AI is incompetent or programmed to not win?

gcampono

Chieftain
Joined
Nov 18, 2018
Messages
42
Hi,

I know the problem of a really bad AI was brought up many times in this forum, but yesterday I finished a game where one AI (Scotland) was doing really really well: it had much more science, culture, religion, military might than me and all other AIs (like 600 to 200).

I really tought I would give up when I saw it was near cultural victory (7 turns remaining), "near" science victory (already had moon landing) and was invading me with Giant Death Robot and Rocket Artillery armies (I declared war) which were one shotting all my units as well as invading all AIs and me with Rock Bands.

I continued on and observed that Rock Bands were doing nothing at all for at leas 30 turns (they only made a few concerts when it was far from the victory but not too much... to avoid an accidental win?), GDRs killed most of my armies but instead of taking my cities they went to take an irrelevant city state and they never progressed in science victory despite me taking another 50 turns or even more to get to the win.

So the question: do you think the AI really programmed to avoid winning (even at the Deity level) or is it only incompetent?

PS: I know that the AI can aggress you really well at the beginning (I lost or probablly should I say give up on many games at the Deity level), but I got the feeling that developers forbid the AI to win a game after some turns (and since I heard sometimes the AI wins... before some really high limit) to not make the player feel he wasted too much time.
 
I have wondered this. I've played since release and the AI has gone through a lot of changes. To me it feels like when the game first released AI was far to easy to beat but improved in the time after with patching and balancing. However since the major expansions, especially GS I think the AI has gotten worse due to the fact their are now a million paths to go down, many more features to process and the game has become much less balanced in general.
You can make yourself a tougher game (Deity with 12 civs is pretty hard for most players) but I feel that the AI needs more direction and it should go for a victory after a certain point and also be much better at hindering yours. I've never actually won on Deity but I have on Immortal and the beginning of those games are pretty tough just due to the bonuses but the AI does not make the most of those bonuses later in the game.
 
I'm actually curious now.. Does AI do better in Vanilla Civ 6?

I've seen people complain about the AI not pushing for the last steps towards Science Victory but I have managed to lose a couple of times because of that information. But it's not like the AI is bee lining the last steps - or priotise the parts that provides faster travel speed so they can win faster.
 
Yeah in my recent immortal game (my first one - which I lost!!! yay!!!) two civs launched their final spaceship step which then caused me to (happily) rage quit. So it just happened in one of my games.
 
They also for sure manage religious victory and go for diplo victories !
Never saw one win domination or culture though, but I'd say there is only Domination that they seem to somehow never fully go for.
 
Last edited:
It wouldn’t surprise me at all if Civ was designed so that the player can simcity and still have a chance to win, at any difficulty level. I personally don’t mind this, as I am the simcity sort. In fact, i think one of the new game modes should be a mode where the AI tries to win, as strange as it sounds, so there are options for both types of players.
 
In one of my last deity games, Australia was progressing really well along on their science victory, and though I won before they started the last project, they were queuing them up on the spaceport one after another. I don't think the AI is too intent on religious victories as much as they just want to spread/maintain their current religion; I've seen multiple times a religious civ just ignore converting another civ because the latter civ had only a handful of apostles outside. It was like they deemed it to not be worth the effort and just went back to running around their own lands. Culture victory also seems hard for them, for some reason. Domination's out of the question.

Diplomacy is the scariest victory for the AI to pursue, but it does seem like they achieve it rather accidentally. I don't think they're necessarily going for it, as much as they're just attracted to competing in all the WC competitions and spending huge lumps of favor on their interests. The diplomatic victory is sort of like a time-cap on the player, the AI will win it if it's completely ignored by a certain turn.
 
I'm pretty confident the AI's late game simply does not exist at all, as it doesn't really know how to do anything in those eras.

