Better AI in CIV 6 ?

[LDiCesare]

Secondly, while modding is an important aspect of the game, Firaxis is responsible for the AI they build for THEIR game NOT for the modded game.

I disagree with this. Their AI must be able to adapt to mods. It should look at combat values instead of unit names for instance. Otherwise, you just can't plain mod the game.

If you are playing Single-Player, with N other Civilizations in the game, what do you believe the expected win % should be for the player in order to call the AI in the game 'challenging'?

Desired player wins: 1/N at the level they want to play (i.e. same as MP).

Some of the very bad aspects of Civ V ai were the diplomacy. As in, resurrect an AI player who'd been killed, and see this fool with a single city and hardly any (backward) army hate you and DOW you a few turns later. It's clearly not optimal in terms of survival, and it's also bad from a RP point of view.

 

[Ulthwithian]

Acken: I am not trying to 'entrap' you. The reason I am asking the question is exactly to gather the data you correctly note will vary from person to person. Maybe I should write a poll?

Would you be interested in helping me craft the questions?

And I know that you didn't ask for optimal AI. My point is that there is no such thing as 'base AI' when you start asking for 'challenge' from one.

I will disagree with you, though, about the AI algorithm scaling well with bonuses. It might scale well with its direct bonuses, but not with the consequences of those bonuses.

As one example, it is possible to write an adequate 1UPT AI for a given number of units that is hopelessly inadequate if you double the number of units. You get clogging, which is exactly what occurs. On the flip side, if you write the AI to handle large numbers of units, it may be limited to only using a large number, and it is inadequate to effectively use just a few units.

So if you increase production in AI cities as difficulty increases, as one example, then given a fixed infrastructure cost, the AI will have more time to build military units. And then if/when it produces more units than it can effectively move, you get 'dumb' AI.

stealth_nsk: If you use too many parameters, of course, you get the situation where the AI isn't even comprehensible. It's a fine balance.

OTOH, I can remember times during my play of CiV where 'always/never get Wonder X' was a very well known 'strategy'. That may be due to erroneous AI parameter values, but you still get it.

Re: Eurekas, sure you don't need to expose it to the player. I was pointing out that Eurekas working for a computer may actually help the AI do its work.

Chess is an aberration here, I think. I would think the approach you would want to use, tactically, is some kind of cellular automata at the lowest level, but you have strategic objectives and grouping algorithms that would mitigate the congestion that can and does occur. Chess has way too many pieces that can move way too far, compared to Civ. Do you know if shogi algorithms are any cleaner?

And yeah, I know way too much of having to alter another developer's code... ><

LDiCesare: Thanks! I find that interesting, though. I would never play a game (for fun, anyway) where I only had an equal chance of winning.

 

[Acken]

Clogging is linked to the tactical ai which i noted is the bad part and also a design issue more than an ai issue. No matter how smart a human can be if you triple his units hellhave clogging issues.

 

[wayneb64]

I would not expect the AI in Civ VI to be a grand master player, or even close to on par with a skilled human player, but I would expect the AI to stop doing such completely mindless things like sending GP or siege wandering through enemy territory alone or moving every unit every turn, including jumping in the water to be slaughtered.

 

[Stringer1313]

I disagree with many of the comments here on the nature of AI. I think people are confusing the concepts of an "equal" with an "obstacle."

One misfortune of strategy games is that players expect the AI players to present an "equal" challenge rather than just a "obstacle." If you are playing Super Mario, there is no argument about the placement of coins and enemies having to match the human player and only win through strategy, with no "cheats." The levels in Super Mario are an obstacle. The AI players in Civ and other 4Xs are an obstacle.

This is further complicated by the fact that in order to be fun, 4X AI has to play irrationally. There is nothing rational about agendas; they inherently make AI unequal. Agendas are actually weaknesses for exploiting. Played rationally, the AIs would ditch their agendas and be aware they were in a game. They would do whatever it takes to win. It would be like if the turtles in Mario World suddenly became aware of the game, ate a mushroom, and proceeded to race the player to Bowser's castle.

Viewed this way, the AI has always been pretty good as an obstacle. It's just when you want it to do the impossible and also be an equal that it's disappointing. While I'm not going to argue against some improvements to the nature of the challenge, an unequal opponent isn't necessarily not an obstacle. Just like a huge pit and lots of enemies with no mushrooms or fireflowers can still defeat you in Mario Brothers, only people don't complain there because there was no expectation of equal.

