Do we need strong combat AI?

[stealth_nsk]

Many people complain about weak combat AI, but is it an issue? You could increase the difficulty level and have a lot of fun fighting skill vs. numbers. Or you could go multiplayer and fight skill vs. skill.

As I see it, the problem is not what AI combat is weak, but what it doesn't recognize it, so it seems to totally underestimate the power of human players. Consider those cases:

1. Player wants to go peaceful and builds small military for defense only. Neighbors think the player is weak and attack. Player is able to defend, but that's far from peaceful way.

2. Next time the player builds enough military for AI to recognize the strength. But it becomes clear what peaceful way doesn't make any sense at this point, since the military is enough to just conquer the world.

3. Player wants to conquer the world. AI don't consider player's army as significant threat until their own military is already ripped off.

So, as I see the simple solution will be just adding "AI stupidity" coefficient to military power estimation. So if AI calculates the power of an opponent, the human player's army strength needs to be multiplied by some number. There could be some variables here, but I think even fixed value (i.e. 1,5) will be ok. Surely, city strength shouldn't be multiplied the same way.

P.S. I don't really familiar with the XML files. Probably this type of coefficient exists already and doesn't work for some reason?

 

[Teutorian]

AI tweaks in certain aspects would be a welcome change.
You are right about the higher difficulty compensating, though.

I play the game on Immortal and right now I am in a game where i'm in a three way race with a very hostile Mongolian Empire to my left and a friendly but score leading American Empire to my right. I have to invade one of them very soon. If I betray America and try to cut them down to size I know the Mongols are going to denounce me and declare war soon after and if I attack the Mongols first i'm concerned America might start pulling too far ahead.

My current military cannot handle a two front war against immortal difficulty AI numbers. (six rifleman and 4 cannons)

In short, higher difficulty certainly makes it more interesting.

 

[spfun]

But the AI are so bad its impossible to lose a war, i'd like to see them pose a threat. For example Hiawatha rushed me early i had 1 spearman in my capital with oligarchy without walls. I stopped the rush dead which was comprised of 2 warriors 2 mohawks & 3 archers... Tactical AI is just brain dead currently & all the fixes so far haven't done much to it.

 

[DemonMaster]

Sure the AI or Combat AI if you want to narrow it down is not the strongest horse in the stable but I hardly see the problem. When I have played Civ 1-4, the only time the AI wins a war is by number. Civ 3-4 with the SoD is particularly easy since you can lure the SoD to screw up so it lands on a tile you want before you "handle it". I can't say that I ever have lost a war ever in the Civ history, until Civ 5.

I have lost a war in Civ 5. Sure as I said, the Combat AI do strange things sometimes, but I have actually lost a war in Civ 5. I can admit that the wars I have lost was by my miscalculation or my me screwing up (but that's how wars are lost).

I don't think this is the biggest issue with Civ 5. Except for some players that have some sort of exaggeration competition in various forums (soon we have someone that claim that this is a boring game because he always win with just a warrior on Deity).

I do would like to have the AI to evaluate muzzle better. On old civ games you could be at the top (counting muzzles) and act as a total ... and bully smaller countries around (for the fun of it). Here that is (has become) impossible. No matter what you do or what you have, the answer is "come and get it".

So I don't think the Combat AI could be so much better or it wouldn't much of a difference. I (as most players) will win (most) of the wars (unless miscalation and/or other mistakes). But I remember pre-patch (actually pre-pre-pre-patch or so) then I could make a show-off at the border so and demand gold or luxury and by having a strong army, one single small AI wouldn't dare attack. I could also have a strong army and be a builder (attacked perhaps only with multiple of AIs). Nowadays I have never manage to demand anything (regardless of size of the "gift" or my army) and I have been attacked by a one city AI (he was alone) when I have ten.

So some changes to how the AI do their evaluations wouldn't hurt.

 

[Teutorian]

But the AI are so bad its impossible to lose a war, i'd like to see them pose a threat. For example Hiawatha rushed me early i had 1 spearman in my capital with oligarchy without walls. I stopped the rush dead which was comprised of 2 warriors 2 mohawks & 3 archers... Tactical AI is just brain dead currently & all the fixes so far haven't done much to it.

