Why Civ 6 AI is terrible - Part 1: Wonders

[Siesta Guru]

For me, the worst part of the AI is the tendency to found cities 1 tile away from rivers/lakes. I mean, it's "fine" once you unlock aqueducts, but those extra settlers on emperor+ always end up in the worst locations...

From what I've seen so far, that seems to be because it doesn't boost a city for being adjacent to a river, instead it boosts the score of a position through a count of all nearby tiles adjacent to a river/lake/coast. It absolutely loves settling exactly between rivers, because that's where it gets the highest total count of tiles adjacent to rivers.

 

[kb27787]

Seriously? First off, with Magnus in the game, the human player should be getting every wonder they wish as long as they beeline the tech/civic. It doesn't really take that many harvestings to get a wonder.

Also, building a wonder does two things:
1) Give yourself the bonuses
2) Deny other players the bonuses.

So basically, if they build Chichen Itza in a city with just one rainforest tile... they get a measly 2 culture, BUT they deny that wonder to another civ (usually happens to a player and then we hear the complaints) who has a city with 10+ rainsforest tiles somewhere. Now I guess since they figured that they are toast anyway under the invasion of your army, they might as well frustrate the other players one last time by denying him 20+ culture which would've been available if the player builds it in a suitable location.
It's the same in BNW where you had to beeline for the great firewall after internet tech to deny the deity AI from building it against your cultural victory.

Actually in BNW the AI actually drops building the wonder temporarily (confirmed with spies) once war is declared... so a "bluff" DoW despite your empire being half the world away or a war bribe was actually useful in helping you beat it to a wonder... people thought it was a cheap trick... now in RnF they made AI stick to finish building the wonder undistracted and people still complain.

 

[SharTeel]

Seriously? First off, with Magnus in the game, the human player should be getting every wonder they wish as long as they beeline the tech/civic. It doesn't really take that many harvestings to get a wonder.

Also, building a wonder does two things:
1) Give yourself the bonuses
2) Deny other players the bonuses.

So basically, if they build Chichen Itza in a city with just one rainforest tile... they get a measly 2 culture, BUT they deny that wonder to another civ (usually happens to a player and then we hear the complaints) who has a city with 10+ rainsforest tiles somewhere. Now I guess since they figured that they are toast anyway under the invasion of your army, they might as well frustrate the other players one last time by denying him 20+ culture which would've been available if the player builds it in a suitable location.
It's the same in BNW where you had to beeline for the great firewall after internet tech to deny the deity AI from building it against your cultural victory.

Actually in BNW the AI actually drops building the wonder temporarily (confirmed with spies) once war is declared... so a "bluff" DoW despite your empire being half the world away or a war bribe was actually useful in helping you beat it to a wonder... people thought it was a cheap trick... now in RnF they made AI stick to finish building the wonder undistracted and people still complain.

Sure, they "deny" the human player the wonder, but they still spent 700 hammers for an useless chichen itza (well, more like 380 on deity or something like that), hammers that the human player is spending building the army that will crush them! (or districs, or whatever). The issue is that the AI sucks at evaluating the opportunity cost of such wonders, which is probably the main point of the video.
It has nothing to do with "boo hoo, I'm bad and the AI steals every wonder before I can build it"

 

[ThERat]

Well, though this is about wonders, I can only assume the crappy district planning. Or should I say, no planning, just drop down anything. As mentioned by others, Korea kills its own campus bonus of +4 by building lots of other districts next to it. Playing this game against the AI is just downright awful. No challenge, no fun.

 

[PhilBowles]

Seriously? First off, with Magnus in the game, the human player should be getting every wonder they wish as long as they beeline the tech/civic. It doesn't really take that many harvestings to get a wonder.

Also, building a wonder does two things:
1) Give yourself the bonuses
2) Deny other players the bonuses.

As the human player, how often do you build Wonders to deny them to the AI? It's not that it's never a viable option, just that as a rule it's of marginal value - so that it shouldn't be something that weighs significantly in AI decision-making.

Of course the human player has the ability to simply take those wonders by capturing the cities, which the AI won't realistically do, but it's still not much more than a niche benefit - you aren't usually going to do that just to deny the AI a Wonder, only if you want it for yourself.

 

[aguliondew]

On high difficulty you can plan for a wonder with harvest but getting multiple early wonders like the AI is not possible.

 

[kb27787]

As the human player, how often do you build Wonders to deny them to the AI? It's not that it's never a viable option, just that as a rule it's of marginal value - so that it shouldn't be something that weighs significantly in AI decision-making.

Of course the human player has the ability to simply take those wonders by capturing the cities, which the AI won't realistically do, but it's still not much more than a niche benefit - you aren't usually going to do that just to deny the AI a Wonder, only if you want it for yourself.

In deity BNW I did that quite often since for cultural victory back then things like CN tower or the great firewall on the cultural runaway means you are doomed to wait until t400 if you want to win by culture. By then someone would have already launched the spaceship.

Some wonders give bonuses only to the city it is built so even if you capture a crappy petra or chichen itza it wont change the fact that the AI had denied you 20+ cpt or 20f 20p 40g had you been allowed to get the wonder built in the right city.

