So apparently, the AI has not been improved

[Aristos]

It's why recently I was asking modders and other folks specifics (lack of modding access, exactly how to break down issues in high-level player / AI play). I feel those kinds of things are super-useful, heck, for me personally if nobody else (knowing why people have grievances is far better than just reading said grievances). I'm not here to say "this is exactly what you should say", that's a ridiculous thing to do. I'm just trying to contribute in the way I best see fit, and here I am arguing against ways that I don't think help.

Another reason why maybe it's not a bad idea to have a subforum dedicated to AI topics (and yes, complaints/reports/denouncements/whatever they want to call it).

 

[ohbuggerit]

I am long time civ player but only occasionally contribute here though I visit often. Civ is my favourite game franchise, but the poor ai of the past two iterations have made me fall out of love with the game. I so desperately want the game to be challenging again, but I feel it may not be possible eith the current desiign. To have the fear factor of civ 4 with the civ differentiation of 6 would be great. I do feel bit should be possible. Both the total war and paradox studios have made progress to the point where I feel challenged and it does not feel like the aigets insane bonuses. So sorry to be one more moaner but I am another person who feels wary about the expansion.

 

[Giskler]

When the DLL is released, there will be improved AI mods just like there was for Civ 5.

 

[Casworon]

If the DLL is released... I prefer Civ VI so nuch more than Civ V. If the DLL is never released it will be heartbreaking

 

[Leathaface]

I'm disappointed TBH, I hope the AI would have been improved a good bit, 2nd expansion and all.

Another AI complaining thread?
It has all been thrashed to death

Moderators, can we have an AI subforum, there are enough threads for this.

I like that there are AI discussion threads on here because Firaxis read these forums and hopefully they'll be aware of our thoughts.

If the DLL is released... I prefer Civ VI so nuch more than Civ V. If the DLL is never released it will be heartbreaking

I agree 100%. Looking at the state the AI is in right now, a mod like Vox Populi for Civ V would immensely improve the game.

 

[Cagarustus]

Improving AI is not cost effective for Firaxis. The best thing for them to do is give dll access to modders and let the latter code the AI to community standards. Heck, some modders could even make a buck out of doing it though patreon!

 

[Ferocitus]

We maybe need to just have one “Ultimate AI Thread”. Otherwise, this forum is going to look like the Steam Chatroom.

Good idea - the pre-triggered doomsayers and self-appointed industry paladins can repeat themselves in there ad infinitum.
Reminder to the AI ignorant: A much better AI is available - but you don't have the hardware, nor the sponduliks to pay for the electricity required to run it.

 

[Cagarustus]

Good idea - the pre-triggered doomsayers and self-appointed industry paladins can repeat themselves in there ad infinitum.
Reminder to the AI ignorant: A much better AI is available - but you don't have the hardware, nor the sponduliks to pay for the electricity required to run it.

But how were they able to achieve it in Vox Populi?

 

[Ferocitus]

But how were they able to achieve it in Vox Populi?

Easy. Because Civ6 is a different and far more complicated game.
The same code is not transferable.

 

[Elhoim]

Easy. Because Civ6 is a different and far more complicated game.
The same code is not transferable.

Except that the tactical part of the game is practically the same in both games, so there is definitely room for improvement there. In any case, time will tell, no need to talk out of our rears about what's possible or not without looking at the actual code nor having any of us done any AI code in Vox Populi. Only Gazebo and his team would know.

 

[Ziad]

But how were they able to achieve it in Vox Populi?

Years and years of development and testing by a group of very dedicated people.

Even then it's only balanced on very specific Standard settings. I prodded Gazebo for some balances to Marathon (around a year or so ago) and he made quite a few changes but it's not fair to demand even more testing for different speeds so I deal.

 

[Ferocitus]

Except that the tactical part of the game is practically the same in both games, so there is definitely room for improvement there. In any case, time will tell, no need to talk out of our rears about what's possible or not without looking at the actual code nor having any of us done any AI code in Vox Populi. Only Gazebo and his team would know.

I've done enough with engineering AI and optimisation to know that civ5 is an enormously simpler game, and enough about the "curse of dimensionality" that kills most amateurs ideas of what is possible. What's possible for that game is not directly relevant or comparable to Civ6. Bolting some successful part from VP doesn't mean that it will just slot into Civ6, R&F, or GS without a huge amount of testing to determine whether there are consequent bad side effects, and whether it can run in an acceptable amount of time per turn.

 

[oedali]

I also vote for AI subforum. Not only can we discuss positives AND negatives but can also run a community wish list of improvement areas for Devs to consider and we can do polls etc

During upcoming Q&A can someone ask when or if ever they plan to make the DLL available? If no answer that won't be a good sign..

 

[Gedemon]

Easy. Because Civ6 is a different and far more complicated game.
The same code is not transferable.

Not directly transferable, maybe, but as already said we have a resource that the devs doesn't: time.

If the VP tactical AI is better, then it can be reused, the rules didn't change much there.

Diplomatic AI is more a matter of design than AI.

As for "Grand Strategy", yes, a lot of testing may be required, but on that side we have numbers too.

 

[Ferocitus]

Not directly transferable, maybe, but as already said we have a resource that the devs doesn't: time.