Reminds me of back then when I used to script for Starcraft 1 and I didn't want to write the end game for some of them. It'd look like:

--final attack--

train x unit
attack @ x units
goto final attack
 
It is a programming issue IMO. AI pursuing a certain victory condition pretty much ends at “produce lots of faith/science/culture “ - it has no clue exactly how an actual victory condition is achieved


AI is not “programmed not to win” - which would be deliberate and more sophisticated

AI is simply “not programmed to win” eg that part of the code is virtually nonexistent
 
Last edited:
It is a programming issue IMO. AI pursuing a certain victory condition pretty much ends at “produce lots of faith/science/culture “ - it has no clue exactly how an actual victory condition is achieved


AI is not “programmed not to win” - which would be deliberate and more sophisticated

AI is simply “not programmed to win” eg that part of the code is virtually nonexistent

Indeed.. They wrote the AI to be roleplaying at best, a mere NPC for the human player. If it wins it's mostly by accident. However they do seem to focus a bit more on SV. Probably because it's was a really easy part to code. Just research and build space parts.
 
It is a programming issue IMO. AI pursuing a certain victory condition pretty much ends at “produce lots of faith/science/culture “ - it has no clue exactly how an actual victory condition is achieved


AI is not “programmed not to win” - which would be deliberate and more sophisticated

AI is simply “not programmed to win” eg that part of the code is virtually nonexistent

Agree on this. I can't tell you how many times I get a message in the mid- to late-game that an opponent has decided to stop pursuing a domination victory when they haven't even attacked a CS since the game began.
 
Indeed.. They wrote the AI to be roleplaying at best, a mere NPC for the human player. If it wins it's mostly by accident. However they do seem to focus a bit more on SV. Probably because it's was a really easy part to code. Just research and build space parts.

Agree on this. I can't tell you how many times I get a message in the mid- to late-game that an opponent has decided to stop pursuing a domination victory when they haven't even attacked a CS since the game began.

I agree with both of these comments, but it's mildly amusing seeing them next to each other. If they wrote the AI to be roleplaying, then why have gossip about their victory condition? I find those messages jarring because (1) the idea of "victory" pulls me out of the game universe, and (2) it makes me realize how incompetent the AI is at these victory conditions.

Semi-related: does anyone know what it means when you see gossip that the AI is "considering war" on another AI? Those seem to pop up a bunch, with no subsequent war declaration.
 
Semi-related: does anyone know what it means when you see gossip that the AI is "considering war" on another AI? Those seem to pop up a bunch, with no subsequent war declaration.

On the contrary, I do see AI actually declaring war on each other after the "considering war" gossip, esp. on City-States.
 
In vanilla civ5 the AI was coded to play to win- a lot of people hated that.
AI playing to win in that case took the form of “everyone dogpile the human if he gets close to winning.” Unrealistic, the players said.
 
The AI should not be aware there is something called a win condition at all, that is really immersion breaking.

They should have however an agenda that's more of a purpose - like cultural agenda means they build lots of Theatre Squares, religious agenda means they will try hard to spread their religion etc.
 
The AI should not be aware there is something called a win condition at all, that is really immersion breaking.

They should have however an agenda that's more of a purpose - like cultural agenda means they build lots of Theatre Squares, religious agenda means they will try hard to spread their religion etc.

Then why even have victory conditions, in the first place? Why does the game series simply assume that human history will end circa 2050 CE, with one civ standing as the lone victor? Doesn't that mindset kind of reek of cold war mentality?
 
In vanilla civ5 the AI was coded to play to win- a lot of people hated that.

Really? That would explain a lot about the Civ 6 AI. Because yeah, the vanilla Civ 5 AI was psychopathic. ["Seriously, Gandhi? Nuking me TWICE, on Emperor?"] The DLCs tamed it, in a good way, but in 6 it's downright polite. "Oh, are you trying to win? I'll get out of your way then. Carry on!"
 
Last edited:
The AI should not be aware there is something called a win condition at all, that is really immersion breaking.

They should have however an agenda that's more of a purpose - like cultural agenda means they build lots of Theatre Squares, religious agenda means they will try hard to spread their religion etc.
If you think win conditions are immersion breaking you can just turn them off.
 
Back
Top Bottom