• Civilization 7 has been announced. For more info please check the forum here .

My Yearly "Is the AI Any Good Yet" Thread

kaspergm said:
Great write-up indeed.

Going slightly off topic (and very off game), this reminds me of what is easily one of the funniest things I ever read online. This is not about Civ, but anyone who ever played Heroes Of Might And Magic should go to section XI of this page and read about an AI experiment someone did back in the days of Heroes IV. Caution: You may just die from laughter. (The rest of the page is funny as well, but section XI tops it all.)
Click to expand...

Man that was hilarious ;-) Thx
 
no sense the NF pass the AI gotten way worst or the mods Im using to make the AI better have not been patched yet.
 
If all ya'll would allow me to don my tinfoil hat; I think there's a good reason they haven't released dll access. It would reveal that the civs are deliberately programmed to stop growing and contesting other civs. The illusion of challenge would be completely gone, and the perceived quality of the game would plummet. It'd be revealed as just being a particularly advanced clicker game. It's a shame, there's enough talent and dedication in this modding community that they could probably fix it to the point where the AI could at least win in non-accidental ways and put some pressure on the player.
 
I want a tin foil hat as well because I agree with ya. I simply want to see the ai handle base game things like taking walled cities. If they could effectively do domination then when they make demands at you it may cause you to say yea ill give you what you want because your stronger than me. Right now I just laugh and decline.
 
Zerazar said:
If all ya'll would allow me to don my tinfoil hat; I think there's a good reason they haven't released dll access. It would reveal that the civs are deliberately programmed to stop growing and contesting other civs. The illusion of challenge would be completely gone, and the perceived quality of the game would plummet. It'd be revealed as just being a particularly advanced clicker game. It's a shame, there's enough talent and dedication in this modding community that they could probably fix it to the point where the AI could at least win in non-accidental ways and put some pressure on the player.
Click to expand...

I'm not going to argue against this because it would make sense in form how expensive it would be to build an AI capable of winning a domination victory. I do however think most people underestimate the effort needed to make an AI capable of competing with human players. Civ 6 is like a really advanced chess game and even chess AI took many years to get on level with a human player.
AI in rts games is completely different as you can speed up things to give the AI an advantage over the human player plus the decesion tree are often over simplified for rts games. In reality civ is probably just becoming too advance to be able to price a really decent AI.
 
Mahi said:
I'm not going to argue against this because it would make sense in form how expensive it would be to build an AI capable of winning a domination victory. I do however think most people underestimate the effort needed to make an AI capable of competing with human players. Civ 6 is like a really advanced chess game and even chess AI took many years to get on level with a human player.
AI in rts games is completely different as you can speed up things to give the AI an advantage over the human player plus the decesion tree are often over simplified for rts games. In reality civ is probably just becoming too advance to be able to price a really decent AI.
Click to expand...
I don't think anyone here imagines for one second that creating a strong AI is an easy task. If it were, then every game would have one. But do you honestly believe that Civ has evolved beyond the capacity of Firaxis to give it even a decent AI? We have evidence to the contrary -- Civ V ended up with quite a good AI, largely because because more tools were given modders to make one in what ended up as the "Common Patch." It would be hard to argue that Civ VI is so much more complicated a game that it has crossed some threshold of AI complexity. In fact, I don't think that it is ANY more complex than Civ V -- it just has a few more gimmicks like districts, in which you can build pretty much the same buildings that you could build in V.

What's most maddening -- to me, anyway -- is that Firaxis isn't even trying to make it better. Look at the most recent patch notes. They barely mention improvements to the AI, and those few mentions are, to put it charitably, minor ones. Ok, fine. If we the playing community don't demand a more challenging game, then we surely don't deserve one.

One final observation. I for one do nor see a categorical difference between AI's in rts versus turn-based games. Paradox's games all are rts games, as is well known, and the AI in those games range from "meh" to "bad". Meanwhile, the three games I suggested above in this thread that have superior AI all are turn-based games.
 
Last edited:
If Humankind is good then Firaxis will likely take improving the AI seriously. There is no better motivator than competition.
 
I remember playing this game at launch and again soon after Gathering Storm, don't take me wrong, both times was a blast, but the very bad AI always kept me from buying it for myself to keep playing it. I even tried some mods, better but still far from good, I supposed that with time the devs would make it better. Did they? I mean, the game is great, but the AI is very bad at it. Increasing the difficulty just give AI more cheats making the early game frustrating until I catch up and they are garbage again. How is it now? They are still updating the game with new content and DLCs, so are they also working on the AI?

