Is AI really brain dead?

This is one of the issues with Civ. Everyone has different desires of how the game is played.

  1. The Human's intelligent production empire plans vs the AI's difficulty based production bonuses
  2. The Human's intelligent strategic picks of techs and civics vs the AI's free techs and civics
  3. The Human's tactical use of era appropriate units and diplomacy vs the AI's hordes of units fueled by bonues
  4. The Human's tactical use of era appropriate units vs the AI's tactical use of units
This is how I see the current paradigm as well. But when I reach back to what I expected of civ when I first started playing (V vanilla) and had no idea of optimal strats (and how absurdly easy they make the game), the solution is obvious:

5. The human's empire plans vs "the empire itself", i.e. intrinsic, and very difficult stagnation mechanics requiring balanced focus - with AI invasions an added threat only on higher difficulties.

I don't feel like VI laid the groundwork to do this, though. Or to be anything but an excruciatingly slow city builder with distracting tactics chores laid on (I love 1upt in V but war and conquest had a dramatic, story-like pacing that VI's cartoon tedium just distracts from).

Edit: I think the reason reviewers and casual players are happy with VI is because they perceive the game as being in condition 5, where just the city building game is a dramatic challenge, and the AI threat is non essential to enjoyment. Well, I would love if I could play VI in the same perspective… but none of the mechanics have any huge impact on win ability; everything does nothing in a different way; the tech tree ends itself; there's a lot of thick-mud mechanics and not any "you overextended and now your empire will die" mechanics like V's global happiness, which is a better braking system because it can allow for a more active early game before applying the brakes; etc
 
Last edited:
So working like it does IRL then, where, for example:

I'm sorry, but I think you're coming from a place where you want to defend the AI at all cost and are trying to rationalize its erratic and often illogical behavior.

The AI will declare on you when there's no point to war. And even when there is a point or a chance for them to win, they can't seem to get the job done.

Some examples:
1. AI's declares when their territory is on the other side of the world
2. AI has no units in place to wage/sustain a war
3. AI declares, then sends no troops to my territory - then peaces out after minimum war-length, granting me hugely favorable peace-deals
4. Even if the AI has units in position to attack my capital, they shuffle those troops around for no reason instead of attacking

And denouncements regularly come without reasons in my games. Or they come because the AI automatically maneuvers itself into a diplomatic hole with me because it is a slave to highly specific likes/dislikes (agendas). I have yet to manage to cultivate long term friendships with AI players, something which was totally possible in Civ IV for example.

And generally speaking: Calling the AI in this game "realistic" because they behave in an erratic and downright idiotic manner is like calling a car with a broken engine "green" or "frugal". Sure: It won't use any fuel, but that's not because it's a marvel of engineering ... it's because it's broken.

S.
 
I'm sorry, but I think you're coming from a place where you want to defend the AI at all cost and are trying to rationalize its erratic and often illogical behavior.

Then you are completely wrong. I've admitted here and elsewhere that it is faulty in parts.
That doesn't mean all of its constituent parts is faulty as you suggested.


[QUOTE="Sascha77, post: 14552372, member: 291245"
The AI will declare on you when there's no point to war. And even when there is a point or a chance for them to win, they can't seem to get the job done.

Some examples:
1. AI's declares when their territory is on the other side of the world
2. AI has no units in place to wage/sustain a war
3. AI declares, then sends no troops to my territory - then peaces out after minimum war-length, granting me hugely favorable peace-deals
4. Even if the AI has units in position to attack my capital, they shuffle those troops around for no reason instead of attacking

And denouncements regularly come without reasons in my games. Or they come because the AI automatically maneuvers itself into a diplomatic hole with me because it is a slave to highly specific likes/dislikes (agendas). I have yet to manage to cultivate long term friendships with AI players, something which was totally possible in Civ IV for example.

And generally speaking: Calling the AI in this game "realistic" because they behave in an erratic and downright idiotic manner is like calling a car with a broken engine "green" or "frugal". Sure: It won't use any fuel, but that's not because it's a marvel of engineering ... it's because it's broken.

S.[/QUOTE]

I don't think it's realistic because Civ itself is not realistic.

You are over-blowing what I said. I did not say the game was "realistic" because it behaves erratically.
I said one aspect of the game, the seemingly irrational denunciations, could be seen as realistic
because of historical precedents.

