I figured out whats wrong with the AI, have a possible temp fix

[chillbaka1]

So ive played about 170 hours on civ 6 so far, but have yet to finish a game. My first couple of games were the typical "well ive won, and theres no point in clicking through 40 more turns." King was way too easy, so I bump up to emperor, lag behind for a while, but then after 100 turns im obviously going to win. Then the same thing on immortal. I personally am not a super competitive player, I play to create a story of my civilization. For example, in civ 5 I loved getting enough GPT to be able to fund wars against the next biggest power, or using city states for a sort of "cold war."

Anyways, sorry for the rant but its killed me to see that its all gone in this game. In my last 3 games I was trying to do something similar, giving whoever is at war with the biggest power (usually kongo) like 8-10k gold along with whatever strat resources they needed, but nothing happened. Try funding the largest power to have him take over the world (like genjis khan and shaka zulu did in V) and nothing happened. Nobody EVER takes cities which is extremely frustrating, so I decided to run a test.

I set up a game on 6 leaf clover, deity, abundant resources, balanced start, start in modern era, domination victory only. Starting in the modern era eliminates religion so the AI cant even devote resources to it (i assume that the AI doesn't know when certain victory conditions are disabled.) I get 4 decent cities and just produce units to line my border with an impenetrable defense. Produce some spies and scouts, get open borders with everyone and place the spies and scouts so I can see everything happening in all possible battleground zones. Here's what my conclusions are after about 200 turns.

-The AI will not attack a unit unless it has a chance to kill it that turn. this is why they rarely kill your units, even on deity. Barbarians do not have this mechanic which is why they seem to have a much better AI and seem to be more difficult, even when you have +5 combat strength.
-The algorithms used for bombard units are absolutely useless. They do not know that they can't fire after moving so they try to move, realize they cant fire, then pillage. Repeat until dead. In the rare occasion that they don't move, i've only seen them fire on units.
-Bombard units and normal units do not communicate (in coding terms) with the other units that are trying to take the city. This is why cities are only taken in the classical/medieval era due to battering rams and siege towers. Say we have 12 mamluks and a siege tower (which i've seen take a city of mine.) They "know" that with the tower/ram they will do X amount of damage per turn and receive X damage, and therefore can take the city in X amount of attacks. If the ratio meets some value requirement, then they will take the city. The problem is that along with bombard units not knowing they cant move and attack, they do not know that their job is to destroy the walls. Without something to destroy the walls, normal units calculate their damage to themselves and to the city and the attacks/units required to take the city. The ratio does not meet the requirement so they never try.
-Units value self-preservation too much. This is why in the late game, we see 20+ helicopters/tanks looking like they are preparing for an invasion, but only seem to line up against each other.
-The AI values short term gains over long term ones. This is why we see them pillaging everything to receive the loot rather than taking the city.
-The AI is completely incompetent with air units, and i question whether or not air units actually have any algorithm at all.
-Ranged sea units seem to be completely lacking an appropriate algorithm. They act more like bombard units rather than ranged units, and seem to do absolutely nothing.
-The AI seems to declare war with the purpose of killing units and pillaging rather than wanting to take a city.
-The AI values resources/greatworks/gold over cities, and does not seem to compare a cities science/gold/culture output to the long term value of trades.

So whats my solution? Firaxis, please please PLEASE use the same mechanics and algorithms used in civ V until you are able to fix these problems. If need be, use the same movement cost system, bombard units needing to set up, and other things that allowed games in civ V to actually be fun. Let me know what you guys think.

 

[Gorbles]

I've had AI units go for my units with no solid chance of killing any one of them. I think the problem with perceptions of AI is that it's wildly inconsistent. And no, this isn't me defending it. But it makes it hard to identify trends and even harder to work out some kind of solution to problematic trends.

Also things like "I wonder if they have any algorithm at all" is grossly misunderstanding how software works. A unit could be referenced by ten individual methods in code and still do nothing ingame. There's very little correlation there.

 

[teks]

I'm pretty sure the AI behaves exactly like cockroaches. Their units look all scary, then you turn the light on and they scatter into the shadows.

