So they still haven't fixed the AI with regard to war.

[Infixo]

@Aristos That was a polite way to say „I am not going to continue this discussion”.

 

[Council 13]

Last night I started a joint war against the Cree with Georgia. Georgia and I had 2 of the Cree's cities down to 0 HP and defense as we surrounded the island with battleships. Georgia had 2 armies sitting right next to the cities on land. For 10 turns they just shuffled around and never took the cities. I ended up losing interest and moved my units elsewhere and by the end of the war another 2 dozen turns later, the cities were still there. MY only guess is that something in the loyalty or grievance mechanic is preventing them from taking cities. Thats the Civ6 AI in a nutshell. It happens all the time and you can't help but feel that something is really broken.

Up until GS, one of my favorite aspects of Civ 6 from a war perspective were barbarians... as they actually seems relatively smart and challenging, at least at first. Now barbarians are broken too...

 

[rbricker99]

Question to people in this thread... How would you feel if the AI completely smashed you in war without any cheating?

The Verge just did an article about how game designers make AI "fun" for gamers not the hardest.

It would seem that modern AI's could be trained to do Civ combat very easily. Likely to the point where it would crush or at least frustrate human players.

I remember I stopped playing chess on my Apple IIe when it would beat me every time. Thoughts?

 

[shaglio]

Question to people in this thread... How would you feel if the AI completely smashed you in war without any cheating?

The Verge just did an article about how game designers make AI "fun" for gamers not the hardest.

It would seem that modern AI's could be trained to do Civ combat very easily. Likely to the point where it would crush or at least frustrate human players.

I remember I stopped playing chess on my Apple IIe when it would beat me every time. Thoughts?

I wouldn't enjoy being "crushed" at all. A tough battle that I ultimately lost is acceptable, but I'd at least prefer a fighting chance.

For a bit of perspective, today at lunch I played a couple of Arena Battles in Star Wars: Galaxy of Heroes. The first one was a back and forth struggle that I didn't win. I was disappointed, but I was okay with it. In the second one, I got to make one measly attack and never got to go again. It was a complete slaughter and it took every ounce of willpower not to whip my phone across the room.

 

[rbricker99]

I think that's the issue. It would be an interesting Mod, but I think the Game Devs would regret making an AI that made combat something human players dread.

Maybe a ramping AI that gets smarter the longer the war lasts.

 

[Victoria]

I think this speaks to how differently people approach civ, and how different their expectations are.

Winning is not what it is about for me but I do get that victory is a target that makes you think in V.
To me any game is like a game of golf, it is really about bettering yourself.
You won a map in V because you restarted until you got 3 cows... I quite like playing on deity on a grass start and being thrashed but going down fighting.
I can better myself on any game, it is whether I enjoy playing it. I find VI has more choices than V personally.

 

[glider1]

Question to people in this thread... How would you feel if the AI completely smashed you in war without any cheating?

The Verge just did an article about how game designers make AI "fun" for gamers not the hardest.

It would seem that modern AI's could be trained to do Civ combat very easily. Likely to the point where it would crush or at least frustrate human players.

I remember I stopped playing chess on my Apple IIe when it would beat me every time. Thoughts?

Good question but the answer is just to use difficulty levels as now. The reason you stopped playing chess on your Apple wasn't because of you but because the app was badly designed. Todays AI play handicap chess really well even better because they make human like mistakes at lower difficulties you can learn from which was not the case in the past. It didn't even need new tech to achieve that. Stockfish engine on level 0-3 is very good for novices to learn from using traditional technology.

Firaxis mistake is to build a new civ from scratch rewriting the game at huge cost to themselves and then expect experienced players to wait for years while the game is brought back up to the standard it was before and then pay full price for that? I would only pay full price for newest civ if the company did much more incremental fine grained evolutionary improvements to the AI engine on a regular schedule over many years in a steam beta program. Then I would pay full price and not wait for it to go on 75% specials.

An even better idea would be that Firaxis could price development cost into the product. You could buy a developers version of Civ which costs more but you get special access to beta AI development as it progresses.

 

[Victoria]

I remember I stopped playing chess on my Apple IIe when it would beat me every time. Thoughts?

100% agree, I mentioned the exact same thing a while back. Everyone used to love chess games until they started beating people even on easy levels.
The point is there is a large audience out there and only a small number of hard core strategist.
They can fit lots of demographic requirements into the game but hardcore would require its own dll coz it certainly is not compatible with many others.

 

[megabearsfan]

I wonder how much of these military issues are actually economic issues in disguise.
Especially around amenities, producing more units when they lose them, brutally bad card slotting, that sort of thing. If the AI has growth bonuses it will constantly be running into amenity problems. Other than adjacency, we would expect that this huge emphasis on just having infrastructure would make yield management pretty easy since the AI only needs to place lots of campuses etc.