Moderator Action: Merged with existing AI discussion thread -- Browd
 
Last edited by a moderator:
SuperFly47 said:
If Humankind is good then Firaxis will likely take improving the AI seriously. There is no better motivator than competition.
Click to expand...

Money is the motivator, not competency. And humankind will at this stage not take any money away from firaxis. If anything there is the risk of just stopping developement and support if civ sales drop.

We have to be realistic about how this works.
 
yegafiru said:
Did they? I mean, the game is great, but the AI is very bad at it. Increasing the difficulty just give AI more cheats making the early game frustrating until I catch up and they are garbage again. How is it now? They are still updating the game with new content and DLCs, so are they also working on the AI?
Click to expand...

They have improved the AI in the past, including the first NFP patch. Despite new features have been added since realease, the AI has gotten better. Still numerous issues are left, though and I wouldn't call it exactly a good AI yet.

We don't know what the rest of the NFP will bring, but I consider further AI improvements as likely. The question is just how big those improvements will be. Which likely depends mainly on how much of the development ressources are spared for it. And AI improvements take their toll here - it is easy to open a thread and complain about the bad AI, but it is along way from there for Firaxis to improve it. Already the 2nd (finding out, what makes the AI bad) and 3rd (providing a useful save) steps are harder and rarer and if Firaxis has to do this steps themselves...well, then there is less time for coding the actual fixes.

That's the reason why I prefer opening threads in the bug section with examples of stupid behaviour supported by a save over just discussing the AI here - I feel that this is the best we can do.
I have no illusion - not everything reported will get fixed, some stuff is probably just beyond what can be done or afforded. But each such special report is a change for an improvement - and since Civ6 AI suffers - among more fundamental and/or complicated issues - alone dearly from a couple of clear issues (e.g. Spaceport spam, not improving ressources, bad deals, switching governments backwards, ignoring visible threads on the tactical level), even some fixes could result in a much better preformance.
 
Its not just combat though, i see daft stuff regularly.

For example, i notice that if the AI drops down a farm... then it turns out there is a resource such as niter there- they seem to just leave the farm (and not because it is part of a triangle etc)

In the late game i see the AI moving warrior corps around.

If the AI finds itself on an island, it doesnt adapt, it still builds its programmed obsession (such as endless horses if scythia)

The thing is though, firaxis want to make a successful game, commercially- and looking at the stats on steam i doubt most players will care? - i would argue people on here are not typical players.
 
Kriegsspieler said:
I don't think anyone here imagines for one second that creating a strong AI is an easy task. If it were, then every game would have one. But do you honestly believe that Civ has evolved beyond the capacity of Firaxis to give it even a decent AI? We have evidence to the contrary -- Civ V ended up with quite a good AI, largely because because more tools were given modders to make one in what ended up as the "Common Patch." It would be hard to argue that Civ VI is so much more complicated a game that it has crossed some threshold of AI complexity. In fact, I don't think that it is ANY more complex than Civ V -- it just has a few more gimmicks like districts, in which you can build pretty much the same buildings that you could build in V.

What's most maddening -- to me, anyway -- is that Firaxis isn't even trying to make it better. Look at the most recent patch notes. They barely mention improvements to the AI, and those few mentions are, to put it charitably, minor ones. Ok, fine. If we the playing community don't demand a more challenging game, then we surely don't deserve one.

One final observation. I for one do nor see a categorical difference between AI's in rts versus turn-based games. Paradox's games all are rts games, as is well known, and the AI in those games range from "meh" to "bad". Meanwhile, the three games I suggested above in this thread that have superior AI all are turn-based games.
Click to expand...
I do actually belive civ 6 and 5 is more complex to develop a good AI to. Modders improved 5 AI without a doubt but everyone can still beat the best AI in 5 if we'r only talking about units vs units.

I have lot of things against the AI so I agree with most of the things you said. The most annoying part for me is, when I attack an AI it doesn't necessary start producing military units and once I've taken out the base army of the AI its just a walk in the park from there. It should be an easy coding fix to set the AI to focus on military production the instance it gets attacked. At least that would slow down the steam rolling a little bit.
 


Here is the answer to OP question:

The city is under siege. There are plenty of targets. What does AI city guard chose to attack? Useless encampment far away!
 