 

[PhilBowles]

In deity BNW I did that quite often since for cultural victory back then things like CN tower or the great firewall on the cultural runaway means you are doomed to wait until t400 if you want to win by culture. By then someone would have already launched the spaceship.

Some wonders give bonuses only to the city it is built so even if you capture a crappy petra or chichen itza it wont change the fact that the AI had denied you 20+ cpt or 20f 20p 40g had you been allowed to get the wonder built in the right city.

In Civ VI you can rebuild Wonders in razed cities, so if you capture a bad Petra you can just destroy it and build it in the right place - though it may be too late to be relevant.

As for culture Wonders etc., not only are none of the Civ VI Wonders that important to victory, the AI can be delayed winning until turn 400 in any event - again a consequence of just how poor it is. I spent much of my Deity Dutch game focused on a too-slow war with Korea (I had mostly low-production areas and agonisingly slow production of melee units) and past turn 250 or so I had no real expectation of winning - I'd had alerts that one civ was one culture away from winning and several space projects were approaching the finish line while I had no spaceports.

I started focusing seriously on getting spaceports up and running at about that time (probably closer to turn 300), sabotaged the ongoing space projects with a mix of spies and nukes, and managed to win not far shy of turn 400. In one thing the AI was evidently capable - one of the AI civs had so much culture that the tourism bar was over 800, and the two leaders never got more than halfway there.

 

[Zuizgond]

Hope part 2 will be about policies.

Firaxis has no budget for a team of AI developers?

 

[Ryika]

The point of production bonuses is to make the AI a difficult opponent, not to compensate for its weaknesses. It already needs production buffs to compensates for the extra unit losses it takes as a result of being poor in combat, but that's much easier to compensate for than producing the wrong thing.If the AI never gets to a point where it's producing things that advance its victory progress, it's irrelevant how big its production bonuses are. Giving it an extra 700 because you anticipate it building Chichen Itza is irrelevant if it just ploughs the excess into a one-flat-desert-tile Petra.

That's an artificial distinction you're drawing there. The production bonus is simply to allow the AI to produce more things. Whether that's to allow them to replace units, or whether that's to compensate for bad production priorities when it comes to other stuff, is completely irrelevant. After all, even if it constructs all the wrong things, having more of the "wrong" things still pushes you ahead.

The only situation where that's not true is literally the situational wonders, but when it comes to that, I think you're overstating the impact that the construction of wonders has, because wonders are limited, there simply aren't dozens of wonders available all the time. Even if an AI builds nothing but useless wonders in one city all game long, then all the other cities still benefit from the production bonus. That one city would lack infrastructure, but given that you expect the strong AI players on the map to have more cities than the player anyway, that doesn't really matter that much. And if it does, well, turn up the production bonus further.

Of course that doesn't mean that it's not a problem to have AIs be that bad in how they choose to do things, but the building priorities are in no way as important as, for example, unit combat. Simply because unless the player goes out of their way to find out what the AI is doing, they will never see just how incompetent the AI is, because production bonuses are a good way to hide it during gameplay, it's not "in your face", and the downsides of that system that aren't that severe. If the AI is bad at combat, you are reminded of it whenever you and the AI clash. Thus, those things should be priority work - but of course they're also much harder to do right.

/edit:
When it comes to buildings though, I think they partly made the system the way it is to save processing power. It would not be much of a problem to make the AI "evaluate" how useful even a situational wonder is (you'd just have to look at the effect type and if it's a yield bonus for a specific type of tile for exampe, loop through or index all city tiles and see how many tiles benefit from the effect), it would just make the process of evaluating these things a lot slower than just evaluating the fixed yields.

 

[Bad Wolf]

Firaxis REALLY needs to make one of two choices for their next Civ (or 4X) game. Either:

(1) Scrap 1UPT. It's nice in theory, and it does make combat more strategic... but this pretty much ONLY works to the advantage of the player, as the AI just can't handle the complexity of it, and a human player who knows how to play the game can run circles around the AI tactics-wise;

or,

(2) Invest some real time, effort and resources into developing an AI that can actually compete with an average human player without overwhelming and unfair bonuses.

 

[SammyKhalifa]

I don't know. There are certainly things that can be fixed. But I hear things like "doesn't Firaxis have a budget for AI" and I think.

IBM is a corporate behemoth. The blew a jillion dollars and hundreds and hundreds of hours of work from the brightest minds in the world to teach a computer how to play chess.

Chess doesn't have a different board every time, or different kinds of pieces at play a different times. Deep Blue didn't have a limitation where it had to pretend that it couldn't see large parts of the board.

A Civ AI doesn't need to be as good at civ as Deep Blue was at chess, but I think there might be more to it than people think--especially if you expect to play at home using a PC and not IBM server hardware.