 

[chillbaka1]

I've had AI units go for my units with no solid chance of killing any one of them. I think the problem with perceptions of AI is that it's wildly inconsistent. And no, this isn't me defending it. But it makes it hard to identify trends and even harder to work out some kind of solution to problematic trends.

Also things like "I wonder if they have any algorithm at all" is grossly misunderstanding how software works. A unit could be referenced by ten individual methods in code and still do nothing ingame. There's very little correlation there.

I am curious to hear what algorithmic functions you think AI air units might have. If it was an unfair, misunderstanding claim, its because Ive played games that i let get all the way until 2200 AD waiting for a single city to be taken without seeing a single air attack, while two of the civs in the game had the secret agenda where its says "likes to have a strong air force and tries to always have the strongest air force, dislikes civilizations with a weak air force."

 

[Lucius_]

The AI definitely needs fixed or changed or anything honestly would be better than what it is. I almost feel like someone tried to make the AI too smart and failed miserably at it. For example, unit self preservation. This is good! This was a problem in CiV, the AI would just hopelessly slam their units into a brick wall and be slaughtered. Now, they wouldn't do that anymore, but they can be yielding an army of Ginsu knives and wouldn't slam it into a sheet of paper. I've seen the AI DoW me, and roll in with catapults, warriors, battering rams, and horses but never actually attack my city, garrisoned with an archer with no walls whatsoever. A city they could take in probably 3 turns, but never actually attack it. I don't know anything about coding but I feel like the AI is so bad at war now that it must be simple to improve it. Even if the AI never learns to use naval, ranged, air or bombard units, if they would just learn to use melee units, it would be a vast improvement.

 

[AriochIV]

The idea that they would disable all the new features just because you think the AI is inadequate is one of the most ridiculous things I've ever heard. If I want to play Civ V, I'll play Civ V.

 

[teks]

Immortal standard match, I was facing a doom army of mech Infantry and AT Infantry and my defense force was antiquated knights, pikemen, field cannons, and artillery. On top of it I was outnumbered 2 to 1

The AI was one turn away from my capital. They declared war. They failed to even pillage a tile. I negotiated a city of theirs in the peace agreement along with the big cash money wad.

Yeah, AI needs some work.

Their mechanical Infantry drove past multiple defenses and encampments, getting whittled down along the way, one by one. The AT guys got stuck in the hills. One would attack fortified artillery, I focused him down, and the others scattered, just to repeat the same failed strategy 3 more times.

And literally all they had to do was move right into my capital and attack it, or at the very least pillage my districts.

 

[oedali]

As the Civ community we should post complaints like this about the AI DAILY so the devs take notice and get the sense of urgency and patch some aspects fast. I am not expecting a master tactician AI to the level of Civ 5 VP mod overnight, but with some basic fixes (like attack timidness) it could at least make the game playable in the short term. They definitely have the resources to make this happen if they want because they still have a team working on the DLCs and whatnot

 

[georgjorge]

Thanks to the OP for shedding more light on what exactly is wrong with the AI. I myself didn't know why the AI was broken in such detail, I just knew it was broken.

Do you also have any information on how the AI determines how many units it builds? I usually see a lot of ancient/classical units from the AI but in my first Deity game I arrived at Brazil's 7-city-empire in 1200AD and faced exactly three units in the war. I killed those three units, and Brazil's power rating (on the victory ranking list) went down to exactly zero. I still wonder how you cannot build more than three units with seven cities in 1200AD. The exact same thing happened with the next two AIs on that map, which also had numerous cities and were very advanced.

 

[Gorbles]

I am curious to hear what algorithmic functions you think AI air units might have. If it was an unfair, misunderstanding claim, its because Ive played games that i let get all the way until 2200 AD waiting for a single city to be taken without seeing a single air attack, while two of the civs in the game had the secret agenda where its says "likes to have a strong air force and tries to always have the strongest air force, dislikes civilizations with a weak air force."