If the VP tactical AI is better, then it can be reused, the rules didn't change much there.

Diplomatic AI is more a matter of design than AI.

As for "Grand Strategy", yes, a lot of testing may be required, but on that side we have numbers too.

Ideas about tactics and strategy can be transferred, sure, but you know that the details are the killer in these cases.
And any AI will be way behind human players in long-term strategies.
Again, I agree with you that the best resource we have for improving civ is the huge modder base with a lot of time that no company could afford to pay for their bug-hunting skills, new ideas, and the sheer will to implement them.

 

[Metecury]

Improving AI is not cost effective for Firaxis. The best thing for them to do is give dll access to modders and let the latter code the AI to community standards. Heck, some modders could even make a buck out of doing it though patreon!

It is if people will eventually stop buying their product.

 

[kaspergm]

I've done enough with engineering AI and optimisation to know that civ5 is an enormously simpler game, and enough about the "curse of dimensionality" that kills most amateurs ideas of what is possible. What's possible for that game is not directly relevant or comparable to Civ6. Bolting some successful part from VP doesn't mean that it will just slot into Civ6, R&F, or GS without a huge amount of testing to determine whether there are consequent bad side effects, and whether it can run in an acceptable amount of time per turn.

There are some things that can still be tweaked that seem fairly binary (correct me if I'm wrong). One thing is that AI puts too much emphasis on not loosing units. This is on paper a sound strategy, but often hold AI back from doing actual damage to the human player. This is quite evident when you compare AI civs to AI barbarians: The barbarians will sacrifice any and all its units to kill a player unit, which makes the barbarians a real threat and nuisance. Imagine if the barbarian units could heal and get promotions ...

I know you can't extrapolate from just one point, but I had a very clear example of the AI prioritizing wrong in my last game (where the AI performed remarkably well all-round, and which I would probably have lost, if not some sort of mod error made me fail to finish it). Anyway, I was starting on eastern coast and soon found Australia to my immediate north-east. I decide to forward settle the only empty spot between our civs and quickly put up an encampment and build walls, because I know this will not please him. Sure enough, just before I finish walls, he shows up with 4 warriors, 2 archers and a battering ram(!) and DOW me. We're still in ancient era, and my only defenses are a slinger and two warriors. I manage to finish my walls, but he immediately destroys the walls on the encampment with one warrior and the battering ram, moving on towards my city. In a couple of turns, walls on my city are also down (battering rams are so stupidly overpowered, and this is first time I've seen AI use it so effectively), and while I spam out warriors in Edingburgh to move in, he soon has my city down in red health.

At this point he has one or two archers left, and two warriors who are both down in red/yellow health. If he had been willing to sacrifice one of the warriors, he could have taken the city in the next turn. Instead he moves the warriors out and put them to heal, leaving the battering ram undefended. A few turns later, he agrees to make peace (costs me my only tea resource). If he had taken the city, he would have lost one warrior, but obviously this is a trivial cost to pay in return for conquering a city (and destroying my archer stationed in it). He would easily have been able to hold the city, and he would probably comfortably have been able to move on and conquer Edingburgh subsequently, once his units had healed and he had brought in a couple of new warriors.


(PS: Like I said, AI performed remarkably well this game. Australia never got off well, but on my continent, Egypt was massive tech leader, with Cree doing very well also. On the other continent, Korea and Greece (Pericles) were running miles ahead in science and culture, respectively, with Greece just being generally an era ahead of second AI. Later I found out that Pericles had Athens situated in a jungle/hilly region and had managed to build Hanging Gardens, Apadana, Colliseum, Chichen Itza and Estadio de Maracana all in Athens (screenshot), which would probably account for his massive run-away. This was on Immortal difficulty)

 

[Hatchie]

I would point out Marbozir and all streamers were given sort of early pre-release build (probably built sometime early January). It was feature complete, but still contained bugs (for example in his last lets play he was placing terrace farm early in the game but it told him it would produce coal, despite coal was far away from discovering and probably shouldn’t be mined by terrace farm). The AI quality was wildly fluctuating between the diferent official livestreams showing it was indeed worked on and it’s the sort of thing that is being polished last, after all features are done and at least rougly balanced. So I would believe the AI at release will be better than what Marbozir currently have (for example in the air units department)!

 

[L4Psha]

There also some kind of bug or something like that, i even wrote at tech support about this. It's that AI sometimes builds a ridiculous amount of support units - military engineers, supply convoys etc. Using that production he could build a formidabable army. And as saw in the stream that kind of behaviour is still there. Which also strange how it could be overlooked during QA or just how it can be logically explained if they think that this is a normal AI behavior.

 

[Gorbles]

There also some kind of bug or something like that, i even wrote at tech support about this. It's that AI sometimes builds a ridiculous amount of support units - military engineers, supply convoys etc. Using that production he could build a formidabable army. And as saw in the stream that kind of behaviour is still there. Which also strange how it could be overlooked during QA or just how it can be logically explained if they think that this is a normal AI behavior.

Something being noticed in QA, or being recognised as undesirable behaviour, doesn't automatically translate into the work hours required to address it. A bug being in the game doesn't mean QA missed it - this is one of the most misleading claims made about QA in general.