And because it behaves in ways you don't like at the diplomatic level, that doesn't mean
it's other sub-systems are useless or broken.

It certainly does punish war-mongering heavily in some cases, and seemingly hypocritically. That's who
your opponents are. Shrug.
A nuke happy Gandhi is not realistic, and he might decide to nuke you for seemingly irrelevant reasons.
Tough: that's the Indian leader you have to face in the game.
 
Playing as France, so I know all the hidden agendas.

Long term friendly ai will denounce out of nowhere.

Long term friendly AI will declare both surprise wars and in some cases official war with no previous denouncement (not sure which CB, none seem to apply...)

The AI diplomacy-wise is inscrutable, it feels worse than when Civ5 was released. Seeming the modifiers makes it even more confusing because it makes no sense.

Why is denouncement even a thing? There seems to be no enemy of my enemy mechanics.
 
i've pretty much grown accustomed to pretty much ignoring the AI altogether in this
 
Another thing besides pointless wars/denouncements for no real reason:

The AI's tendency to aggro-settle.

In my current game (which I'll now abandon due to this), I have Teddy and Sumeria as my closest neighbors. Both have good lands available to them away from me, but decided to plop down cities right on my borders. Teddy went south of my empire, Gilgamesh went east. Both these settlements are at minimum distance from me, in pretty poor spots (no luxuries, not a lot of food, no fresh water) and in Teddy's case it's only his third city and miles away from his other two cities (20 tiles plus IIRC).

Teddy and I are friendly BTW and Sumeria was the one with the unavoidable early aggro in this game.

Both of them are developing religions, and both their cities are so close to me that me having a garrison in my city closest to theirs will trigger their "your military is on my borders"-whine. So that's two sources of constant conflict, since both are within pressure distance from my holy city and will get auto-converted.

And thanks to the insane warmonger penalties, I can't really do anything about this if I want to avoid going full-on conquest

The more I think about it, the more I come to realize that you can't really avoid wars in the early eras and the game seems to be geared towards this: Low to no penalties for early war, an AI that will nearly always declare during this time, or an AI that will actively box you in. Plus on higher difficulties (emperor and up), the AI-boosts are so insane that you actually have to take a few cities from potential runaway-AI-leaders if you want to stay in the race.

I do like conquest victories and playing aggressively, mind you.. but it seems pretty clear to me that my initial plan of building up enough science to stay competitive and enough military to defend/hunt barbs/deter the AI and then going after culture as soon as the Acropolis comes online probably won't work the way I imagined it.


S.
 
Another thing besides pointless wars/denouncements for no real reason:

The AI's tendency to aggro-settle.

In my current game (which I'll now abandon due to this), I have Teddy and Sumeria as my closest neighbors. Both have good lands available to them away from me, but decided to plop down cities right on my borders. Teddy went south of my empire, Gilgamesh went east. Both these settlements are at minimum distance from me, in pretty poor spots (no luxuries, not a lot of food, no fresh water) and in Teddy's case it's only his third city and miles away from his other two cities (20 tiles plus IIRC).

Teddy and I are friendly BTW and Sumeria was the one with the unavoidable early aggro in this game.

Both of them are developing religions, and both their cities are so close to me that me having a garrison in my city closest to theirs will trigger their "your military is on my borders"-whine. So that's two sources of constant conflict, since both are within pressure distance from my holy city and will get auto-converted.

And thanks to the insane warmonger penalties, I can't really do anything about this if I want to avoid going full-on conquest

The more I think about it, the more I come to realize that you can't really avoid wars in the early eras and the game seems to be geared towards this: Low to no penalties for early war, an AI that will nearly always declare during this time, or an AI that will actively box you in. Plus on higher difficulties (emperor and up), the AI-boosts are so insane that you actually have to take a few cities from potential runaway-AI-leaders if you want to stay in the race.

I do like conquest victories and playing aggressively, mind you.. but it seems pretty clear to me that my initial plan of building up enough science to stay competitive and enough military to defend/hunt barbs/deter the AI and then going after culture as soon as the Acropolis comes online probably won't work the way I imagined it.


