Has the AI improved?

[Deaga]

Hello everyone!

I've been a Civ player since the very recent times of Civ5 and remember being pumped for Civ6 in the pre-release period. However, after release, despite the game having a lot of great and fun systems, I couldn't enjoy it much due to a very simple reason: the AI seemed incapable of using all those new systems! The end result was that me, being an Emperor player in Civ5, could consistently beat the Deity AIs in Civ6, provided I survived the early game warmongering. This made Civ6 not really fun to play, as I knew that if I survived past turn 50 or so, I had already won; the AI couldn't really use districts or eurekas properly, so once I survived their early rushes, I'd very quickly get ahead of them and click Next Turn until I won.

Has this been improved, all these years later? I really want to get into Civ6, districts are a lot of fun, but without an AI that's at least a bit competitive, that's kind of hard. I guess mod suggestions would also be helpful, if the game itself still has bad AI too. Thanks in advance!

 

[Starwars]

There are parts of the AI that have been improved but not in the way you describe sadly. The whole "if you survive the early rush" thing is very much a thing in my experience. The tension and interesting gameplay lies in the early game and once you have your empire up and running, it's smooth sailing to victory if you know how the game works.

There is a mod called Real Strategy which improves a bunch of things about the AI but there is no mod that really improves it to the point where it's competitive that I know of. A bunch of things about the AI are still hidden in the source code (if I understand things correctly) so modders are quite limited in what they can do to improve it.

 

[Tech Osen]

Some parts were improved. Land combat is a tad better. The AI used aircrafts decently for a while until GS changed the resource system. It seems the AI can't handle the new resource system all too well in general.

 

[Loderingo]

I would say the AI has improved a lot since the initial version of Vanilla but it is much stronger when it is being proactive (it attacks you) rather than reactive (you attack it).

There are still a few things that hold the AI back:

1) Military civs care too much about warmonger points
2) All civs heavily target aid contests even if they are not going for diplomatic victory
3) The AI is bad at repairing sabotaged spaceports
4) I have never seen the AI build the space victory projects (la grange and the other one)
5) The AI tends to focus too heavily on religion, which is a hard victory to reach on standard maps and larger

If you want a tough game, the AI tends to well with Germany, Cree and the Dutch, followed by Korea, Nubia and Australia. It tends to do less well than the player with more niche civs like Hungary and Maori.

 

[Haig]

It has improved a lot since when the game was released, but still gives very mixed results.

Sometimes the units shuffle around a city and cant focus a proper attack. Sometimes they surround a city with good forces and capture it.
Sometimes the ai even builds jet bombers and bombs you!

I just wish the good tactical moments were more frequent

It defends better than it attacks though.

 

[King of Prussia]

It has improved a bit, but it still has issues. especially in the military front. They don't build enough naval vessels.(played an unmodded civ 5 game the other day and the difference was noticeable) They have a harder time building an airforce because they need both an aerodome and the resources. This isn't getting into empire management either that effects things like policy cards, improvements, and district placements.

the more checks/barriers you have, the harder it will be for the AI to overcome them. Players can plan ahead or compensate for said barriers while the AI cant or has way more trouble with it. Personally i wouldnt mind removing the aerodome requirement for planes. Cities can still only have one in them at a time. I am for the removal of requiring resources and resource upkeep for troops, and instead turn those resources into an empire wide boost.(perhaps boost to building said units) Not only would it make it easier for the AI, it would mean that late game isn't as RNG dependent for the required resources. You would be less likely to see early game units running around.

I suppose another option is to have buildings provide the required resources, like in Civ V. if there is another expansion, maybe they will tackle this with a global market system.


on a similar note. I wish the AI was more aggressive at trying to win. Especially the military focused Civs. They are too afraid in the late game to actually push hard. Not just take a city or 2, but push until you get the capital or wipe them out. You could toss in repairing and guarding spaceports.

 

[nauberry]

4) I have never seen the AI build the space victory projects (la grange and the other one)

I lost recently a very good and interesting game, where I was on my way to culture victory, but left korea unchecked for 50 turns. They were not the biggest empire and had even fallen back in science, but in those 50 turns they managed to shoot off the last mission to space. I started invading them to stop them from winning, but they launched both of those projects trun by turn and their ship was flying 7 light years a turn. Lost the game only a couple turns before I would have won the culture victory.

 

[Sostratus]

It’s definitely true that the AI just does phenomenally with the flavor weights of certain civs, even though it’s normally meh.