I don't know what function calls pertain to the AI in general, nor how this drills down to unit code references. Because I don't know the game code. But you didn't respond to the important part of my post. The point about you misusing the term "algorithm" was just a secondary point.

 

[alberts8]

Using CivV game mechanics or Ai in CivVI isn't a real option. The only valid fix is playing CivV with Vp until the devs fixed and finished their game.

Next time wait until the game is in a playable state before buying and playing it. These days you can think of the release of a game as the start of the public beta. I feel really stupid for buying CivVI too early.

 

[Krajzen]

Firaxis is the worst game company regarding AI I have ever interacted with. Their incompetence is total and ruthless.

 

[teks]

Firaxis is the worst game company regarding AI I have ever interacted with. Their incompetence is total and ruthless.

Et tu Rome 2. Sega was never far behind.

 

[ProMeTheus112]

Didn't buy it

 

[Gorbles]

Firaxis is the worst game company regarding AI I have ever interacted with. Their incompetence is total and ruthless.

You like making these statements, but you rarely respond to countarguments or criticism raised in return. Nevermind the complete semantic mess that is "ruthless incompetence", AI in general is a problem area for a lot of games. It's one of the largest complaints around the Dawn of War franchise, for one.

AI is a difficult thing to get right. I'm not saying we shouldn't ask for better, but you have to ask what your comments on the subject actually contribute, here.

 

[AmtrakQuebec]

While I agree that the AI needs considerable work, I'm still having fun. When I want real competition from my enemies, I just play some Multiplayer games with good players... even bad players are better than the AI.

 

[chillbaka1]

You like making these statements, but you rarely respond to countarguments or criticism raised in return. Nevermind the complete semantic mess that is "ruthless incompetence", AI in general is a problem area for a lot of games. It's one of the largest complaints around the Dawn of War franchise, for one.

AI is a difficult thing to get right. I'm not saying we shouldn't ask for better, but you have to ask what your comments on the subject actually contribute, here.

Honestly, the only thing i want is for the AI to be able to take cities and conquer weak civs. That evolution of the "survival of the fittest" in V was probably my favorite part.

 

[Severus]

but you have to ask what your comments on the subject actually contribute, here.

I have no desire to cause anger or start anything but you might like to ask yourself that. In another thread you were arguing something when you haven't even played the versions of Civ pertinent to that discussion.

At least the OP was trying to help by posting his thoughts. It is, after all, a forum.

Do you try and put people down on every thread or just the two I've read I wonder?

 

[Kougar7]

Yeah, the V AI at least could capture targets and units.

Some other bugs with the VI AI logic:

In VI the AI will intentionally ignore barbarian settlers and barbarian workers. I've watched AI path their military units AROUND them, while the barbarian civilians try to attack the city center or random nearby units. Watched it for 20 turns because I couldn't believe it.

In VI the AI will buy settlers, then the next turn buy workers in sieged cities. I've seen it happen over and over, it's a programming issue. The city can be completely defenseless or even almost dead. The AI buys civilian units when a military unit is fortified in the city because the AI refuses to MOVE that military unit in order to buy military reinforcements. Even when it is deliberately given the ability to do so... instead it buys civilian units and moves them out between attacking forces. Even when the city is capture-ready in one turn it will continue to buy civilians.

Similar to V the AI in VI will spam cities in places that have zero resources. Snow, tundra, or grassland smack in the middle of another Civ's empire. At least the AI in V would try to put them on luxury hexes, in VI they don't even do that anymore. Then because of this the AI has to pay absurd sums of gold for lux amenities to anyone willing to sell them by late game.

 

[Gorbles]

Honestly, the only thing i want is for the AI to be able to take cities and conquer weak civs. That evolution of the "survival of the fittest" in V was probably my favorite part.

I get that completely, but I've seen the AI do that. Not consistently, though. The same goes for Barbarians - general reports is that Barbs generally know what they're doing better than the faction AI (and I actually think I'd agree in general), but I've had times where Barbs have just milled around one of my cities not knowing what to do.

Identifying why the AI falls over occasionally would go a long way to resolving the issue I think. This is purely based on my own experience, though.