Yes and no. It is very difficult to lose out to an ancient era rush by the AI. I normally defend my city while going for the NC from a very angry Civ (or 2) with one archer and one warrior without too much worry.

Late game, though, it gets easier to lose wars and territory if you're playing on higher difficulties. In the game I detailed above, if I invade the Mongols and America denounces and attacks i'm going to lose a good half my Empire at least.

Defending a city against 6 or 7 warrior units is one thing. Defending against an army made up of about 25 units, lancers, rifleman, cannons, while your army is on the other side of the map is another.

The primary problem with the AI is that it's just a sucker for the the Battle of Thermopylae tactic. Every game i'm Leonidas and the AI is Xerxes. It seems like every war the surest way to go is to set up a front line with ranged in the back and let them come to you, preferably in a bottleneck, and be grinded to pieces until they have nothing left, leaving their Civilization defenseless. I'm sure this is what everyone does.

 

[Callonia]

Because, in the game i just finished..

I ripped through three medieval kingdoms of Songhai, Aztecs, Russia

with two in renaissance usa and ottomans, I dec'd on all of them at same time cuz it'll be too easy otherwise.

I managed to get panzers out and rolling and supported them with infantry.

They were helpless. Nor built many units to make me choke on their dead.

They need to learn how to recognize a threat and deal with it properly.

They can't communicate with each others that a lone industrial power wants us dead and we should work together to kill him.

Oh well, it was funny at least, wiping out the 2/3 of the world with panzers and infantry.

took about 20ish turns on huge map. difficulty prince.

And me feeling free to backstab them and show lack of respect helped me alot.

You kinda get cold and hard after a couple of games i've played with douches staying at war for thousands of year and you can just kick them out of ur lands as you wait for ur tech to catch up with their military.

Throw tanks against my knights and caesar loses.

I throw tanks against AI medieval kingdoms and I win.

yup.

 

[aimlessgun]

I agree that adding some multiplier to the AI's evaluation of player strength would make diplomacy a LOT more realistic. It would get rid of the infinite wars, and some of the ridiculous DoWs, and generally make the diplomacy AI seem less idiotic.

I think 1.5 is too low as a multiplier though. More like 3x

Ultimately the AI does not have to be perfect, but it also does break immersion when they handle their army not only poorly, but in ways that make no sense in any circumstance, like putting archers in the front, fortifying a unit to be bombarded until it dies, running industrial units into the water to be picked off by triremes, etc etc.

 

[DemonMaster]

<>
Oh well, it was funny at least, wiping out the 2/3 of the world with panzers and infantry.
<>

Interesting. I just finished my German campaign and I broke all (took some cities but not all) my civs on my continents with Landsknechts (they could match that but I outnumbered them).

But during my panzer era I attacked them again and also a civ on a nearby continent. Those AIs on my continent had only manage to climb to Renaissance (at best). The one on the other continent was bigger and was on rifles but since I had so many oils and so many panzers, I overrun them all with strength and number.

The last civ was actually a match (regarding tech) and had a lot of units as well. I panicked and nuked him and hoped that if he have nukes he will only nuke my cities on his continent (newly taken cities from Persia and Denmark) and not my units (put away). It worked.

I won this campaign by ...

Points

Because the last capital was extremely heavily defended. I had nuked/bombed it down to low HP and manage reach it with a unit on turn 499. To late.

A problem with a destroyed country is that it is destroyed, very hard to move around. I should have started going nuts a little bit sooner.

But back to topic (sort of). Didn't I read somewhere that it was harder on the techs at the end or something like that. Doesn't that mean that If you have a hard time during the medieval times (like my neighbours had with me) you are screwed. Because I have never encountered such a huge time gap in Civ 5. It happened a lot in older civ (the famous tank vs. spearman) but until this patch. Most civ are at least on riflemen (perhaps they still have a few muskets around) when I have riflemens. Even small civs have that. They usually even have riflemen before me (that can be a priority but still). But with the latest patch, I am on the tech lead even with the military (and that's new, since I can be a little bit lazy with military techs sometimes). This was the reason for my domination victory attempt, to late attempt since I was planning for a science victory, but changed my minds when I entered the industrial era and saw that it can work.