Their district placement is also bad, though; I wonder if there is a way to improve it. I often raze captured cities because of the poor district placement. Part of it is planning, but perhaps a simple model of "flat land=farm, hill=mine, resource=improve resource" With a tweak for environmentalist civs to use lumber mills. Then you'd know how the terrain will be improved in advance and can,for example, plan IZs.


Stack building may be appealing for Civ4 vets, but dropped into civ6 as is would probably result in stacks dancing in your city strike range instead! The addition of the City as a static unit and ranged units in general, as well as the corps system has, IMO, made stacks "obsolete." Stacks' AI upside is force concentration- simply having much earlier corps/armies in the game would also give them that advantage (because they also have very high unit production bonuses.)

I too suspect that some of the AI's problems in the mid-to-late-game on high difficulties could be the result of economic problems. The growth bonuses leading to low amenity is one possibility. I think the AI gets amenity buffs to compensate. But also, if they get buffs to tech and civics, then they may also be racing ahead in the tech tree early. This may seem nice, but it can cripple military development. You end up unlocking high-cost units before you have the industrial infrastructure to train them in a reasonable time, or the economy to buy them. This is especially problematic given that the A.I. is notoriously bad at district-placement and not getting the maximum yields from them. If their Industrial Zones aren't getting adjacency bonuses, then that is slowing down their production mid-game, and if their Commercial Hubs aren't getting their adjacency bonuses, then the AI isn't going to be able to afford buying or upgrading units. I haven't looked under the hood to confirm my suspicion, so I may be totally wrong about this being part of the problem. I do notice that once you take out AI units, they are extremely slow at replacing them (if they bother to replace them at all), which is part of what leads me to this suspicion.

This is why, in the past, I've advocated for the idea of handicapping the player to slow them down, rather than handicapping the A.I. to speed them up: http://www.megabearsfan.net/post/2017/06/03/Frustrations-with-Civ-difficulty-levels.aspx.
This way, the pace of the game is maintained, and the A.I. isn't shooting itself in the foot with unsustainable growth.

100% agree, I mentioned the exact same thing a while back. Everyone used to love chess games until they started beating people even on easy levels.
The point is there is a large audience out there and only a small number of hard core strategist.
They can fit lots of demographic requirements into the game but hardcore would require its own dll coz it certainly is not compatible with many others.

Isn't there a line in 2001: A Space Odyssey in which HAL confides to Dave that he is programmed to let humans win about 50% of the time, in order to keep their morale up?

 

[Kimurae]

I think the core problem is you the AI they give isn't really designed for a tactical challenge.

I mean.... they could have just programmed the map generation to just give them their cities with optimal district placement, ignore the game rules. They get tech and free upgrades at regular intervals; they spawn units at regular intervals; and then just focus the AI on using those units, ignore everything else. Make them like barbarians; an obstacle to survive until you win one of the victories.

 

[glider1]

100% agree, I mentioned the exact same thing a while back. Everyone used to love chess games until they started beating people even on easy levels.
The point is there is a large audience out there and only a small number of hard core strategist.
They can fit lots of demographic requirements into the game but hardcore would require its own dll coz it certainly is not compatible with many others.

With all respect you aren't understanding AI. The better the AI gets, the more closely it resembles making human like mistakes when you handicap it. See Stockfish chess levels 0-4 and that is still an old tech chess engine.

I think the core problem is you the AI they give isn't really designed for a tactical challenge.

This is no longer true since VP mod for Civ5 proved 1UPT works on old tech and without redesigning the AI. Firaxis would be at the same level as VP mod if they didn't keep rewriting the game and evolved the game on two separate development branches - a feature branch that remains private in house and AI branch that is open to (EDIT: opt-in) public beta.

 

[Victoria]

The better the AI gets,

We all now have to learn AI in my job, been doing it for a couple of years.
I was simply making the point that most people do not like losing.
As I originally said, when quantum is commonplace, things will change, currently you have 2 hopes of Firaxis investing because they are investing for a small audience and it’s not cheap yet.
I also do not see them releasing the dll while there is another expansion likely.
@megabearsfan indeed, I have just been rereading a fave book of mine, one of the greatest composers asks the hub mind if it could write a symphony in his style that no-one would be able to spot and the reply is, I could do it faster but not with the same satisfaction.
It is swings and roundabouts.

 

[Sostratus]

If their Industrial Zones aren't getting adjacency bonuses, then that is slowing down their production mid-game, and if their Commercial Hubs aren't getting their adjacency bonuses, then the AI isn't going to be able to afford buying or upgrading units.

