Combat AI

[aziantuntija]

Hi everyone! I started a new game (with the latest patch ofcourse) a couple of days ago and I found out that theres still one huge problem with civ5; weak combat AI. I play on immortal difficulty and my games are quite interesteting until to the point (usually later in the game) that I realize that I can just take over everyone who leads me in points (I only have time and conquest victories enabled). So basicly im saying that the combat AI is so weak that I know that I can win every game because of that. And no, I don’t have to conquer the whole world, I just take enough cities from that (or those) who has more points than I do (yes sometimes I must take many, MANY cities before I overtake them in points but nonethless), then just press the next turn to the end. The combat AI is so weak that it feels almost like cheating when im using my (smallish) military to gain victory, but it also feels stupid to NOT to use it since it is a part of the game, so technicly although it FEELS like cheating its not. The AI is surely having enough units to beat me but it just fails to execute. It also feels like that the AI only gets more stupid later in the game because then there are fighters, bombers, more ships (wich are also even relatively more powerful than early ships), long range artillery, AA guns and helicopter units and it just fails to use them effectively.


I also understand that there are propaply people who are satisfied with the current combat AI, so therefore I have this idea: Firaxis improves the combat AI but they also add difficulty setting just for the combat AI. Ok lets say that there would be three different difficulty settings for the combat AI; easy, normal and hard. This way someone can have deity difficulty for the game and easy difficulty setting for the combat AI, or vice versa, settler difficulty for the game and hard difficulty for the combat AI. Im not saying that many people would chooce those examples that I gave but im just simply saying that this way you can still have a poor combat AI paired with the AI having lots of bonuses (a.k.a. the deity and the immortal difficulty level).


I must admit that I don’t know anything about programming AI’s, so I wouldn’t be surprised if someone would say that it would be impossible to give different difficulty setting for the game itself and also for the combat AI in a game that wasn’t designed to have two different difficulty settings. But I still think something should be done so that using the military doesnt feel like cheating.

 

[Teutorian]

The combat AI is the game's biggest failings right now, I agree.
I just finished a game on Immortal and had the all the AI's but one completely beaten militarily by 70 AD with about 6 units, 1 of which I didnt even use (archer).

If they could fix the combat AI so that it could execute war it would make the game a completely different animal. I would have to be much more careful in how I proceeded with a war mongering victory and war mongers would have to fear worldwide denouncements. Hopefully they're working on it.

 

[Danild111]

It's not because some people are satisfied with the current AI, that they don't make it better.
Making a good AI takes a lot of money and time that they would rather spend elsewhere. It will also make the turntimes even longer.

 

[Teutorian]

I would take longer turns for a competent AI, honestly.
The game is getting boring because of the AI, and I say this as someone who loves the Civ series and strategy games. It sucks being able to just steamroll anyone.

 

[Danild111]

But who knows what they're up to. It wouldn't surprise me if they soon come up with an expansion that makes the AI better. Like BtS for civ4.

 

[aziantuntija]

I would take longer turns for a competent AI, honestly.

I agree.
If I could make a game AI better (any game! not just civ) just by having a bit longer loading times I would do it instantly. Besides, I don’t belive that making the combat AI better in civ5 would add a whole lot to the turn loading times on average

But who knows what they're up to. It wouldn't surprise me if they soon come up with an expansion that makes the AI better. Like BtS for civ4.

I can honestly say that I would gladly pay some € for better combat AI as a DLC or something (maybe someone should make a poll about this??). But I must also say that since my girlfriend also plays civ5 on my PC the better combat AI must be able to turn off somehow, OR that there would be different difficulties for the combat AI.

 

[Danild111]

It will automatically become less difficult when it has 3 units instead of 5, because of the handicaps at lower difficulties. No need to change the combat AI.

 

[aziantuntija]

It will automatically become less difficult when it has 3 units instead of 5, because of the handicaps at lower difficulties.

In that case, it will automatically become less difficult only if the combat AI stays at the same level.

For example, if you put horrible combat AI with 5 units against uber great combat AI with 3 units I wouldnt be surprised at all if the uber great combat AI would beat the living ***t out of the horrible combat AI with 5 units.