 

[Becomedeath]

I don't think there is anything fundamentally wrong with the combat AI, it just doesn't appear to make a particularly dedicated effort to taking a city.

Where as a human player will tend to try and take a city in the lowest number of turns and then rethink their strategy if it's not working. The AI doesn't make these choices and will keep on sending units to their ultimate demise.

It doesn't pull back, restock and re-engage.

I have to say that I find the AI to be good at strategy, but weak on tactics. And I think that the difference is creativity. With so many unit types available, the AI behind combat is incredibly sophisticated, but lacks, and will always lack, creativity.

Even the most sophisticated chess AI systems work by assigning values and working out what it believes to be the human players logical response. But there is no real creativity there and Deep Blue wasn't exactly cheap or easy to make.

True tactical AI would be wonderful to play against, but as someone else in the thread has said, you want multiplayer for that. The computer will ultimately always loose because it can't come up with creative solutions or reprioritise effectively.

If it could, we'd live in a very different world!

 

[aziantuntija]

Do we need strong combat AI? Not necessarily. Do we need stronger combat AI than what we have now? Absolutely!

The point is that for me the game is too freaking easy! Its pretty much solely because of the bad combat AI. I play currently on immortal difficulty but there&#8217;s really nothing to play even on that difficulty level, I mean I can just easily conquer everyone I please and there&#8217;s nothing the AI can do about it, even though it has WAY superior unit count compared to mine AND its at the same tech level than I am. So because there is this huge loophole called the combat AI it pretty much waters down the rest of the game, at least for me it does. In civ4 I played on King because there the enemy army composed a real threat for my civ (mainly because the SoD was more easier for the AI to handle) so I didnt want to make any harder in order to survive, but in civ5 I play on immortal and still I can easily take down the point leader with realatively small army. So because they went to 1upt (wich is really fine by me) in civ5, Firaxis should of done better job with the combat AI in order for it to stay as competitive as it was in civ4.

Raising the difficulty level can only do so little, yes the AI gets more troops but it still doesn&#8217;t seem to have a clue what to do with them. Also it doesn&#8217;t seem to even separate ranged units from close combat units other than the ranged units are ranged, so it (still) doesn&#8217;t seem to understand to actually guard those ranged units. Also the only upgrade it seems to give to its units is instant heal, wich isn&#8217;t really good idea in the long run. These are just few examples.

Also I really don&#8217;t care to read answers about how hard it is to code a decent AI. Why? Because if they cant code a decent AI for the 1upt and they knew it before they even started it, then perhaps they shouldn&#8217;t of gone for the 1upt in the first place. Now that they went with the 1upt, they should also give a (at least) somewhat competitive AI to pair with that system so that it stops to feel like cheating when using military.

 

[aziantuntija]

Do we need strong combat AI? Not necessarily. Do we need stronger combat AI than what we have now? Absolutely!

With this I mean that there is a huge gap between current state of combat AI and a strong combat AI, so we dont necessarily need a this kind of giant leap from bad AI to strong AI. But what we DO need is improvement.

 

[Babri]

Another thing that can be done is that AI remembers the past wars. So if a human player has killed rifles with longswords, then AI would think twice before attacking such a player & will bring a much larger force to fight before declaring war.

 

[stealth_nsk]

I play currently on immortal difficulty but there’s really nothing to play even on that difficulty level, I mean I can just easily conquer everyone I please and there’s nothing the AI can do about it, even though it has WAY superior unit count compared to mine AND its at the same tech level than I am.

What if in case of your warmongering, AI would consider you much bigger threat? I mean denouncing you, making joint attacks, etc.? I think that would make it more difficult to do the world conquest. I think having this kind of improvement for several hours of work (including testing) would be just great.