S.
There are certainly some bloody awful starting positions dished up!
But that's the deal you got. This time they got the better position, another
time you'll have an advantage over them, and you will be able to play a
particular style and not the one that's the most appropriate in the circumstances,
and you'll win at that level.
Some are completely unplayable, or just don't look like fun.

I'm curious: suppose that in a galaxy far far away, the game was balanced, bug-free,
and the AI was working flawlessly but fairly.

What percentage of games would you think was fair to win at immortal level?
All of them with perfect play, or with a few blunders because you chose to
reroll until you got an advantageous starting position?
10% because Immortal is inherently difficult, even with a great starting position?
Something else?
 
They make the AI feel schizo. Diplomacy right now seems to consist of everyone being permanently furious with each other and no one explaining why. Not that you can do much even if you know why they are upset.
Couldn't agree more.

Agendas are too specific and, well, "luck-based". If you happen to be on the same continent as Vicky, she'll love you. If you dare to improve your first luxury when he still has none, Monty will give you an instant dislike (just happened to me today). As I've said in another thread: Too often the AI will maneuver itself into a diplomatic hole because it has to act according to those super-specific agendas.

I would prefer a system where:
1. Upon initial contact, there's a "grace-period" where agendas don't really matter. Heck, why not disable agendas until AI-players hit a certain era or tech? That would give you time to get on an AI player's good side before his inevitable dislikes kick in.
2. Agendas that are more based on *my actions* not on random factors like who is on which continent. You know? Stuff like warmongering, settling too close, adopting certain policies/governments, breaking diplomatic promises, etc.
But not stuff that will be unavoidable if the player is running a half-way decent game. I mean: Nobody in their right mind will pass on a wonder just because China is on the map. And nobody in their right mind will pass on Great People just to please Brazil, etc. They could still have a problem with that stuff, mind you.. but the hits shouldn't be so big as to make normal relations impossible.


I've been thinking about all this and the AI's ineptitude in war, and I started wondering how they built the AI. More specifically: I wonder if they had the AI in mind and whether it could cope with some of the new mechanics when they set down the new rule-set for the game. I don't know the first thing about creating an AI, so this may sound a bit naive: IMO, any changes to rules/mechanics should take into consideration how easy/hard it would be to write an AI that can master these new rules. I think it's fair to say that Civilization, primarily, is still a single-player game. So IMO any version of the game that has a bumbling AI or one that behaves irrationally, silly or plain incompetently can't be regarded as a success.

But it seems to me that revamping the AI wasn't very high on the priority list for Civ VI. Looking at the insane boosts AI-civs get on Emperor and up seems to indicate this. And cheating in this fashion feels like a crutch, and makes me avoid anything above King.

What percentage of games would you think was fair to win at immortal level?
All of them with perfect play, or with a few blunders because you chose to
reroll until you got an advantageous starting position?
10% because Immortal is inherently difficult, even with a great starting position?

Oh, I have no problem with getting my ass kicked on Immortal or Deity... that, and the fact that higher difficulties demand a much more "streamlined" style of play is the reason why I avoid higher difficulty. I don't want to be stressed out, trying to catch up to AI-players that start with multiple settlers, free tech and a couple more military units than myself - or who get insane production boosts.

Immortal should be friggin' hard, but from what I've seen so far in Let's Plays, people are beating the game on Immortal and even Deity pretty easily. One of the reasons being the AI's poor performance in war, of course.

That said: The situation I described happened on King and I've seen them do this aggro-settling on other difficulty-levels as well. I have no problem with the occasional "aggro-town", if the AI has a good reason to put a city in that spot (like lack of spots to settle in or getting a rare strategic/luxury resource). But plopping down cities in poor spots within minimum distance of another civ's capital when there's still tons of unclaimed land with better resources around.. well.. that's just annoying.
And it makes me long for the days of Civ IV where such cities could be culture-flipped pretty easily.. :D



S.
 
Last edited:
I've been thinking about all this and the AI's ineptitude in war, and I started wondering how they built the AI. More specifically: I wonder if they had the AI in mind and whether it could cope with some of the new mechanics when they set down the new rule-set for the game. I don't know the first thing about creating an AI, so this may sound a bit naive: IMO, any changes to rules/mechanics should take into consideration how easy/hard it would be to write an AI that can master these new rules. I think it's fair to say that Civilization, primarily, is still a single-player game. So IMO any version of the game that has a bumbling AI or one that behaves irrationally, silly or plain incompetently can't be regarded as a success.