I know there’s a mod out there to prevent the AI from even engaging in holy sites, which probably makes them play better. This is also a big boost to the Kongo AI.

The AI isn’t that dumb- I have seen them do reasonably clever things. The real issue is they are programmed to play in a horribly inefficient way that kneecaps them at every single level. This is why AI Korea etc can open up 3-4 era leads on everyone - they are weighted to play in a more “meta” way. Success really begets success here- when they have a military edge they can make you feel the pain on higher difficulty.
If you think their district choices are bad just wait until you look up your game logs and see their policy card load outs.

 

[enKage]

The Ai is somehow better at defending now, but still extremely dumb

1) AI is able to build caravels in 1tile lakes
2) In my recent game Eleanor threw at me massive force besiegieng my 2 cities for 20+ turns. The problem was mainly cats and xbows and were never able to move a knight toward surrounded city to do the final hit, just because their Xbows occupied every single tile around the city
3) One day AI were building too many farms, now they cannot grow to unlock district slots. These ends up often with culture/science output in modern era equal to player output in medieval era. There are many cases where civs having 10+cities have maybe 1 or 2 campues in whole empire
4) AI can compete to become a suzerein of Jerusalem to over 20+ envoys (and other religious city states too) leaving you with suzereinity over Babylon, Geneva, Kumasi, Antananarivo forever with 3 envoys only
5) AI can use all diplo favors on resolution to ban a certain ammenity for 30 turns
6) AI Maori constantly suicides every single game by settling high populaed land
and many more

 

[steveg700]

A lot of what most consider "dumb" AI behavior is, I suspect, intentional design. It's what I think of "marking time" behavior. That is to say, the AI follows a routine where it expends its production on certain activities without regard as to whether or not they're going to bear meaningful fruit. This is why the AI chases religion and wonders so feverishly. It's not working to be competitive. When it's not building military units for war, it's just looking for something to do that kills time.

So, for instance, every AI scrambles like mad to build Machu Pichu as if building it was the most crucial activity they could possibly undertake. Only one's going to actually get it, and even then, It's not that decisive an accomplishment. But it marked time. Now let's go chop down all our forests and build farms so that we aren't forced to deal with more production than we care to manage.

 

[steveg700]

4) AI can compete to become a suzerein of Jerusalem to over 20+ envoys (and other religious city states too) leaving you with suzereinity over Babylon, Geneva, Kumasi, Antananarivo forever with 3 envoys only

And don't forget that after sending a ton of envoys to a CS, they will then decide to declare war and conquer it because they aren't the suzerain. And the suzerain is in an alliance with them, so the ally has to wait for it to expire in order to declare a Liberation war.

 

[ThunderLizard2]

A lot of what most consider "dumb" AI behavior is, I suspect, intentional design. It's what I think of "marking time" behavior. That is to say, the AI follows a routine where it expends its production on certain activities without regard as to whether or not they're going to bear meaningful fruit. This is why the AI chases religion and wonders so feverishly. It's not working to be competitive. When it's not building military units for war, it's just looking for something to do that kills time.

So, for instance, every AI scrambles like mad to build Machu Pichu as if building it was the most crucial activity they could possibly undertake. Only one's going to actually get it, and even then, It's not that decisive an accomplishment. But it marked time. Now let's go chop down all our forests and build farms so that we aren't forced to deal with more production than we care to manage.

Is that what you observe or what you think devs intended? Hopefully the AI is dumb because of poor coding not intentional design choices. They just need to release dll code now so mods can step in and fix like they did for Civ V.

 

[steveg700]

Is that what you observe or what you think devs intended?

Well, both. It's not as if nobody's pointing out to them that the AI doesn't use aircraft, or that using bombers is too clever for the AI's to work out. Same goes for weird policy choices, like not side-stepping Rationalism or other victory-focused policies in favor of more amenities and housing. Not hard to code a no-brainer policy choice. An AI civ settles its third city and starts building an Entertainment Complex of all things. There's poor coding, and then there's its cousin, easy coding.

 

[Lily_Lancer]

Ai is still dumb.
However, Civ 6 is a good game for MP, much better than Civ 5.

 

[kaspergm]

A lot of AI problems boils down to AI playing the game as developers intended - which would not be quite so big of a problem, if they developers had actually bothered balancing or closing the loopholes that human player will obviously use to his advantage (like campus spam, cf. also the other thread about mountains being OP).