 

[Becomedeath]

Also I really don’t care to read answers about how hard it is to code a decent AI. Why? Because if they cant code a decent AI for the 1upt and they knew it before they even started it, then perhaps they shouldn’t of gone for the 1upt in the first place. Now that they went with the 1upt, they should also give a (at least) somewhat competitive AI to pair with that system so that it stops to feel like cheating when using military.

1upt makes a significant difference to the way the game is played. AI stacking units on top of each other to make stacks as strong as possible and then just barging into terriotory isn't a game, it's not even difficult to program in terms of "intelligence", mainly because there isn't any.

Forcing players and AI to think about what units they want to put forward first, and then how the AI makes a threat assesment.

Think about it like this, is Modern Armor more of a threat than Artillery? What if there are 3 Artillery in range and only one Modern Armor? Add to that a mountain range and your defending units being 2 moves away. Which unit should the city attack first?

Civ5 doesn't apply threat effectively, because it will start shooting at the Artillery blindly. And why shouldn't it? Because the final blow will be dealt by the modern armor. If it takes out the Modern Armor first, the Artillery are a cake walk. But the AI will tend to leave units that aren't shooting at it when others are.

In that example, there are of course dozens of other factors to consider. But overall I'd say that the AI holds up rather well, although it needs to have the capability to make more educated decisions, it needs to consider more in it's logic rather.

My point is, that the difficulty in creating the type of AI these threads complain about is actually integral to the discussion. The gameplay has been overhauled in terms of military conquest and largely for the better.

If you're finding it so darn easy, I'd suggest Multiplayer, where you'll find your opponents being rather more creative. And if that's not an option, I'd strongly suggest you consider why you continue to play such a mindlessly easy game as you've indicated you find it.

As someone else has said, the AI needs to remember more. But not just within the confines of one game, it needs to build up a bank of data regarding your tactical playstyle.

You know the Aztecs are likely to be feisty and that an Alliance with the Mongols is likely to go sour, you also know how the computer fights it's battles. It doesn't have the same advantage as you and so it comes across as being rather dim because it's not adequetely armed to defend itself.

But I'm yet to see a game where the AI poses a significant threat without having an insane advantge. The fact is, you won't. Because ultimately, the challenge has to come from another human player, and that's the future of pretty much most strategy games IMO.

 

[aziantuntija]

1upt makes a significant difference to the way the game is played.

Yes, compared to SoD&#8217;s yes it does. And as you propably noticed(?) that I realize this and I also already said this same thing myself in my post that you were quoting.

AI stacking units on top of each other to make stacks as strong as possible and then just barging into terriotory isn't a game, it's not even difficult to program in terms of "intelligence", mainly because there isn't any.

Yep, and this is why the &#8216;combat AI&#8217; in civ4 was more competitive than combat AI in civ5, because it was easier for the AI. So I really do understand this, its not rocket science. I just said that I don&#8217;t need anybody writing an essay about &#8220;how hard it is to program a good AI for a 1upt&#8221;. The &#8216;combat AI&#8217; in civ4 is more competitive than the combat AI in civ5 is, and that&#8217;s the problem, they need to make it more competitive.

But I'm yet to see a game where the AI poses a significant threat without having an insane advantge. The fact is, you won't.

In case you haven&#8217;t noticed: Im not asking for them to build AI where it would pose a significant threat without having an insane advantage (although it would be nice but unfortunately it is impossible).

If you're finding it so darn easy, I'd suggest Multiplayer, where you'll find your opponents being rather more creative. And if that's not an option, I'd strongly suggest you consider why you continue to play such a mindlessly easy game as you've indicated you find it.

Im sorry but I don&#8217;t like to play multiplayer, I don&#8217;t really play any games in multiplayer, infact the game that im anticipating right now (Deus Ex: Human Revolution) doesn&#8217;t even have multiplayer at all. Also, I haven&#8217;t said that civ5 is mindlessly easy game, what I said is that it has a giant loophole, kind of like cheat system wich at the same time is also kind of a big part of the gameplay, at least in the higher difficulties. I also must say that suggesting what I should do or shouldn&#8217;t do isn&#8217;t really your concern at all, by saying &#8220;I suggest don&#8217;t play it&#8221; to people who complain about some games AI isn&#8217;t too constructive behavior.

