Is AI bonus too small?

[AsH2]

In Civ6 we all know that Deity AI gets bonused by 80% in production and gold, 32% plus free eurekas in science and culture, and +4 in combat. They also gets free settlers in their first few cities.
..
But why do designers design to cut off AI bonuses so severely? Civ6 is designed to be a good game, I know it is much more balanced in MP than Civ4 or Civ5.

However, is the AI bonus too small that make Civ6 standard speed PVE games dull and uninteresting?

I'm really looking forward to a higher level of AI bonus, for example, give Deity AI 150% on gold and production, 80% gold & resource discount on upgrade and 50% bonus on science and culture. Also we shall add free housing and amenity and loyalty for AI cities.
Can this match the level of difficulty and make the standard PVE games as interesting as Civ4 or Civ5?

No, giving AI more gold is like trying to solve candy thievery by giving babies more candy - that would just make the bigger problems worse as to further increase the exploiting gameplay and snowball effect on deity level. Is that really what you're looking forward to?

The district concept is coming out early and strong; which AI can't handle well and pushes human player to start on the run with a preset strategy instead of really "playing the map".
Midgame wouldn't be so dull and uninteresting if cities would be more organic instead of this board game approach that just reiterate same issues.

I think some people might be losing sight of deity- it’s supposed to be unfair. In Civ4, the description for Immortal was “Only the best players will beat this difficulty.” Deity’s was “Muahaha, good luck sucker!”

Anyways, people would do well to think of the difference between facing AI generic civ and AI Sumeria, Greece, or Korea. The AI isn’t actually that good at planning their districts, but the fact that Korea is told to spam seowons all day and Greece deploys acropolis everywhere leads to them becoming tech monsters. There’s no reason the AI couldn’t just be told to engage in actions that we now know, 3 years later, are near optimal. Imagine if every deity civ spammed campus+theater. They might not need so many bonuses then...

Warfare would be halfway solved if they just kept building units. I don’t know what prevents it, but fixing that will make them extremely scary, if the above change in district focus doesn’t make them all too strong to begin with.

Sure, I think deity AI beelining do make all sense, but still it will fail to adapt and "above average human player" will exploit it; weigthings won't solve fundamental issues like district planning because of too much uncertainties.

 

[leandrombraz]

I think some people might be losing sight of deity- it’s supposed to be unfair. In Civ4, the description for Immortal was “Only the best players will beat this difficulty.” Deity’s was “Muahaha, good luck sucker!”

What I don't like is that it affect specific things without making the overhaul experience harder, so instead of feeling like a challenge, it feel like I'm being limited to a narrow scope of possible strategies. I'll win just as easily, I just won't have as much freedom to choose how I want to win. My ideal "good luck sucker!" mode would be a difficulty where everything is equally harder, which is an idealized idea of balance that is unlikely to be ever reached, mostly in a game where one mechanic is inherently more powerful than any other in the game (war).

Anyways, people would do well to think of the difference between facing AI generic civ and AI Sumeria, Greece, or Korea. The AI isn’t actually that good at planning their districts, but the fact that Korea is told to spam seowons all day and Greece deploys acropolis everywhere leads to them becoming tech monsters. There’s no reason the AI couldn’t just be told to engage in actions that we now know, 3 years later, are near optimal. Imagine if every deity civ spammed campus+theater. They might not need so many bonuses then...

Warfare would be halfway solved if they just kept building units. I don’t know what prevents it, but fixing that will make them extremely scary, if the above change in district focus doesn’t make them all too strong to begin with.

That is a good solution but it conflict with how Firaxis tries to give each leader a specific playstyle that is aligned with their abilities, so I doubt they would set every AI to focus on campus+theater.

 

[Alexadamz]

That is a good solution but it conflict with how Firaxis tries to give each leader a specific playstyle that is aligned with their abilities, so I doubt they would set every AI to focus on campus+theater.

Yes. I guess the problem is that the second or third agendas are weaker for the AI behavior than the leader's exclusive one. Maybe tweaking that could change something.