This x1000000 Agree x1000000

This has always been my attitude towards Civ AI. I see each Civ as an obstacle, not a human equivalent.

 

[Malachi256]

I would not expect the AI in Civ VI to be a grand master player, or even close to on par with a skilled human player, but I would expect the AI to stop doing such completely mindless things like sending GP or siege wandering through enemy territory alone or moving every unit every turn, including jumping in the water to be slaughtered.

This.

AI should provide a reasonable challenge (I think everyone can agree with this), but what I also want is an experience that isn't totally immersion breaking. When the AI behaves bizarrely, that breaks immersion.

Ultimately it would also be nice if higher difficulty levels actually involved more complex and devious strats on the part of the AI, as opposed to just growing bonuses, but I understand that's a tall order.

 

[darko82]

Frankly, darko, it seems that you want Firaxis to make an AI that you can be proud of.

Maybe. I can talk about myself because some people may have even higher expectations

But my reference has been to CP, basically. It's not perfect, but it offers satisfying level of AI as of now (for me).

And I will be referring to it as long as Civ VI AI is "worse". Anyone can see the list of improvements CP did. The list is long.

 

[Lexicus]

Don't expect the VI AI to improve much, if at all, on the V AI. Hope to be proved wrong.

 

[darko82]

Don't expect the VI AI to improve much, if at all, on the V AI. Hope to be proved wrong.

Yeah, but they have had many years to prepare good AI for Civ VI. Almost ten years? considering it's not the first game in the series... they should evolve it over this time, not replicate poor outcome just to sell the brand again.

It is the mindset of the creators themselves. They need to feel it; they need to feel to develop, move further, have ambitions to be proud of themselves, create something unique, special, outstanding in terms of the AI, create the best AI in the game industry etc. etc.

And CP shows that it is at least possible to constantly create better AI, which people actually like.

So I hope they will not abandon this, and they will support and deveop the AI to the very end, to the level that is at least satisfying for most people.

For instance, I expect the AI to be able to take cities and manage their troops well to achieve that. But the list is long...

 

[Miravlix]

Civ V's computer controlled decision making was really strange, they didn't even allow it to step and shoot with ranged units (A mod later implemented it, so it's not like the engine couldn't handle it.)

They had one patch where the AI could do these "surprise" attacks, where it amassed a large army and then went for one of your towns. Most of the time, unless you got lucky and noticed the AI actually took the one city it focused on, but I think it was flawed and resulted in the AI getting it's units stuck, if it had trouble getting to the target.

After that, I think they simply gave up and went back to ultra basic decision making. Looking at the Barbarians city harassment system, it's nothing but a gimmick (Looks like it's one that doesn't break things though) on top of the same flawed decision making of civ v. The Barb's still dances around like in Civ V, with no possible decision path in the code that allow it to attack.

 

[Valmighty]

Is this something to be proud of? I do not think so... I would rather pose a question. Is THIS AI BETTER than Vox Populi, commonly known as Community Patch for Civ 5? Then, I would be kind of impressed. If it is not, then this is just a lame attempt to sell a product that is not up to the standards or expectations - unfinished product, so to speak.

Some systems implemented into the game are really nice, but is the AI capable of using these systems logically? Is the AI any threat to a human player? Can the AI even take a cities etc....?

No, but it is my answer. They virtually won't be able to make it worse than Civ5 AI. If it's about my personal wish, of course i want a MUCH better AI especially with the multi threaded performance. As some people here said, not obstacle/cheat but smarter AI

 

[darko82]

No, but it is my answer. They virtually won't be able to make it worse than Civ5 AI. If it's about my personal wish, of course i want a MUCH better AI especially with the multi threaded performance. As some people here said, not obstacle/cheat but smarter AI

Maybe, my question is: are they able to make it better from the CP AI? Even untill the last patch this game receives - let's give them another two years or so to make it better? Are they capable?

It can't be worse than Civ 5 vanilla AI, btw.

 

[stealth_nsk]

When people speak in terms of "this AI is good" and "this AI is bad", that's categories usually used by preschoolers. To have some real arguments, first comes the understanding what the AI is just one of the game mechanics to provide challenge/obstacles as pointed out several times in the thread. So with this we could start looking at some real thing.

