incorporate those into the product (hey, they at least worked!). I am very disappointed with all of this. I can mod other things, but a poor AI is not one of them I can fix.

I never played with AI mods, but by the end of its life cycle Civ V's default AI wasn't far off being able to pull off the same sorts of naval attacks. Which makes it even less excusable that Civ VI can't.

Firaxis has hired modders in the past to lead Civ projects - why can't they hire some of the modders responsible for AI improvements?

 

[Acken]

Because civilization VI is not civilization V and have different rules. If you mean why they did not make civilization VI an improved civilization V, why would they?

Anyway throw in house rules that try to limit the advantages you have over the ai and the game may become a bit more enjoyable. To be fair civilization V and IV also have their weakness tied to ai.

-There's enough similarities between the 2 games to work on top of it. Also some AI-unfriendly changes are really questionable like harder movements. Civ4 to Civ5 was messy because of 1UPT. It was time to show they know what they are doing when it comes to AI because at least that part was similar.
-Easier to just not play Civ6.

Denkt, Acken wrote the mod that makes the AI an actual war machine in Civ V (and the reason I have so many hours of Civ V in), so his question has deep credibility. I think he is referring to why the tactical AI was changed and made worse, considering they could have instead chosen to build upon what was learned during the creation process of V and it's expansions. Acken's programming of the AI was superior.

Let's give credit where it's due though. Civ5 BNW AI is good enough that through balancing it can work (especially the whole melee vs range balance). My moding of the tactical aspect of it is minimal and if any has been copied from parts of Artificial Unintelligence from @Delnar_Ersike.
Most of my AI modding is more in the Building department and helping warmongers achieve their goal.

Im not sure it is playtesting that is the problem, more likely is that the developers limited budget do not allow for much ai improvements that may get completely obsolete with the next patch/expansion. Modders don't have the same problem.

Yes that is true. However, as a customer it's not my problem. I also understand that their business model may be better that way. I just rant as a fan of the series.

But that is probably more to faulty game design than faulty ai. The later game is pretty poorly done in civ VI. They add stuff such as the neighbourhood and mechanized agriculture but they have not really made these stuff interesting or worth it. Why have they not made the later game about industrlization, like they could have made the industrial zone work like the neighbourhood and buildable in multiple copies so you get massive cities with huge industrial capacity instead of what we currently have is just about building as many cities as possible. Maybe they will do that in an expansion. Atleast civilization IV you is encouraged to change your economy but that is something they did not do in V and VI.

Civilization IV and V encouraged you to build great cities. Civilization VI encourage you to build as many cities as possible because each city allow for multiple copies of the same districts and there is not much reason to grow your population because it is so expensive for so little gain.

If a problem, a lot of these can be fixed through modding rather easily especially if they give the dll source.
The AI however is another matter. Some modders can potentially do it but it's far from an easy task if the base code is really no good. Especially the 1UPT part. It's also daunting from a modder perspective that since they have changed the whole algorithm it requires to learn everything again...

Either way it comes down to 2 options regarding the AI. Fix it or give the dll.

 

[Archon_Wing]

I think the more complex movement (which I sometimes don't get myself!) is one of the big culprits, since that is a key change from 5 to 6.

And 1 upt is certainly part of it, but it's not the whole story, since not garrisoning proper defensive units, not aiming to build archers or walls, and not escorting settlers would be a huge mistake even if we didn't have 1 upt.

If the AI tries to send a huge army and fails to take my coastal mountain city, I wouldn't fault the AI because such a place is very hard to attack. But here, I think there's something fundamentally wrong with the AI itself.

And stop it with the forwarding settling unless they want to war. I don't know why 5 and 6 love it so much since it ends up in disaster as they cut themselves off.

Honestly, just have the damned things prioritize archery. If they meet another Civ, then they should get masonry. If their power sinks below a neighbor, build even more archers. It won't be optimal, but that's why they get bonuses.

And they should stop obsessing over religion. That's what got the Ai killed in 4 as well as it being kinda lame the player can't really go for religion as a result.

Expand, and then war because you can't expand. Don't just war because you have nothing better to do or because HURRR PLAYER HAS LESS POWER SUICIDE.

The terrible unit micro can stay the same. I just want to see the above. Is anything that unreasonable?

 

[Zuizgond]

and not escorting settlers

Yea. It's not even linked to 1upt and so easy to code that it's beyond me it's still a thing after 9 months...

 

[earlc]

I think I tried capturing a settler in the information era last game and it just disappeared. If that's the case, that kind of bandage implies there's a bigger issue with unit escorts.

That is what I would call stupid.

Just because they get bonuses to mask said stupidity, doesn't change that.

EDIT: Started an emperor game. This is a good example of something that would be a bad idea, independent of bonuses. Can you spot two things wrong with the AI already? And that DoF was just declared, so I doubt it was thinking that far ahead. (the DoFs are good though)

img

Sigh. As expected, difficulty doesn't excuse the AI. Kept playing this emperor game and I actually just stopped here because honestly this feels terrible. (Though the lack of CQUI hurts) Saladin forward settles my capital and has a bunch of archers and warriors. Guess what he garrisons the city with. If he had put the archer inside he would have put up a long fight. But that's not the worst, I arrive to his main cities and this happens... really? There is no excuse for this.

I wasn't even trying to rush if he hadn't forward settled me like a jerk. Pretty sure if I optimized my BO or picked my civ instead of randoming one, or picked the Forge pantheon, this would have been even worse. And we already had some no-name die offscreen by 2000 BC. Their build orders aka strategy are just fundamentally horrible. No wonder why they can't handle NPCs.

Now if you'll excuse me, I'm going back to Prince to build wonders.

I don't know man, I just had the opposite experience on immortal. Gilgy also had walls up on at least all his forward cities early, before I got them up on mine. But the point isn't that the actual functioning of the AI is coded to improve at higher difficulties, it's that it reaches a certain tipping point where the AI feels better being about aggressive and is more successful at doing so.

The no-name who died offscreen would be a city-state, which is pretty frequent on the higher difficulties. On King I think I've only seen, like, Freddy take over a couple city states, and that's usually later in the game.

And yeah that definitely is stupid, but it wouldn't be an issue if it was also balanced out by AI civs that have their crap together.