I was mainly talking about people who dont want any tactical combat in civ game or just simply people who dont care to spend any thought to it at all. They could still be massacring tons of brainless zombie armies if they wanted to, but also those who want real challenge from combat AI would get what they want to.

 

[snarzberry]

It is a very difficult task. Just think about how many possible moves there are during a war in a turn of a regular Civ game. How good play requires that you take into consideration things like terrain features, unit types and enemy unit types, cohesive group maneuvering, limited objectives etc. Those are the things the human thinks about. To produce a cost effective AI that can compete with a smart human in such a complex game world ( as opposed to, say, a 64 tile, two-toned chess board game world) by brute force following programmed rules is, it seems to my laypersons appreciation of the subject, a pretty hard thing to do.

For the time being the advantage in battle tactics seems to rest firmly with the human, resulting in the need for numerical advantage being with the AI.

Never mind, probably in 100 years the AI's of various games will be complaining to each other on AI-only discussion boards about having to play against stupid humans all the time.

 

[aziantuntija]

It is a very difficult task.

Im sorry but I disagree with you.

Think it like this: They have already build (at least somekind of) combat AI wich (at least somewhat) takes consideration of terrain features, unit types, enemy unit types etc. So I wouldn’t say that it would be that difficult task to actually improve the existing system. With that im talking strictly about the land combat, because it seems like they forgot to give really ANY kind of AI to naval units, I mean those aren’t always even shooting back at you but instead they wait enemy to sink them for crying out loud.

Never mind, probably in 100 years the AI's of various games will be complaining to each other on AI-only discussion boards about having to play against stupid humans all the time.

Too bad we are not around to witness this happen

 

[snarzberry]

Im sorry but I disagree with you.

Think it like this: They have already build (at least somekind of) combat AI wich (at least somewhat) takes consideration of terrain features, unit types, enemy unit types etc. So I wouldn’t say that it would be that difficult task to actually improve the existing system. With that im talking strictly about the land combat, because it seems like they forgot to give really ANY kind of AI to naval units, I mean those aren’t always even shooting back at you but instead they wait enemy to sink them for crying out loud.

Well, I'm definitely arguing out of my field here but it seems that it must be at least fairly difficult to do, cost effectively, as there are so few games that have good AI! I know the AI already takes into account unit types and terrain etc. but it's putting all of that information in and spitting out moves that we think are good moves that seems to be the trouble.

 

[aimlessgun]

To help imagine how difficult the AI task is, take a battlefield. Say 8 units on each side. 16 total. Some units are ranged, some units have 3+ moves, some units have 3+ moves AND are ranged. Then there are flanking bonuses, terrain bonuses, wonder and policy bonuses, promotion bonuses applied variously across all of those units.

The human brain is pretty amazing. We can look at this battlefield and immediately discard the junk moves, all the stuff that doesn't make sense or is 'obviously' bad. That's like 95-99% of the possible moves. Then we can focus and really think about the manageable subset of possible decent moves.

The AI cannot do that. It has to calculate everything.

And then say you come up with a plan, a model in your head of what you're going to do. You make your first attack, and the damage number is unlucky (or lucky). That changes your plan, and you have to adjust. For a human this is easy.

An AI has to recalculate everything again (assuming it's programmed to play perfectly)

Now I'm not an AI programmer so I don't know what techniques there are for getting an AI to discard bad options quickly, or if they can be optimized to better recalculate over the course of their turn while battle results happen. But I assume those things are very difficult to deal with.

 

[aziantuntija]

@snarzberry I might be wrong but to me it sounds like that you are saying that basicly they have made too complex game compared to the AI they can build? I must say that I agree with that, after all, the war IS more easy in civ5 than it was in civ4 wich had more simpler combat system.

I agree that there are very few games with good AI, and by good AI I mean that the AI isn’t necessarily ‘good’ but its good for that particular game (if you know what I mean). For example UFO Enemy Unkown (1994) is a great game and it had a great AI, the game wasn’t too complex (although it was fairly deep in that sence) so perhaps that’s why the AI looked very good in it. Im sure civ5 has WAY better AI than what the UFO Enemy Unknown had but the complexity of the game makes it look bad.

