Why Civilization 6 AI Sucks - Part 2: Settling
- Thread starter Zike
- Start date
SupremacyKing2
Deity
I agree that generally, you want to settle adjacent to fresh water for max start housing. However, I think it is still ok to settle in areas with no fresh water when you have a lot of great resources. There are other ways to get housing than just fresh water. The other resources might be worth it. In my current game, I think I found a spot with 3 wheat and 1 rice. There is no fresh water but the abundance of food resources made it a prime spot for me.
Disgustipated
Deity
I have had the AI settle Kilamanjaro, about 3 turns before me which pd me off. Of course then they stupidly put a holy site next to it. As you can guess, I razed the city when I took it.
What you aren't mentioning is the AI likes to keep their cities in clusters. You can clearly see it on the video in all your examples. Especially important in R&F. It's obvious they value this more than water (which they don't value at all obviously)
What you aren't mentioning is the AI likes to keep their cities in clusters. You can clearly see it on the video in all your examples. Especially important in R&F. It's obvious they value this more than water (which they don't value at all obviously)
Eliminator_Sr
Prince
- Joined
- Dec 5, 2016
- Messages
- 566
Wow that's pretty awful - almost like they programmed the AI to actively avoid good locations.
Disgustipated
Deity
I can't dispute the video above, but I just loaded up my game, and yes there were some bad spots, especially be Egypt. And Georgia put one city in the Tundra, but she's out of good city locations to expand to, I would do the same thing if I had a spare settler with no place to settle.
9 out of 34 is what I counted with all the cities I can see right now. 34 AI cities (some were conquered by me in my Chandra game and I'm including them in the tally). 9 bad cities out of 34 which is about 26.4%. Egypt didn't have a lot of good locations either, they settled where they could.
The biggest problem I see is they quickly run out of space, and they put cities anywhere. Like the last video, the AI sees they can do something, so they do that something.
9 out of 34 is what I counted with all the cities I can see right now. 34 AI cities (some were conquered by me in my Chandra game and I'm including them in the tally). 9 bad cities out of 34 which is about 26.4%. Egypt didn't have a lot of good locations either, they settled where they could.
The biggest problem I see is they quickly run out of space, and they put cities anywhere. Like the last video, the AI sees they can do something, so they do that something.
GoatsHeadSoup
Chieftain
Has civilization ever had good AI outside of mods? I have played civ 4,5,6 and revolution and they have all had pretty crap AI.
Melliores
Warlord
- Joined
- Feb 9, 2018
- Messages
- 137
The AI looks to settle a close network of cities. As such, they do not need to grow them big and won't be building any housing improvements. I have also noticed the AI building a lot of farms around the Medieval era, so it does get a good 2 to 4 housing just from tile improvements.
If their cities have a lot of overlapping tiles, they won't be able to work more than 8-9 per city. Growing bigger than that can be counter-productive in terms of happiness. More cities also allow them to build more districts.
As noted, placement is the biggest consideration, not wonders or housing. And the AI does a pretty good job at that. I can provide a counter example to the loyalty point. I had a Cultural alliance with Sumeria and so they decided that it will be safe to settle in the small spaces between my cities. They were safe till the alliance expired. As I did not renew it, the two cities flipped pretty fast.
However, the point to make is that none of the other AI's tried to settle close to me, taking into account the fact that I will be applying pressure on the new city.
If their cities have a lot of overlapping tiles, they won't be able to work more than 8-9 per city. Growing bigger than that can be counter-productive in terms of happiness. More cities also allow them to build more districts.
As noted, placement is the biggest consideration, not wonders or housing. And the AI does a pretty good job at that. I can provide a counter example to the loyalty point. I had a Cultural alliance with Sumeria and so they decided that it will be safe to settle in the small spaces between my cities. They were safe till the alliance expired. As I did not renew it, the two cities flipped pretty fast.
However, the point to make is that none of the other AI's tried to settle close to me, taking into account the fact that I will be applying pressure on the new city.
isau
Deity
- Joined
- Jan 15, 2007
- Messages
- 3,071
I have mixed feelings about whether this is poor strategy. Its definitely not ideal in some cases. On the other hand putting cities close together often makes them harder to conquer. A city 3 tiles away eliminates many spots an enemy can place Siege units. So IMO it's not a total loss for the AI to cluster them together, even if it's annoying.
I do wish AI valued canals though. Seeing them settle one tile away from an important canal location makes me want to die.
I do wish AI valued canals though. Seeing them settle one tile away from an important canal location makes me want to die.
Last edited:
Disgustipated
Deity
Not too much harder to conquer, but they often put encampments which have their own ranged attack (with walls) in nearly all of those cities. So if you aren't careful, you could get triple or even quadruple bombardment, so it can be annoying conquering stacked up cities. Not to mention the loyalty pressure will be very strong until you get that second city. So it does have its advantages for the AI.