But it seems to me that revamping the AI wasn't very high on the priority list for Civ VI. Looking at the insane boosts AI-civs get on Emperor and up seems to indicate this. And cheating in this fashion feels like a crutch, and makes me avoid anything above King.

I have used "AI" in applied maths and for enginereering design for about 30 years
and that still does not mean I know anything about the AI used in Civ.
There are too many possible techniques, and subtle and blunt variations.

I can say, however, that it is possible to do a lot of balancing and tweaking with
large amounts of data, and that's exactly what is being collected as people play the game.

Have a look at the log files and you'll see there's quite a bit of information that
can be used for a myriad of adjustments.
 
Have a look at the log files and you'll see there's quite a bit of information that
can be used for a myriad of adjustments.
That's what I'm hoping for.

Sadly, Firaxis seem to be eerily quiet on the matter at this time - I hope this indicates they're too busy tweaking the game to make any lengthy announcements. ;)

However: A simple acknowledgement of the fact that there *are* problem-areas would be nice.

S.
 
I must say the AI's performance in war is quite poor but also slightly unpredictable. In my Aztec King game there was a moment when I moved an injured musketman forward one tile too many and found it in range of FIVE full health crossbows. In Civ 5 my unit would be dead. In fact generally if you put a unit into an exposed position in Civ 5 you lost that unit (I certainly did anyway). Instead what happened was that three or four of those crossbows ran away. There doesn't seem to be any willingness on the part of the AI to fight a war of attrition that saps the human player's will to continue. In Civ 5 even weaker and smaller AIs would make you earn every yard of territory and peace deals were sometimes hard to come by. In my Aztec game Rome ceded Rome and Antium, and gave me Brussels and two other cities at the first time of asking. Why did Trajan take this attitude?

I know the AI *can* focus fire as it happened once later on vs. Kongo and also Sumeria did so against my naval units on my later Emperor game, plus the AI is capable of buying a heavy chariot in a city and ruining an besieging archer's day. So there is some aggression there albeit it seems modified out too much. They don't go after cities as they did for sure, at least not in Prince-Emperor so far.

Going back to Rome's crossbows running away and then Trajan giving me an overly generous peace deal - maybe combat performance is being overly affected by agenda and/or some kind of background leader 'state of mind' or attitude to an aggressive/powerful human civ that is overriding the basic 'make them bleed for every yard' mentality that AIs defaulted to in Civ 5. In my Aztec game Trajan was desperate to be friends with me from early on, probably saw what happened to Arabia. Even when I attacked Germany, who was his ally and around whose cap he had tons of legions, Trajan did nothing.
 
And generally speaking: Calling the AI in this game "realistic" because they behave in an erratic and downright idiotic manner is like calling a car with a broken engine "green". Sure: It won't use any fuel, but that's not because it's a marvel of engineering ... it's because it's broken.

S.

I just wanted to say this is awesome analogy :)
 
Edit: I think the reason reviewers and casual players are happy with VI is because they perceive the game as being in condition 5, where just the city building game is a dramatic challenge, and the AI threat is non essential to enjoyment. Well, I would love if I could play VI in the same perspective… but none of the mechanics have any huge impact on win ability; everything does nothing in a different way; the tech tree ends itself; there's a lot of thick-mud mechanics and not any "you overextended and now your empire will die" mechanics like V's global happiness, which is a better braking system because it can allow for a more active early game before applying the brakes; etc

Makes sense.

Many casuals and reviews wanted a civilization city builder with the option of war like a mobile game (such as Dominations). The AI is perfect for them.

Making a AI for:

Civ city builders
Civ base builders
Civ tactical wargamers
Civ strategic generals

might be tough.

It would be great if you could adjust the mentality of the AI to meet the needs of the different Civ players.

Basebuilder (focuses on districts, UIs, and wonders. Bonus to building producing and non-aggressive)

Warmonger (focuses on carpets of doom of UUs. Bonus to unit production and grabs all army techs to upgrade carpets)

Commander (focuses on armies/corps/fleets/armadas of UUs and support units.)

Diplomat (focuses on agenda and civis. More bonus civics. Double agenda approval or disapproval)
 
Back
Top Bottom