Inability for Civ5 AI to move and shoot. Overall it decreases AI ability to provide challenge, but not significantly. It's more immersion-breaking thing than actual gameplay issue. Still, as far as we know, AI is capable to move and shoot in Civ6.

Inability to actually take cities in Civ5 without huge advantage in unit numbers and quality. That's real problem, because just adding more bonuses to AI will have consequences n other area of the game. But:
- The problem is only partially in the area of AI and partially in the area of too strong city defense in Civ5. Decreasing city attack, defense and regeneration would do the trick just well.
- We didn't see an AI taking city from human in Civ6, as far as we know, but it would be strange to expect this on Prince level against experienced Civ players. It launched some relatively decent unit groups for attack, though.

That's the level of discussion which I'd be happy to see here.

 

[Acken]

If Ai problems are due to a balance issue, for example ranged units being way too efficient in a human hands then people expect it to be balanced.

Im not sure many users care exactly what is changed for the game and combat to be more challenging. At least thats my experience with users of my mod. Be it balance or ai coding. Usualy the best way is simply a combination of both. The problem in civ5 is that neither was done and the same kind of combat seems to be making a comeback

 

[darko82]

When people speak in terms of "this AI is good" and "this AI is bad", that's categories usually used by preschoolers.

So I am/I am not satisfied by the challenge the AI gives us is a pre-schooler category?

To have some real arguments, first comes the understanding what the AI is just one of the game mechanics to provide challenge/obstacles as pointed out several times in the thread. So with this we could start looking at some real thing.

It is one of the things without which other things will not make any sense.

Inability for Civ5 AI to move and shoot. Overall it decreases AI ability to provide challenge, but not significantly. It's more immersion-breaking thing than actual gameplay issue. Still, as far as we know, AI is capable to move and shoot in Civ6.

Move and shoot is just one of the many things the AI should be able to do, but that is very basic. There are thousands of other things it should be able to do. So I would not say that this make the AI "good" or "bad".

But I agree, any stupidity seen by the naked eye is an immersion breaker.

- The problem is only partially in the area of AI and partially in the area of too strong city defense in Civ5. Decreasing city attack, defense and regeneration would do the trick just well.

The problem is the tactical AI and suicidal attacks, inability to move units or attack according to the game mechanics. Inability to use siege units, horse units etc. For instance, how stupid it is to attack a city with a legion, and then not be able to take a city by the catapult? The AI should be able to put and organize siege weapons properly, and be aware of their goals and limitations. Goal is to take a city. Limitation are the catapults that are used to decrease the defence, not to take a city etc.

There are thousands of problems probably.

- We didn't see an AI taking city from human in Civ6, as far as we know, but it would be strange to expect this on Prince level against experienced Civ players. It launched some relatively decent unit groups for attack, though.

Prince level is actually the base level, so yeah, any behaviour we can see on prince will apply on higher levels.

What we saw was pretty the same thing we could see over five years ago.

I do not like the idea that the AI can only take a city when it gets some bonuses, which you do not have, even though it still makes immersion breaking stupid illogical moves on the map. That's the "bad" AI, or one of the things that make the "bad" AI.

That's the level of discussion which I'd be happy to see here.

So is this discussion exclusive for AI programmers or hardcore gamers?

 

[Thormodr]

This.

AI should provide a reasonable challenge (I think everyone can agree with this), but what I also want is an experience that isn't totally immersion breaking. When the AI behaves bizarrely, that breaks immersion.

Ultimately it would also be nice if higher difficulty levels actually involved more complex and devious strats on the part of the AI, as opposed to just growing bonuses, but I understand that's a tall order.

I agree with this. The AI is participating in a story that you are telling.

I want the AI to play fairly competently and not shatter the immersion.

 

[stealth_nsk]

So I am/I am not satisfied by the challenge the AI gives us is a pre-schooler category?

I'm speaking about using terms "good" and "bad" without demonstrating understanding of what this means. "Like" and "dislike" aren't very different terms either. What really bothers me is what most of the people on the forum are really clever. So when I see discussion like:
- Civ5 AI was bad, hope Civ6 will be better.
- Civ5 AI was so bad, anithing will be better.
etc. - my brain just blows. The people can't be serious.

It is one of the things without which other things will not make any sense.

Yes.