Not if it is a 4 pop city with no fresh water, granary or aqueduct.
kaspergm
Deity
- Joined
- Aug 19, 2012
- Messages
- 5,577
Wow, this is so depressing. The basically seem to have a fixed settle pattern? Settle first city, settle next city two tiles east + two tiles southeast from capital, settle next city two tiles east + two tiles northeast from capital, so you get a perfect triangle. Look at it, it's exactly what China and Australia does, and it's exactly what prompts the Zulu to plop that city down in the snow. I wonder what Zulu would have done next, now that the final city spot is blocked by a mountain. Maybe that's what gets their settlers to wander?
Holy horsehocky I didn't even notice this. They virtually do have a fixed settle pattern... They very frequent try for the triangle, same thing here in the france game. 1 tile from river because AI wants the triangle....
This explains so much. AI settling logic is based around grade 5 geometry..
PhilBowles
Deity
- Joined
- Nov 20, 2011
- Messages
- 5,333
I've seen the AI consistently settle Natural Wonders since R&F, often immediately adjacent to them. I've noted it specifically because pre-R&F, just as in Civ V, the AI not only didn't prioritise settling NWs, it appeared to actively avoid them. They won't actively settle them early as a human would, but in my games I've had an AI settle the exact spot I would have next to Yosemite (had I been close enough to avoid loyalty pressure), the exact spot I did (after my city was raised) next to Potopahi, and a tile or two away - but still within reach - of a spot I'd have chosen next to the Great Barrier Reef. I'm fairly sure they headed towards my Uluru spot at one point before I beat them to it as well.
When you ran down the housing issue I assumed the AI was simply coded with housing buffs, but the slow growth of the cities indicates this is evidently not the case. I can't understand Firaxis' reasoning here - now you may be overvaluing coastal tiles, since the housing bonus rarely compensates for the low yield of half of your workable tiles, but not settling rivers and more importantly not building granaries and aqueducts is inexplicable. Even if the AI ignored housing the granary is still a food production building - that should give it a reasonably high priority by itself.
For whatever reason there must have been a deliberate decision by Firaxis to ignore housing in AI decision-making, but if baffles me what the underlying reasoning could be.
The wander issue is likely to have been reduced by R&F simply due to loyalty (the AI still forward settles, often losing cities as a result, but not as frequently). Much of the time it seems to be the same issue as Civ V: the AI appears to set its destination when the settler is produced and can't adapt if a unit sits on the spot when they get there, or someone else has settled that area. They don't redirect the settler somewhere else, it just wanders around randomly a bit.
When you ran down the housing issue I assumed the AI was simply coded with housing buffs, but the slow growth of the cities indicates this is evidently not the case. I can't understand Firaxis' reasoning here - now you may be overvaluing coastal tiles, since the housing bonus rarely compensates for the low yield of half of your workable tiles, but not settling rivers and more importantly not building granaries and aqueducts is inexplicable. Even if the AI ignored housing the granary is still a food production building - that should give it a reasonably high priority by itself.
For whatever reason there must have been a deliberate decision by Firaxis to ignore housing in AI decision-making, but if baffles me what the underlying reasoning could be.
The wander issue is likely to have been reduced by R&F simply due to loyalty (the AI still forward settles, often losing cities as a result, but not as frequently). Much of the time it seems to be the same issue as Civ V: the AI appears to set its destination when the settler is produced and can't adapt if a unit sits on the spot when they get there, or someone else has settled that area. They don't redirect the settler somewhere else, it just wanders around randomly a bit.
Last edited:
PhilBowles
Deity
- Joined
- Nov 20, 2011
- Messages
- 5,333
The video points out at length that the AI doesn't use any of those other methods, however.
Makes sense for the AI. Holy sites get a large adjacency bonus for Natural Wonders. The first time I saw this, pre R&F, I was actually somewhat impressed because (a) it was the first time I'd seen an AI settle a Natural Wonder, and (b) it was the first time they seemed to have made a sensible decision based on adjacency bonuses. Given the choice I'd rather they were coded to do that where it's suboptimal than avoid doing it at all - and the AI now builds fewer holy sites than it did anyway.
It's never been good, but it's never been as bad on so many axes as it is in Civ VI. A lot gets excused based on the hard-for-AI-to-handle combat and (in Civ V) diplomacy systems, but these are errors of a type and magnitude that didn't happen noticeably in those games. Even when you couldn't see things like poor district placement (because those systems didn't exist), without looking under the hood you could tell their cities were growing at an adequate rate, and their progress towards victory was faster. In Civ IV and earlier you could see their city outputs and even the tiles they were working with spies.
It really isn't just a case of more complex mechanics and more visible features they can screw up like districts in Civ VI - the AI is objectively worse in many areas, including ones where the mechanics are largely unchanged from older games. Civ IV AI could handle health (though I think it got health bonuses that Civ VI appears not to based on these examples), which was functionally almost identical to Civ VI housing.
As depressing as the two videos so far have been, what's more depressing is that these are actually areas in which the AI is relatively strong. It seems to follow simpler rules across the board than in other Civ games, but in my experience (as one of the apparent few here who plays all Civ games without AI mods), its settlement behaviour tends to produce better results on average than Civ V, one of the very few areas where the Civ VI AI is now superior to its predecessor.
Last edited:
Similar threads