Because ultimately, the challenge has to come from another human player, and that's the future of pretty much most strategy games IMO.

I don&#8217;t think that the human player vs human player is the future of the strategy games (well of course people are still going to play chess against each other). I personally think that the AI keeps on improving as the computers keeps on becoming faster and faster.


Also this I picked from another site, its actually a comment (last one) but I suggest reading the whole page because theres also Jon Safer talking about AI:

&#8220;There is an enormous number of players out there who don't get into strategy games because multiplayer is too competitive for them and the AI is Terrible. Playing online multiplayer against jerks or people who are far better than you (even good matchmaking seems to fail often) is a stressful and annoying experience and puts a lot of people off of these games. Not to mention that often the most competitive way to play a game is not the most fun way.&#8221; http://www.gamasutra.com/view/news/...ans_Look_At_The_Genres_Biggest_Challenges.php

 

[MkLh]

But I'm yet to see a game where the AI poses a significant threat without having an insane advantge. The fact is, you won't. Because ultimately, the challenge has to come from another human player, and that's the future of pretty much most strategy games IMO.

I doubt that and CiV wasn't definitely designed for multiplay. They have put even less effort to its broken MP section than to the AI. It makes sense as MP doesn't fit well with DLC-model.

 

[Misterboy]

I think the combat AI is pretty competent. I want to see the devs continue to improve it, but I agree with most ppl here that it's not the biggest problem.

I think a bigger problem is the Diplo AI. And I agree that the AIs miscalculation in starting wars is a problem, and that giving the AI a multiplier in reference to humans is a good idea. On higher difficulty levels the AI is more than capable of posing a serious threat and conduct wars with a fair bit of skill. Sure, it'll make silly mistakes, but often, their huge numbers of troops make up for it.

As suggested, if the AI calculated differently, you wouldn't see it enter wars it can't win, and/or, you would see it enter the same war with FAR more troops because, quite frankly, high level AIs can afford to pump them out.

 

[Misterboy]

Ultimately the AI does not have to be perfect, but it also does break immersion when they handle their army not only poorly, but in ways that make no sense in any circumstance, like putting archers in the front, fortifying a unit to be bombarded until it dies, running industrial units into the water to be picked off by triremes, etc etc.

Aimlessgun, do you still see these behaviors with the last couple of patches? I'm honestly curious.

Personally I don't see the AI lead with ranged units anymore. Sometimes it *appears* like they do, but that's because they sent a wave of melee units (which I squashed) and then a turn or two later the artillery shows up by itself. Granted, they should withdraw the ranged units, but I don't see them place ranged units up front any more than I do it (because sometimes it's a good idea).

I also haven't seen the AI fortify a unit to be bombarded to death in a LONG LONG time. Can't remember the last time it happened.

I do still see them jump into the water from time to time though And I do see them leave GGs orphaned. That GG thing is the one thing that still happens that makes me say to myself, "Really... you're going to do that!?"

 

[BSPollux]

The basic idea of this thread, that the AI should at least know that its skillm is low, is definately correct.

I dont play enough to decide how well this is allready factored in.

One other thing it should know is that sometimes its hopelessly behind and that it should take any chance to change it. Like that tiny country next to you that should definately join any big war against you just cause its his finaly opportunity to change its fate.

 

[blasto]

Is it just me or are barbs smarter than the usual AI? I constantly see them doing things that I never see from the usual AI; (always playing raging barbs, which doesn't affect their behavior as far as I know)

Fog skirting...they'll trail my units right outside of my own FOW... as soon as I cross a parallel, there are the same units that have been trailing me... usually accompanied by other units trying to herd me into a trap.

Pillaging and pillage rushing.. if I stop a barbs pillaging, usually more barbs will enter the city limits from the other side then.