 

[acluewithout]

I think a problem is that the role of Immortal and Deity difficulties aren’t really clear.

My understanding is that these were more “hardcore” levels in previous games, and so were sort of deliberately unfair. Instead, for many people here, these difficulty levels are becoming more about just adding in genuine challenge to the game.

As hard core play, I think being pushed to one narrow strategy is totally fine. But as an actual challenge level it’s terrible.

I don’t mind the extra settler at Emperor and Immortal. It’s not crazy and does make the AI more competitive. The huge number of extra units at Immortal+ is more of a problem, because you do just get warrior rushed and crushed which sometimes can be really lame.

The combat bonuses I’m fine with - it sort of adds in the bonuses the AI should be getting via tactical placement but doesn’t because it’s cheese. Although, from Immortal upwards, it means you really can’t have wars at the same tech level, which is a pity.

The endless thickets of XBows at Immortal+ are exhausting. I can’t really blame the AI for building them though.

I hope the AI keeps getting better. I’d like to drop back to Emperor and still be challenged. Beyond that, I think the AI would be improved by giving it some default policy choices and maybe some free units during the course of the game not just at the start, and helping them with unit upgrades (or change the resource system to be more about resources giving a combat bonus rather than being a hard barrier to upgrading units).

In terms of overall challenge, I think Civ could use some empire management - happiness, corruption, gold, whatever - and for that to scale with difficulty more.

Otherwise, I think the difficulty levels are okay.

 

[Lily_Lancer]

I think a problem is that the role of Immortal and Deity difficulties aren’t really clear.

My understanding is that these were more “hardcore” levels in previous games, and so were sort of deliberately unfair. Instead, for many people here, these difficulty levels are becoming more about just adding in genuine challenge to the game.

As hard core play, I think being pushed to one narrow strategy is totally fine. But as an actual challenge level it’s terrible.

I don’t mind the extra settler at Emperor and Immortal. It’s not crazy and does make the AI more competitive. The huge number of extra units at Immortal+ is more of a problem, because you do just get warrior rushed and crushed which sometimes can be really lame.

The combat bonuses I’m fine with - it sort of adds in the bonuses the AI should be getting via tactical placement but doesn’t because it’s cheese. Although, from Immortal upwards, it means you really can’t have wars at the same tech level, which is a pity.

The endless thickets of XBows at Immortal+ are exhausting. I can’t really blame the AI for building them though.

I hope the AI keeps getting better. I’d like to drop back to Emperor and still be challenged. Beyond that, I think the AI would be improved by giving it some default policy choices and maybe some free units during the course of the game not just at the start, and helping them with unit upgrades (or change the resource system to be more about resources giving a combat bonus rather than being a hard barrier to upgrading units).

In terms of overall challenge, I think Civ could use some empire management - happiness, corruption, gold, whatever - and for that to scale with difficulty more.

Otherwise, I think the difficulty levels are okay.

I guess most of you miss the point.

My point is not "Civ6 AI bonus is too large/small", but "why don't Civ6 AI inherit the bonus from Civ4/5"?

In Civ 4/5 people are satisfied with the AI bonuses and therefore the game is very interesting even on PVE, especially for Civ4.

So why do the developers decide to cut off those bonuses instead of inherit from Civ4/5, which is proved to be sufficient? Reducing gold/production bonus from 100~200% to 80%, and some other reductions, really make AIs look like fools, and the game becomes dull and uninteresting.

 

[acluewithout]

@Lily_Lancer No, we got your point. We just decided to talk about something else. Because the internet.

But sure. I’d like to see Civ V bonuses in Civ VI. Fine by me.

 

[ThunderLizard2]

Why not just fix the AI so it's not braindead instead of dumping more and more bonuses on top of a broken AI. Sounds like AI still sucks even with the Deity bonuses.

 

[Sostratus]

Why not just fix the AI so it's not braindead instead of dumping more and more bonuses on top of a broken AI. Sounds like AI still sucks even with the Deity bonuses.

