AI Wonder Building

[Izz]

Does anyone else get sick of the ai beating you to a wonder only to discover that the wonder was built by them pointlessly? The Temple of Artemis in range of only the camp it was built next to, Huey Teocalli built in a one tile lake, Petra built on the only desert tile in a city. All of us have had this happen when we were just a turn away from finishing it.

In my opinion, its frustrating in an unfun way. Its not just the normal frustration of missing your target; its the immersion breaking frustration of having the ai play a defensive metagame against you rather than an organic outgrowth of their situational interests as an empire. It breaks me out of the game and makes it all too clear that I'm playing against an ai trying to deprive me of advantages through metagame strategies rather than that I was outcompeted by an empire who also wanted the wonder.

Does anyone else feel this same frustration? Anyone know any mod that would tweak the ai's wonder building behaviour so that it will build wonders that will actually benefit it rather than spam them to stop you getting them?

[tempest.of.emptiness]

Wonders provide tourism later in the game, which means that no wonder is ever built pointlessly. Add to that the fact that you've denied the wonder to an opponent and suddenly it makes a lot of sense - as a human being - to build every wonder you can even if you don't get much benefit from its "intended" purpose. It makes sense to me that the AI would do the same thing.

Of course, reason aside, it is still very frustrating to get beaten to a wonder you wanted.

[Lemurian1972]

Among the many other things the AI doesn't understand, Wonder and District placement is just one. Just for example, keep an eye out for how bad the AI is at placing a Government Center.

[gimmethegepgun]

Huey is mainly for the yields from all lake tiles in their empire, not the amenities from being next to more lake tiles.

[Izz]

The AI in my game that built Huey had one lake, a single tile lake. Other than denying other empires Huey and future tourism for their pointless boondoggle, they received no benefits at all.

And as for the future tourism and advantage denial, those are metagame strategies. They are the kind of behaviour you expect from players trying to game the mechanics for every advantage; its explicitly gamey behaviour and if you play against humans that do that its part of the package. I don't play against the ai to see it try to replicate the metagame. Imo ai should play as if they an actual civilisation, not like they are trying their hardest to exploit the mechanics for advantage.

I will admit that that is my preference. If I wanted to just play civ as a competitive exercise than I'd go to multiplayer with its many human players who will act as players in a game not civilisations. The design philosophy of the ai seems to be in agreement with my preference though otherwise they wouldn't have tried to give the ai personality through agendas. Having the ai like you for building walls or choosing melee over cavalry is not in the spirit of optimal metagame strategy.

[leandrombraz]

They aren't trying to exploit mechanics, they are merely following their agendas and building wonders because that's what they value. Some leaders are more interested in wonders than others, either because of their leader agenda (China), or because of the Wonder Obsessed hidden agenda. These leaders will build wonders whenever possible. Other leaders will still build but will do less frequently. I assume leaders that are pursuing a cultural victory also have an increased interest in wonders, but I'm not sure of that.

If the AI meet the requirements of a wonder and are interested in building wonders, they will do it, regardless of the wonder effect.

[Izz]

I think that's a problem in itself. There should be a way for the ai to prioritise suitable wonders for them and ignore ones which aren't. It shouldn't be just Ai wants wonder > Build available regardless. Its a catch 22 to defend the behaviour like that; you are saying that the ai isn't acting strategically or metagamey because its actually just too simplistic and dumb to act in sensible ways. I'd just like to see the ai act with intention and purpose about wonders that actually fits what they are supposed to be emulating, ie historical civilisations.

[tempest.of.emptiness]

Wonders are weapons of cultural war. As long as the concept of cultural victory is in the game, it is in the interest of every civilization in the game to build every wonder that they can. Either the wonder will provide them with tourism to support their bid for a cultural victory, or it will provide them with tourism to help protect them from another civilization's attempt at a cultural victory while simultaneously denying that wonder to that cultural civ.

If you were playing against strategically sophisticated humans you'd be seeing similar behavior. Your opponents wouldn't avoid building Huey Teocalli because you have a better spot for it. Maybe the logic behind the AIs decisions is simplistic, but in this case the results are nevertheless realistic.

[Izz]

I think that misses two points I'm making. Firstly, I wasn't saying that me having a better spot for the wonder should prevent the ai building it, I was saying it is slightly strange for them to build it when they receive virtually no benefit themselves for the wonder. I'm not saying that because I have a 6 tile lake and they have a 4 tile lake, or that if I have 5 lakes and they have 3, that they should politely step aside and let me have Huey; I'm saying that if they have a situation like only have a single tile of lake in their whole empire that maybe the ai should be considering other options that would better suit it. Even just in the immediate term there are other wonders it could build, even within the same city, or it could use that time and resource cost to build up something else rather than pour a lot of resources into something that gets them no actual benefit. As it is, the ai seems to just flag wonder building and then select one at random that it is capable of building.

As for whether the behaviour mimics players, that also misses a point that I explicitly made. The ai shouldn't be mimicking player behaviour, it should be mimicking empire behaviour. The fact the devs gave the ai agendas and priorities outside of winning shows that the design philosophy leans towards giving the ai cultural and historical personalities, not aligning them to optimal metagame strategies for beating players. If we want to compete against empires playing that way we have multiplayer. Its the wrong meaning of realistic to say that the ai is mimicking player behaviour imo. I could be wrong but I don't think the devs intended the ai to be playing the metagame to screw players in this way, if they did it seems like it clashes with other aspects of their apparent design philosophy for the leaders.