Move and shoot is just one of the many things the AI should be able to do, but that is very basic. There are thousands of other things it should be able to do. So I would not say that this make the AI "good" or "bad".

I wouldn't use the word "should" here. It's the matter of a balance between amount of work and results. Without looking at the code I can't say how much work is required to fix it. But of course it should be quite easy to avoid such problem in AI written from scratch.

But I agree, any stupidity seen by the naked eye is an immersion breaker.

Yes.

The problem is the tactical AI and suicidal attacks, inability to move units or attack according to the game mechanics. Inability to use siege units, horse units etc. There are thousands of problems probably.

If we put immersion aside, that's not a big problem. Inability for AI to play according to game mechanics doesn't matter as long as it allows player to do so. If AI can't attack cities at all, the player will not have city defense part of the game, for example. But AI is generally provides challenge in all aspects of the game, so basic gameplay functions of it were fulfilled even in Vanilla Civ5.

Of course, we can't put immersion aside, so that's there the problem comes. And it's even bigger since immersion is quite subjective. Sometimes players may think AI is stupid if it does some clever thing player don't understand (due to lack of information).

Anyway, we surely seen some improvement. At least AI uses proper retreat, for example.

Prince level is actually the base level, so yeah, any behaviour we can see on prince will apply on higher levels.

It's not that simple. For example, Marbozir had total military superiority, because he's a good player and AI is on Prince level. For AI it would be smart thing not to attack at all, but it would be actually lack of challenge. If building enough units will guarantee you from AI attack, the game will become even more boring for those playing on low difficulty levels. So, AI HAVE to attack even against overwhelming forces. It tries to focus a city (in one of gameplays England nearly succed in this), it gets beaten, of course, and performs quite effective retreat.

In combination of both tactical effectiveness and game constraints (having to attack even against overwhelming forces), the AI works quite good. But the point is - we can't be sure it's actually good on higher difficulty levels, because situation will be different. AI will have more forces than human player and instead of being able to effectively retreat, we'll value ability to effectively attack.

What we saw was pretty the same thing we could see over five years ago.

6 years ago we've seen preview build played on Immortal. Huge difference.

I do not like the idea that the AI can only take a city when it gets some bonuses, which you do not have, even though it still makes immersion breaking stupid illogical moves on the map. That's the "bad" AI, or one of the things that make the "bad" AI.

In those terms it's impossible to make good AI for a game like Civ on modern hardware. Against more or less experience player, AI could take city only if it has significant bonuses.

And yes, it will make some stupid things. Partially because it's impossible to make AI properly handle a game like this, and partially because sometimes providing proper challenge means being ineffective - like attacking human player with overwhelming force or leaving some space for human player to expand in the first half of the game.

So is this discussion exclusive for AI programmers or hardcore gamers?

You don't have to be AI programmer to understand the principles. Actually, it's more about game design than AI programming. And nobody prohibits from participation in the discussion, of course. But if a person reads the thread and understand it, it will have some basic knowledge enough to discuss the matter on different level than just "good" or "bad".

 

[darko82]

- Civ5 AI was bad, hope Civ6 will be better.
- Civ5 AI was so bad, anithing will be better.
etc. - my brain just blows. The people can't be serious.

In fact, this was bad, and this is a solid feedback by an average user/player. Unless you disagree it was bad?

Without looking at the code I can't say how much work is required to fix it. But of course it should be quite easy to avoid such problem in AI written from scratch.

Firaxis has had the code all the time - including the modded code. If modders can make it work or fix some obvious issues with the AI, why the hell Firaxis cannot do it? They have as much time as they want or need to do this.

Inability for AI to play according to game mechanics doesn't matter as long as it allows player to do so. If AI can't attack cities at all, the player will not have city defense part of the game, for example. But AI is generally provides challenge in all aspects of the game, so basic gameplay functions of it were fulfilled even in Vanilla Civ5.

No, it's not as simple as that. If the AI cannot take cities, this means it cannot do lots of other things, like defending themselves logically, which means taking certain actions that a human player would take according to the game mechanics, which allows to perform such actions etc. Things like running away if the AI can see they are going to lose a unit etc.

It's not that simple. For example, Marbozir had total military superiority, because he's a good player and AI is on Prince level.

Have you seen that gameplay of inexperienced Civ guy who has had war with everyone and nobody even was able to hurt him? He was like beating everyone in the game.