If you glance through the thread, the issue is that making the AI "better" is an extremely hard problem that, even if we on the forum knew a silver bullet fix, probably couldn't really implement without the .DLL being released. Anyone who could make an AI that doesn't rely on simple behavior trees and basic weighting schemes, and can play civ6 competently, shouldn't be making civ6 AI- they should be making millions applying that AI skill to other problems.

So we discussed relatively simple things that we could kinda maybe change that don't rely on redefining the entire AI. For example, depending on your modding knowledge, one can currently directly change the table that defines some of the static bonuses AIs get - +% production, or starting units- and can even add some modifiers to only affect non human players. For example, you could make a modifier that only affects AI players that gives a hidden +2 to every district they have.
I don't know if we can actually change up some of their behavior trees yet, but once we can there are other simple things - like making them want to build more units, or campuses, or slot certain policy cards we know to be strong- that might make them play "better" even though we haven't made them any smarter.

But this runs the risk of actually making the game too hard. For example, try rolling a deity game where all AIs are Korea, to get an idea of what would happen if we made the AI prioritize tech more heavily.

 

[Laurana Kanan]

I don't know if we can actually change up some of their behavior trees yet, but once we can there are other simple things - like making them want to build more units, or campuses, or slot certain policy cards we know to be strong- that might make them play "better" even though we haven't made them any smarter.

AI+ did an admirable job of this (up through R&F) and Siesta had a fairly good understanding of the Behavior Trees.

 

[Lily_Lancer]

But this runs the risk of actually making the game too hard. For example, try rolling a deity game where all AIs are Korea, to get an idea of what would happen if we made the AI prioritize tech more heavily.

The main problem is not their focus. Under the current bonus (which is too small and too small compared with Civ 4 and 5), any rule-based system can be defeated easily, whatever they prioritize.

If they prioritize tech they still cannot survive an archer rush, and, more badly, they actually give the human player a lot of campuses to let the human player take more advantage.

What we want is to increase their bonus, bonus hide the noob face of Ais. For example, when CS begin to start with walls, no matter how noob they are at war, at least you cannot capture them easily.

 

[Bangau]

I agree that the high level bonuses in IV and V made high difficulties more challenging than the AI bonuses in VI do. It was way easier for me to beat deity in VI than it ever was in previous civ releases. I've even done a few deity OCC's succesfully, which I have never been able to do before. There are still situations where the game is challenging though and I still lose some games in the first 50 turns. But I would love to see one or two difficulties above deity where the AI gets even more bonuses and you have to play at maximum efficiency to beat them. How would you call a difficulty above 'deity' though...

On the rest of the discussion: I've recently watched a nice youtube series by TheGameCommon where he does an 'AI championship'. It really shows the main issue I have with the AI: it is totally unfocused. AI having a large science lead, but somehow deciding to hold off on building spaceports. AI with large armies, seemingly randomly moving them around instead of simply sending all of them towards the enemy. I'm starting to believe that the AI is deliberately programmed to be 'in the way of the players victory' instead of 'going for a victory'. In earlier versions of civ there was also (iirc) more resistance from the AI if you were clearly ahead of the rest. They would refuse trading and becoming friends e.g., in VI I see only a little (and too late) of that behavior. In most cases (not always) I can keep my alliances and friendships all the way to victory.

 

[Infixo]

Anyone who could make an AI that doesn't rely on simple behavior trees and basic weighting schemes, and can play civ6 competently, shouldn't be making civ6 AI- they should be making millions applying that AI skill to other problems.

Just wanted to point out that Civ 5 does NOT use behavior trees, and conversly Vox Populi neither. BTs are not necessary to code a good AI, it is just a tool to make the coding easier, more flexible and understandable for humans.

I don't know if we can actually change up some of their behavior trees yet, but once we can there are other simple things - like making them want to build more units, or campuses, or slot certain policy cards we know to be strong- that might make them play "better" even though we haven't made them any smarter.

It just so happens that all the examples you've given are NOT implemented in the trees. These are FavoredItems plus internal scoring subroutines.
The trees are used mainly to govern the units and some tactical processess like city build queues, etc. However, all the nodes that actually do things, are coded in the DLL.