[tempest.of.emptiness]

Originally posted by Izz:

Firstly, I wasn't saying that me having a better spot for the wonder should prevent the ai building it, I was saying it is slightly strange for them to build it when they receive virtually no benefit themselves for the wonder. I'm not saying that because I have a 6 tile lake and they have a 4 tile lake, or that if I have 5 lakes and they have 3, that they should politely step aside and let me have Huey; I'm saying that if they have a situation like only have a single tile of lake in their whole empire that maybe the ai should be considering other options that would better suit it. Even just in the immediate term there are other wonders it could build, even within the same city, or it could use that time and resource cost to build up something else rather than pour a lot of resources into something that gets them no actual benefit. As it is, the ai seems to just flag wonder building and then select one at random that it is capable of building.

I'll try one more time: even if the AI will get no benefit from the abilities of a wonder that they build, they should still build it if they can because that deprives their enemies of an advantage. It's as simple as that.

Originally posted by Izz:

I could be wrong but I don't think the devs intended the ai to be playing the metagame to screw players in this way, if they did it seems like it clashes with other aspects of their apparent design philosophy for the leaders.

Some time when the AI beats you to building a wonder, reload an old save and make some changes (prioritize production if it was not already done, or chop some trees) to finish the wonder one turn before the AI did in your original playthrough. If you do this a few times, you'll encounter some situations where the AI will still beat you to completing the wonder. (This situation has been complained about by other players in other threads in the past, if you prefer to search for those threads instead of trying this yourself.)

The explanation is fairly simple; the AI knows everything that you are building and calculates whether it can beat you to wonders. If it can, it will. So when you back up and alter your completion time it also recalculates and if it can still beat you to the wonder it will, by using the same tricks - increasing production or chopping trees.

It doesn't seem likely to me that the developers would have coded this behavior accidentally, which then belies the notion that the developers don't intend for their AI to metagame.

Originally posted by Izz:

As for whether the behaviour mimics players, that also misses a point that I explicitly made. The ai shouldn't be mimicking player behaviour, it should be mimicking empire behaviour. The fact the devs gave the ai agendas and priorities outside of winning shows that the design philosophy leans towards giving the ai cultural and historical personalities, not aligning them to optimal metagame strategies for beating players. If we want to compete against empires playing that way we have multiplayer. Its the wrong meaning of realistic to say that the ai is mimicking player behaviour imo.

This is a matter of opinion. I don't agree with it, but I respect your right to have a different view of how the game should be designed.

[leandrombraz]

Originally posted by Izz:

I think that's a problem in itself. There should be a way for the ai to prioritise suitable wonders for them and ignore ones which aren't. It shouldn't be just Ai wants wonder > Build available regardless. Its a catch 22 to defend the behaviour like that; you are saying that the ai isn't acting strategically or metagamey because its actually just too simplistic and dumb to act in sensible ways. I'd just like to see the ai act with intention and purpose about wonders that actually fits what they are supposed to be emulating, ie historical civilisations.

I'm not defending, I'm explaining how it works. Off course it would be preferable if the AI made more complex choices. There's a difference between wanting something and actually making it happen, though. The devs are certainly aware that this is a weak point in the game, this subject always come up and the devs read our feedback in forums. They know and they probably agree this isn't ideal.

[Aachen]

“Optimal metagame strategies for beating [human] players” is quite a tall order. I have a feeling you are underestimating what developing such an AI entails.

[Izz]

I was using hyperbole about the optimal strategies, I understand ai for games is notoriously difficult. I also admit that even saying that I'm still probably more ignorant than I know of just how difficult some of the changes that would be my preference are to enact. I'm mostly expressing frustration at a particular solution the devs have come to and questioning if its intentional or not, and asking for other people's opinions on it.

As for tempest's comment, I think it does come down to a difference of opinion in how we'd like the ai to act in our games. They are happy with or approving of the ai metagaming and see attempting to replicate player strategies/anti-player strategies. I would prefer to see the ai act less like its playing a game and more like the leader of an empire making decisions that a historical empire would make. That's just a matter of preference.

Its probable that the ai behaviour in the game is a compromise between the two positions, as well as what's implementable. Certainly there are improvements that can be made to the ai though and a few modders have attempted that. I know there are several mods which modify it pretty successfully. If anyone knows a mod that will tweak the ai behaviour more to my preference I'd appreciate a recommendation.

[GoldenTalon]

Originally posted by leandrombraz:

Originally posted by Izz:

I think that's a problem in itself. There should be a way for the ai to prioritise suitable wonders for them and ignore ones which aren't. It shouldn't be just Ai wants wonder > Build available regardless. Its a catch 22 to defend the behaviour like that; you are saying that the ai isn't acting strategically or metagamey because its actually just too simplistic and dumb to act in sensible ways. I'd just like to see the ai act with intention and purpose about wonders that actually fits what they are supposed to be emulating, ie historical civilisations.

I'm not defending, I'm explaining how it works. Off course it would be preferable if the AI made more complex choices. There's a difference between wanting something and actually making it happen, though. The devs are certainly aware that this is a weak point in the game, this subject always come up and the devs read our feedback in forums. They know and they probably agree this isn't ideal.

OK - but what's being done about it? Civ VI is done but not at all fixed and now they move onto Civ VII?

[leandrombraz]

Originally posted by GoldenTalon:

Originally posted by leandrombraz:

I'm not defending, I'm explaining how it works. Off course it would be preferable if the AI made more complex choices. There's a difference between wanting something and actually making it happen, though. The devs are certainly aware that this is a weak point in the game, this subject always come up and the devs read our feedback in forums. They know and they probably agree this isn't ideal.

OK - but what's being done about it? Civ VI is done but not at all fixed and now they move onto Civ VII?

They still working on Civ VI. There's another patch coming and possibly more paid content, we just don't know when it will be announced or what it is.

[rincewind]

Funny, in my current game I'm the one building Huey Teocalli in a single tile Lake.