Yes, it seems you can have war with everyone, and they cannot even hurt you. All of them. No units etc. The whole diplomacy suffers because of that. You do not need to care about lots of things if the AI does not manage the basics. Thus, certain mechanics become meaningless.

In those terms it's impossible to make good AI for a game like Civ on modern hardware. Against more or less experience player, AI could take city only if it has significant bonuses.

That's a wrong mindset. If you say something is impossible, the results will be mere. The belief in imposibility will become reality. That said.

 

[Ryika]

I'm speaking about using terms "good" and "bad" without demonstrating understanding of what this means. "Like" and "dislike" aren't very different terms either. What really bothers me is what most of the people on the forum are really clever. So when I see discussion like:
- Civ5 AI was bad, hope Civ6 will be better.
- Civ5 AI was so bad, anithing will be better.
etc. - my brain just blows. The people can't be serious.

No, they're obviously not, and everybody already knows that. Your whole scheme is basically: "I interpret what people say ultra-literally and ignore the fact that they're obviously using exaggeration, and are not thinking deeply about every single word they write or being are mildly trolling." That's literalism usually associated with some forms of autism - people who cannot understand words beyond their most literal meaning.

It really doesn't take much thinking to see a person write "Civ5 AI was so bad, anithing will be better." and come to the conclusion: "Oh, they're using a rhetorical device, the hyperbole, to make it clear that they really, REALLY disliked the Civ 5 AI."

They don't ACTUALLY think a randomly generated code would yield better results. Although... given what I've seen in Civ 5 I tend to think that it might.

Spoiler :

PS: That last bit was sarcasm. Please don't waste your time explaining why random code would not work better than Civ 5 AI.

 

[stealth_nsk]

In fact, this was bad, and this is a solid feedback by an average user/player. Unless you disagree it was bad?

I don't disagree. I'm just surprised to see "average user feedback" posted as discussion on this forum.

Firaxis has had the code all the time - including the modded code. If modders can make it work or fix some obvious issues with the AI, why the hell Firaxis cannot do it? They have as much time as they want or need to do this.

It's the matter of amount of work. As I said, AI fulfills all gameplay goals even in vanilla Civ5. And immersion could be so tricky thing - it's possible developers decided it's just not worth trying to improve immersion aspects of AI (as several small fixes will not heal it, of course) and focus on actual gameplay.

Also, a very important thing - in terms of immersion, modders work is estimated totally differently. Modders fixing move & shoot - "they are genius", developers doing the same - "LOL, there are so many problems left".

No, it's not as simple as that. If the AI cannot take cities, this means it cannot do lots of other things, like defending themselves logically, which means taking certain actions that a human player would take according to the game mechanics, which allows to perform such actions etc. Things like running away if the AI can see they are going to lose a unit etc.

Yes, that's not simple
- AI could take cities in some conditions, so you have to put some effort (although not big) into defending cities, at least on King+. Whether it's enough or not - is subjective thing, nt subjective. Formally, AI could take cities, so players have "city defense" aspect of the game.
- It's more gameplay problem than AI. Less strong city would make the situation better.

Have you seen that gameplay of inexperienced Civ guy who has had war with everyone and nobody even was able to hurt him? He was like beating everyone in the game.

Yes, it seems you can have war with everyone, and they cannot even hurt you. The whole diplomacy suffers because of that. You do not need to care about lots of things if the AI does not manage the basics.

You're speaking about Marbozir? He's very experienced Civ player and did his homework on known Civ6 information, so he's surely far from being "unexperienced". Just to note - he managed to maximize tech boosts, plus fully exploit Brazil bonuses (and probably ) into tech superiority, which lead to nearly 2 era advantage over the AI.

That's a wrong mindset. If you say something is impossible, the results will be mere. The belief in imposibility will become reality. That said.

It's practical point of view. Computational complexity is a measurable thing and processing power is limited. We know AI can't play as effectively as human. And, in general, it's non-goal as AI should provide good game experience, not play effectively.

EDIT:

No, they're obviously not, and everybody already knows that.

Well, as a non-native speaker I can't say I could 100% identify sarcasm, but usually I can, like in this your post, for example. To me half of the posters about AI being just bad are serious. If I'm wrong - well, you've just restored a big piece of my faith in humanity.