Why are we getting bad AI’s in todays games? Here’s my view:

Game makers concentrate too much on all the things they can add that the player can do in the game so that they can market all of its new features, but at the same time they seem to forget to make a decent AI for it that would understand these new features. Why don’t they make good AI for it? Well maybe because it sells better to talk about all the things the player CAN DO in the game, so therefore why focus on AI? I mean the big crowd doesn’t notice the horrible AI until weeks later and by then the game has already made its mark (or hasn’t made).

@aimlessgun You are basicly saying the same thing that snarzberry already said before; that its hard to build AI. But I wasn’t asking for a perfect combat AI, I asked for them to IMPROVE the EXISTING combat AI. As I already said it: They have already built the damn thing, now just focus on making it better, it should not be mission impossible to improve some games AI. Especially when considering the state that this game launched in, I very much doubt that they would NOT be able to improve the combat AI of civ5.


BTW sorry if I sound harsh, I sometimes sound harsh when im trying to make my point in english. And yes im a bad speller also, but if you get the point of my text then thats great.

 

[Fabiano79]

The combat AI is weak, but I think this problem is compesate by the huge number advantage on high dificulties.

Theres no playing wrong on civ, but if you are playing only time and domination, of course in the end you are going to win. Just a matter of time. You should try to play with all victory conditions enable, them you will see the real challenge.

The main problem with ciV for me still is the diplomacy and the lack of more complex choices. In my opinion the concept of the AI "trying to win" and the small tech tree are also core problems.

I cant see the combat AI getting much better without a proper expansion. One that I hope bring back religions, corporations and at least 20-30 more techs, units and wonders.

 

[starrywisdom]

Why are we getting bad AI’s in todays games? Here’s my view:

Game makers concentrate too much on all the things they can add that the player can do in the game so that they can market all of its new features, but at the same time they seem to forget to make a decent AI for it that would understand these new features. Why don’t they make good AI for it? Well maybe because it sells better to talk about all the things the player CAN DO in the game, so therefore why focus on AI? I mean the big crowd doesn’t notice the horrible AI until weeks later and by then the game has already made its mark (or hasn’t made).

I think you missed the one main point of software development as a whole. Goal Checkpoints and Money -- very few software developers go out of their way to make a 'bad game'. The fact of the matter is that some things are prioritize and some aren't. When you're doing 'information or objective triage' you have to put into a lot of factors. A.I. is one of the absolute hardest things to code, and grows exponentially when more and more gameplay mechanics are placed into the game.

When it comes to game design and the 'general fanbase' most people are going to want more emphasis on what they, as a player, can do. This means other features are going to get less attention by default. That being said, there's been a ton of games where the A.I. has released and been a lot...lot...worse lol.

@aimlessgun You are basicly saying the same thing that snarzberry already said before; that its hard to build AI. But I wasn’t asking for a perfect combat AI, I asked for them to IMPROVE the EXISTING combat AI. As I already said it: They have already built the damn thing, now just focus on making it better, it should not be mission impossible to improve some games AI. Especially when considering the state that this game launched in, I very much doubt that they would NOT be able to improve the combat AI of civ5.

It's worth noting that multiple patches(AFAIK) have had A.I. tweaks. A.I. is the single worst QA nightmare to patch. One bad line of code, or a weird priority set can make the whole thing go nuts. QA also has to do extensive testing to even the slightest tweak as it could make things go wonky. It's not like balancing a units hp or movement where it's simple testing - so any AI tweaks demand a much longer 'test' period which means large roll outs of A.I. tweaks aren't going to happen frequently.

 

[aziantuntija]

@Fabiano79 The bigger numbers on the higher difficulty levels doesnt seem to compensate it for me, nor does it do so for many others. It does not do so in the land combat and even less so it does in the sea combat, wich is not really combat at all. Also what do you mean with that “if you are playing only time and domination, of course in the end you are going to win”? Do mean just my games or generally every ones games? If you mean just my games then my answer to you is that this is how I want to play, I normally play pretty peacefully but usually I start to use more military towards the end of the game. I don’t want the game to ‘end before it ends’, I want to play the game to the finishline, so for instance thats why I disable culture and diplomatic. Also, if I were to enable some (or maybe all) of the victory conditions then nothing stops me from waging war more early in the game. My point is that the bad combat AI ruins the game for me by making it too easy to the point that it feels almost like cheating.