It's easy to check their placement ability: just capture an AI city. My biggest turn off of conquest is that I inherit city layouts that are sabotage levels of bad. It keeps me peaceful!
Human ability to get 2-4 extra yield from adjacency, especially early, is HUGE. I don't think the AI buys tiles just to place districts either, i could be wrong. But for an IZ specifically, I think they really struggle because they can take a long time to improve their tiles even on 6+, which means it won't get anything from mines that don't yet exist. (I think earlier in this thread I suggested to just program the "normal" AI to farm every flat tile and mine every hill, to make IZ placement more deterministic: place near the most hills.) It hurts the IZ, it hurts their coal plants, it's just brutal all around. (Seowon AI is extra bad, late game they end up surrounded by districts. Excusable though)

 

[Kimurae]

This is no longer true since VP mod for Civ5 proved 1UPT works on old tech and without redesigning the AI. Firaxis would be at the same level as VP mod if they didn't keep rewriting the game and evolved the game on two separate development branches - a feature branch that remains private in house and AI branch that is open to (EDIT: opt-in) public beta.

What I mean is, they put a lot of effort into making it behave along agendas, and have feelings towards your actions. I wish I could just turn all that crap off; its annoying to always be playing against an AI that throws temper tantrums. I'd almost rather have a game with barbarians and disasters.

 

[Council 13]

100% agree, I mentioned the exact same thing a while back. Everyone used to love chess games until they started beating people even on easy levels.
The point is there is a large audience out there and only a small number of hard core strategist.
They can fit lots of demographic requirements into the game but hardcore would require its own dll coz it certainly is not compatible with many others.

To be clear, you would also stop playing chess if you won every single time and the computer was completely inept at putting up any kind of reasonable challenge. This isn't about asking for a hard core challenge. Its really a cry to make the AI at least somewhat challenging at war and at defending themselves.

 

[steveg700]

My first game in GS I had Greece as a the most powerful AI civ. They have the Airpower agenda, so I really and truly hoped to see them produce aerodomes and build aircraft. I let that game run waaaay overlong (over 100 turns past Greece researching Flight) praying and hoping.

What a letdown.

 

[Halcyan2]

The AI's war deficiencies (mainly the lack of desire to take cities even when they easily could) seem to be restricted to major civs and I don't think it's an amenities issue.

Right now, in my GS game, in order to delay an accidental Cultural Victory, I've been propping up America (2nd place in culture) all game, and they are a solid 2nd place.

I'm constantly selling America all my duplicate luxuries (so I'm typically trading 8-10 luxuries to him), which makes me think this is NOT an Amenities problem.

I've been trying to "feed" America some cities in our joint war with the Cree (America is right next to the Cree). Despite the fact that I use Jet Fighters and Stealth Bombers to reduce the Cree cities to 0 health, America refuses to take them with his AT's, Field Cannons, and Cuirassier. After many turns of American inaction, city state Yerevan (which I am suzerain of) has no hesitation to take those cities and raze those cities with their Pike & Shot, Crossbows, and Knights. It's not purely a "comparison of unit strength" since Yerevan is willing to attack with weaker units and American is not willing to attack with stronger ones.

If Major AI's actually used the aggressive AI algorithms of city states, they would do significantly better.

 

[Sostratus]

If Major AI's actually used the aggressive AI algorithms of city states, they would do significantly better.

I believe major civ AIs have their military broken into a few "battle groups" and those groups are assigned tasks. Whereas city states just have their one group, and they usually keep them all in one spot.
it's very curious because I have also observed CS do very well in wars - roll up catapults, drop walls, raze the place with melee units. I wonder if we could easily make the major civ "battle groups" have the same behavior as the city state armies (stay clumped around base of operations, attack target together, etc.)
Early on when the Ai only has like 6 military units they have no trouble dropping CS with nothing but warriors.

 

[Infixo]

I wonder if we could easily make the major civ "battle groups" have the same behavior as the city state armies (stay clumped around base of operations, attack target together, etc.)

Minors and Majors use exactly the same Behavior Trees. Only Barbarians and Free Cities have their own trees.

 

[ThunderLizard2]

I've been holding off on Civ VI mainly because of the braindead AI. Looking for honest feedback on it now after GS. My benchmark for decent AI is Civ IV BTS although I find Civ V with Vox Populi acceptable but not exciting/challenging.

A couple subquestions:
* Does the AI know how to conduct a decent attack and take over cities?
* Does the AI know how to manage it's own cities?
* Does it use all of the newer mechanics?

P.S. I did a search for AI and got no results