 

[Sostratus]

It just so happens that all the examples you've given are NOT implemented in the trees. These are FavoredItems plus internal scoring subroutines.

I should have sat and thought about what I wrote because I knew it wasn’t technically accurate, but I get involuntarily triggered at the phrase “braindead AI”

 

[Equilin]

Why not just fix the AI so it's not braindead instead of dumping more and more bonuses on top of a broken AI. Sounds like AI still sucks even with the Deity bonuses.

And ruin the chance of laptops or non-gaming computers running the game? I would avoid that honestly. Better AI usually comes with more workload for the computer.

 

[bc1]

It does not seem like too much of a challenge to gimp a better AI so it can run on low power devices. A better AI would be nice for non-sandbox players with decent computers.

 

[Tomice]

Comparing Civ4, Civ5 and Civ6, the complexity of the game has increased drastically, especially from an AI perspective.

• In Civ5, the addition of 1UPT made unit placement a much bigger challenge. No longer was it enough to just send a stack of doom on its way. A large army became a hindrance for movement, and narrow mountain passes or landbridges were much more of an issue.
• In Civ6, the "unstacking of cities" added a ton of complexity. Before, the placement of a city was a one-time choice. Now, it's an ongoing challenge. It's extremely hard to plan a city in advance to make use of the best adjacencies even for a human player. Previously, you only sacrificed a single tile that couldn't be improved, now it's several. Previously, each tile only mattered on it's own. Now, each tile influences the yields of its neighbors in a very complex fashion!

It is safe to assume that a lot of the advances made in AI development and raw processing power were swallowed by the changes mentioned above.

Another issue:
IIRC, in the early stages of Civ5, it had an AI that was programmed to act like a human player. It would break the fourth wall and clearly show that it was playing a game just like the human.
Immersion wasn't the priority, and especially in later stages, it would ruthlessly backstab a 1000-year ally because he was about to win.
This wasn't very popular with the average player, and the AI was reverted to act more immersive.
(Please correct me if I remember this wrong!)

And another one:
In earlier iterations of civ, the AI was cheating more than nowadays, especially in regard to the fog of war.
It knew about unrevealed resources, which resulted in weird desert and ice cities that looked inexplicable for the human at first - until massive oil or uranium sources appeared within their borders.
The AI also ignored fog of war in combat, too.
(Again, please correct me if my memory betrays me.)

Finally, a very complex AI might be more challenging. But in a game with as many variables as Civ, it is also at a high risk of utterly failing in edge cases. Imagine a minor balance change for aqueducts in a patch - it might result in a roman AI not building anything else but baths if the AI isn't simple and robust.
Sure, you can partially test this in beta, but often enough it will be very difficult to see the issue and understand what went wrong.

 

[Pure24]

For starters, AI is indeed incredibly hard to fine-tune. I see the issue being about too many elements to the game; to many things to deal with.

But see, I think the problem is that human players can utilise exploits to get ahead. As an example, I just don't see the developers ever programming an AI that chops in wonders to rush them. And thus human players will always be able to beat them to the punch. The best way is to make the AI get to the wonder faster in hopes of being able to build it in time... and we're back to bonuses.

I heard somewhere that Civ 6 AI is programmed to roleplay. It plays the way a civ is designed to work (Korea with Seowons everywhere, focusing on science). Doing that and also optimising gameplay the way a human does is something that isn't really feasible. Heck, cutting edge AI isn't even there yet!

Also, would players enjoy a game against an opponent that chops everything as soon as they're able to? If Deity AI could exploit those things, it would be game over for all but the most experienced players - and even then you'd still need luck!
And how does it work in the context of a civ that can't chop (e.g. Maori), or whose agenda is to not destroy the environment (e.g. America)? Those Civs would remain behind, unbalancing the game to breaking point.