I also understand that there are other issues with civ5, but currently the biggest issue for me is the bad combat AI.

@starrywisdom Yes, as you said it; very few software developers go out of their way to make a 'bad game'. Still for example civ5 without any patches would be one those bad games, all we know for sure is that the paying customer was not in the top of that list of theirs. Still I doubt that Firaxis would have intended to have a bad game at launch but still that is exactly what happened.

When it comes to game design and the 'general fanbase' most people are going to want more emphasis on what they, as a player, can do.

This I agree with. However, this should not automaticly mean bad AI for the game. There are other departments to take away from than the AI, for example graphics, ESPECIALLY in a strategy game! Game developers should always make AI to match the complexity of the game, at least to a certain point. With that being said, there also might be a slight possibility that im just too good civ player, I mean because ive been playing since civ1 perhaps im just too much of a handful for civ5, im just not so easily distracted by all the changed game mechanics, maybe I can see the "core game" more easily . Who knows.. I only know that I want more of a challenge from the civ5 combat AI and im not the only one.

 

[markusbeutel]

Shafer immediately responded to Chick, saying, "I don't think it makes financial sense to make great AI," arguing that budgets can be more efficiently allocated to other areas of development.

http://www.gamasutra.com/view/news/...ans_Look_At_The_Genres_Biggest_Challenges.php

One could interpret from this quote that CIV V put more of a focus on other features, (graphics/UI), as opposed to AI, (and combat AI in particular). That's not to say significant effort wasn't put into AI, but I think the time and effort, (and money), on Firaxis' part - shows more through the Leader screens and the opening video.

 

[sebasm]

All these AI threads keep reminding me of one of the AI exams I had at Uni which only had one problem in it: logically design an AI to make a decision of whether or not it should buy a lottery ticket.

Ask yourself the same question. Did you say No? Why not?

And from here on out you can fill pages of things your brain actually takes into account when making the decision, filled with weights, prioritization tables, logical queries and so on. And these are without even considering human mumbo-jumbo like instinct or personality.

What I'm trying to say is that designing a competent AI (tactical in this case), even when it's restricted to say a chess board, is incredibly difficult! Just read about Chess AIs to get an idea.

And then add in the fact that this AI must also be boxed in the computational limitations of our Home Computers. And believe it or not nothing we can currently build comes even close to the computational power of the average human brain:
The Jaguar supercomputer can do 1.75 petaflops, which is 1,75x10^15 floating point operations per second. The average human brain does 10^25(figures vary per study). How many extra zeroes is that?
And you want the AI to do this competently in the seconds bewteen the turns?! No, no.


Sure, the Civ 5 AI could be better. But asking for it to be truly challenging is too much.

 

[BobDole]

Could it be better? Yes. But right now it's pretty good compared to its initial state, where it couldn't even line up units in a decent formation or coordinate proper attacks. Making a truly terrific AI in a 1UPT game is going to be really difficult, which is why there's all the bonuses there to help prop up the AI. Not anyone's favorite solution, but it's the most feasible.

Anyway, Civ's AI has always been bad, it's just been a lot easier to mask with stacks of doom. Considering that the AI was bad then, it's hard to expect it to be really good now in a more complex game.
Basically.....be grateful it's not more braindead.Hopefully they keep working on it. At least this patch I actually saw a short distance naval invasion launched against me.

 

[BigChiefLizzy]

I agree that there are very few games with good AI, and by good AI I mean that the AI isn’t necessarily ‘good’ but its good for that particular game (if you know what I mean). For example UFO Enemy Unkown (1994) is a great game and it had a great AI, the game wasn’t too complex (although it was fairly deep in that sence) so perhaps that’s why the AI looked very good in it. Im sure civ5 has WAY better AI than what the UFO Enemy Unknown had but the complexity of the game makes it look bad.

UFO was a great game - but I wouldn't say the AI was great; it was good enough for the game.
CiV - The AI cannot handle too many different factors, plus the planning needed to know when to retreat, and what to attack when, on what terrain type, what unit bonuses are available etc, see Sebasms post ....

The AI might get a bit better, and I like 1UPT but thats whats crippled it the most.