Tigranes said:


Here is the answer to OP question:

The city is under siege. There are plenty of targets. What does AI city guard chose to attack? Useless encampment far away!
Click to expand...
What city guard? There's no walls..
 
The weird thing to me is that the AI clearly has the ability to take cities but only seems to do a good job at the very beginning of the game. I have lost a number of Deity games in the first 50 turns when the AI shows up with 4 Warriors and all I have is 1 Warrior and 1 Slinger, and they will take down my captial - even moving to siege it correctly.

However, later in the game the AI seems to intentionaly not take cities. I have played many games where they will redline a city but then either move away or attack a unit instead of taking the city. This is especially true for a just-captured city. For example I will often take a coastal city with a Longship and the AI will have something like 2 Swords and an Archer next to it, easily enough to retake it in one or two turns, but will just march those units around aimlesslessy giving me time to buy reinforcements or chop walls.

I assume this is intentional because it would be “too frustrating” for the player if they captured a city and lost it right back, but then why allow the much more frustrating siutation of being killed in the Ancient Era?

The AI remains terrible at winning the game. To be fair, I feel like it has gotten much better at war over the past two years. There seems to be a lot less instances of a ranged unit simply not firing, sometimes (but not always) the AI will avoid making an obviously suicidal attack like crashing a Warrior into a Knight, and occasionally (but again not always) it will attack with a unit stationed inside a city instead of just leaving it there. A big part of this may be that the AI seems programmed to beeline Crossbows now and also build more Encampments, which at least make it harder to invade.

However, I haven’t lost a city to the AI military for years (aside from Deity early game rush). They just don’t seem to be able to invade and end up at best suiciding a handful of units into your borders.

I wonder how much of this early game prowess and late game ineptitude is intentional (e.g. the programmers not wanting to “spoil” the players 200 turns of fun by winning) and how much is the AI being unable to keep up with the more complex calculations (e.g. 4 Warriors surrounding one city versus hundreds of units sprawled over the map).

Anyway, even though the AI sucks it can be fun to see how fast you can win under different conditions, like a standard game, playing peaceful, One City Challenge, etc.

AI is hard to program, but a few very easy things might make the AI more competitive:

1. Have the AI upgrade units. They always seem to have random Spears and stuff walking around in the late game. Since the AI always seems to have a ton of gold it should really be upgrading these. One of the biggest advantages the player has is the ability to build a half dozen Chariots, then upgraded them to Knights and Tanks. The AI seems to mostly hard-build, which means it will take forever to make an advanced unit.

2. Hard code them to always escort Settlers. Sure, I send out Settlers with no Warrior sometimes when speed is of the essence. But the AI just gives away too many Settlers to human Scouts and since it starts with 4 Warriors, there is no excuse for just walking them around in the open.

3. Fix the endless Friendship loop. Last I played, if you get Friendship or Alliance with an AI, you can renew it the turn it expires no matter what. This makes it to easy to secure yourself against one neighbor while taking out another.

4. Make the AI less crazed about DF. They often give ~15g for one DF which is such a bar deal.

5. Chop more. The Ai seems to almost never chop, which is another big disadvantage since choosing is so strong. I almost hate to say this because I love capturing an AI city on T100 and it still having 6 or 7 forests to chop out a Wonder.
 
AsH2 said:
What city guard? There's no walls..
Click to expand...

This game calls the unit inside the city "City Guard" when calculating city combat odds. Do you see the Arabic crossbow in Medina who was shooting at my encampment? That.
 
Last edited:
Tigranes said:
This game calls the unit inside the city "City Guard" when calculating city combat odds. Do you see the Arabic crossbow in Medina who was shooting at my encampment? That.
Click to expand...
Haha I think I only see mirages.. Are those oasis real?
 
You must log in or register to reply here.

Similar threads

V
Basil problems: when the AI has no troops
Replies
0
Views
524
venus_genetrix
V
The_J
Euronews: "AI models chose violence and escalated to nuclear strikes in simulated wargames"
Replies
12
Views
1K
Bibor
Bibor
planetfall
Why is AI so poorly written?
2
Replies
25
Views
3K
UWHabs
U
A
Is it true that the AI is not smart?
Replies
19
Views
2K
robbyextreme
robbyextreme
planetfall
Where is everyone?
2 3 4
Replies
78
Views
6K
reddishrecue
reddishrecue
Top Bottom