So unless AI gets progressive bonuses to aid them to keep up with the human snowball, I don't see much changing. And still players will complain that they're not playing a fair game. Rubber banding the AI isn't a good idea. So sometimes, we the players demand unreasonably of them. Which is not to say that every complaint is unreasonable.

Combat AI kinda sucks! I've never played a game where an AI was able to (or at least, close to) win domination. Neither can it win Diplomatic victories. So what's the point of playing against an opponent who CAN'T win?

For me, that's what I'd like to change the most

 

[Sostratus]

Rubber banding the AI isn't a good idea.

People say this, but I don't think it's true.
What people complain about is that the AI isn't a threat once the player starts getting a huge lead. This is, of course, true in almost every scenario where one civ has a massive tech advantage, and even doing something extremely blatant, like say on deity the AI will start getting free/1 turn techs so that if you are in era X the AI will be in at least era X-1, then we still have a lead being a lead, but you can't totally phone in the victory; if you get super sloppy at the end then an Ai could conceivably catch you. (I mean even that isn't likely. You get a tech lead by having more science output, so even if you gave them every tech the player has you would still see them just barely keeping up.)

A one era gap, for example, is quite large; You could have steam power and the Ai could just be getting gunpowder - they are only one era apart! But where it would become dangerous - at least on deity - is if you get close to the end, then screw around for a while, the deity AI that is big on production but not science might be able to leverage this that it could put its production towards the rocketship- where before it never could due to not enough science. And it would still be a long period to catch up.

And this is just handing free stuff out. You could have much more clever game mechanics that would keep the civs a little tighter together without anyone noticing a penalty. When done properly, a civ with less science output will never pass a civ with more, ever. For some reason that seems to be a sticking point - the design of the "catch up" is to govern how big that gap is which is turn is the "cushion" players have to work around. You could, even under the egregious cheating scheme i mentioned, easily circumvent it by ensuring your empire is fully built out at all times, capitalizing on your lead, rather than endless science spam that would become fairly worthless except in minimizing the turn you can win on.

 

[acluewithout]

People say this, but I don't think it's true.
What people complain about is that the AI isn't a threat once the player starts getting a huge lead. This is, of course, true in almost every scenario where one civ has a massive tech advantage, and even doing something extremely blatant, like say on deity the AI will start getting free/1 turn techs so that if you are in era X the AI will be in at least era X-1, then we still have a lead being a lead, but you can't totally phone in the victory; if you get super sloppy at the end then an Ai could conceivably catch you. (I mean even that isn't likely. You get a tech lead by having more science output, so even if you gave them every tech the player has you would still see them just barely keeping up.)

A one era gap, for example, is quite large; You could have steam power and the Ai could just be getting gunpowder - they are only one era apart! But where it would become dangerous - at least on deity - is if you get close to the end, then screw around for a while, the deity AI that is big on production but not science might be able to leverage this that it could put its production towards the rocketship- where before it never could due to not enough science. And it would still be a long period to catch up.

And this is just handing free stuff out. You could have much more clever game mechanics that would keep the civs a little tighter together without anyone noticing a penalty. When done properly, a civ with less science output will never pass a civ with more, ever. For some reason that seems to be a sticking point - the design of the "catch up" is to govern how big that gap is which is turn is the "cushion" players have to work around. You could, even under the egregious cheating scheme i mentioned, easily circumvent it by ensuring your empire is fully built out at all times, capitalizing on your lead, rather than endless science spam that would become fairly worthless except in minimizing the turn you can win on.

My issue with rubber bands is that I think should be steam rolling some Civs, just not all of them. I want to have AI fall behind so I can go beat them up. The catch is that I also want some Civs to keep up with me, and those Civs and I can then fight for global domination. And I also sometimes want to be the Civ that’s fallen behind, and now I’m struggling to catch-up or looking at victory conditions that don’t require superior tech.

 

[Pure24]

My issue with rubber bands is that I think should be steam rolling some Civs, just not all of them.

Exactly. I'd like to get rewards for playing (that is, stomp my opponents)... but at the same time, I'd like a challenge.
There's that sweet spot which I think is what would be the best