 

[PhilBowles]

Firaxis REALLY needs to make one of two choices for their next Civ (or 4X) game. Either:

(1) Scrap 1UPT. It's nice in theory, and it does make combat more strategic... but this pretty much ONLY works to the advantage of the player, as the AI just can't handle the complexity of it, and a human player who knows how to play the game can run circles around the AI tactics-wise;

1UPT is no excuse for terrible building AI. Combat gets as much flak as it does because it's the system's most visible weakness, but every other element of AI needs to be addressed with as much urgency. See my above example - you can delay the game to turn 400 on Deity and the AI still can't win peacefully. There's no other Civ game of which that is true, at least in the recent past. In past Civ games you can eschew combat to play a race/builder game where the AI represents reasonably capable competition.

 

[ashendashin]

I don't know. There are certainly things that can be fixed. But I hear things like "doesn't Firaxis have a budget for AI" and I think.

IBM is a corporate behemoth. The blew a jillion dollars and hundreds and hundreds of hours of work from the brightest minds in the world to teach a computer how to play chess.

Chess doesn't have a different board every time, or different kinds of pieces at play a different times. Deep Blue didn't have a limitation where it had to pretend that it couldn't see large parts of the board.

A Civ AI doesn't need to be as good at civ as Deep Blue was at chess, but I think there might be more to it than people think--especially if you expect to play at home using a PC and not IBM server hardware.

Every time with this...
Look, these are entirely different games. All those extra variables make it less than what you think since each individual choice means little in grand strategy. The game needs to have the AI capable of working with the game mechanics and have a general line to follow towards victory. It's been done, so it's not as if this is some crazy CPU busting concept.

 

[juanpavo]

A Civ AI doesn't need to be as good at civ as Deep Blue was at chess, but I think there might be more to it than people think--especially if you expect to play at home using a PC and not IBM server hardware.

That is a valid point, but if you haven't already, check out the Civ5 Vox Populi combat AI. If Civ6 combat AI could even be half as good, I would be happy.

 

[SammyKhalifa]

Well yeah, it's not to say there aren't things they can improve.

 

[historix69]

Maybe it would be easier to define reasonable requirements for wonders than try to teach AI to understand complex game mechanics like wonder effects. Define the requirement in a way that you can only build a wonder in a city when you get an advantage from it.

Example : Chichen Itza
Make the requirement to have 3 or more Rainforest Tiles in city area.

 

[PhilBowles]

That's an artificial distinction you're drawing there. The production bonus is simply to allow the AI to produce more things. Whether that's to allow them to replace units, or whether that's to compensate for bad production priorities when it comes to other stuff, is completely irrelevant. After all, even if it constructs all the wrong things, having more of the "wrong" things still pushes you ahead.

It pushes you ahead relative to where you were, but not relative to competitors who are using their resources more efficiently.

Of course that doesn't mean that it's not a problem to have AIs be that bad in how they choose to do things, but the building priorities are in no way as important as, for example, unit combat.

Here you're making an artificial distinction. The worst result of bad combat AI is that the AI loses units, which requires production to compensate by allowing it to build enough units to make up its losses. Just as you say, that's the same as needing production for extra buildings. And as you argue, the extra cities the AI has provide more slots for unit production, especially as AIs love to spam encampments. The entire reason Civ IV was better at combat than Civ V was that spamming excess units was a surer route to victory with stack combat, while in Civ V/VI simple space requirements on the map give diminishing returns for having too many units in play at once. That's almost purely a result of production bonuses rather than combat ability.

My point was that the existing production boosts help it do one or the other, not both, and that it's simply easier to correct for poor combat AI by spamming extra units than it is to correct for poor building AI by allowing it to build more stuff, when that stuff will often be unnecessary holy sites, excess spaceports in low-production cities, aerodromes and air units they won't use, or extra Wonders. There's far more of this wasted effort than just situational Wonders (and many more situational Wonders than you're crediting. How often do you see Jebel Barkal built by civs that have no need for any of tourism, faith or iron, say?).

When you have an AI that's bad at both building and combat it's pulling the production bonus too thinly, since it needs to compensate for both, and is putting all its bonuses into treading water rather than actually increasing the difficulty for the human opponent.

 

[historix69]

...
The worst result of bad combat AI is that the AI loses units, which requires production to compensate by allowing it to build enough units to make up its losses.
...

It should be possible to define separate production modifiers for AI, e.g. one for each type of unit class (melee, ranged, mounted, naval, ...) and one for each type of building class (e.g. city walls, campus district, etc.) ... So with a +100% bonus on unit production AI could easily replace cheap units but would have normal production for wonders ...

 

[KandAffar]

Actually in BNW the AI actually drops building the wonder temporarily (confirmed with spies) once war is declared... so a "bluff" DoW despite your empire being half the world away or a war bribe was actually useful in helping you beat it to a wonder... people thought it was a cheap trick... now in RnF they made AI stick to finish building the wonder undistracted and people still complain.

People should still complain, because there is a very obvious option #3 - the AI evaluates the strength of enemy units inside its borders in comparison to the strength of its own units and decides whether to drop building a wonder based on that. If you want to beat the AI to a wonder by declaring war and actually sending an invasion force, well, then you probably should get to finish the wonder at the risk of losing the invasion force.