So, is the Artificial Intelligence finally getting some intelligence?

[Pax_Romanus]

Okay, now, I'm a warmonger. I want to crush my opponents, sure, but I want to fight to win, not just steamroll the whole world! A couple Civs here and there, yeah, but not everyone! So far, though, that's all I've done. Once I take that first city, I'm suddenly unstoppable. Unless I'm really unlucky and have no strategic resources whatsoever, the current AI doesn't come anywhere remotely close to even slowing me down, and the problem is the AI is useless at strategy.*I know how to destroy this pathetic three-city civ with half my military might but a full tech-era lead! I'll send my forces in range of his cities and crossbowmen, watch them die, have the survivors retreat, and surrender!* So when I heard that G&D is gonna make combat better, I was happy, to say the least. Now, my question is this: are we gonna see the AI get at least some improvement with strategy, or is it just gonna take me even longer to destroy that bunch of brain-dead lawn gnomes the AI calls an army? And if it does improve, or doesn't, I do have another combat-related question? Is the exp system going to change too? Cause that would kinda suck if it took ten times as long to get my keshiks to logistics. I don't really expect any definitive answers, but I haven't exactly followed all the updates, so if there are any answers, or even educated guessing, I'd really appreciate it.

 

[theadder]

It sounds as if they are going to compensate for the combat AI with the addition of the higher HP for units. This ought to help out, because it gives the AI more time to work out that it has made a mistake and then move the unit back.

I'm very interested in the naval AI and the idea that cities can be captured with melee naval units; this ought to add a lot to the game.

Supposedly they are also polishing up the AI in general.

It is too soon to know, but there are reasons to be very optimistic.

 

[Agent327]

Now, my question is this: are we gonna see the AI get at least some improvement with strategy, or is it just gonna take me even longer to destroy that bunch of brain-dead lawn gnomes the AI calls an army?

A question that has been raised by an actual game designer (and already repeatedly mentioned), and the basic answer is that 1UPT combined with the map size used in basic CiV simply will not result in a game that´s challenging AI wise. (In other words, the AI used with CiV can´t cope in a strategy game of this size; it works with Panzer General, it does not work with Civ, which uses a different map size.) The only way around this is if the community comes up with enough detailed-map-scenarios (sized like The Hundred Years War, for example) to make for any kind of interesting game. That CiV designers did not realize this (indeed in the early stages of development of CiV) is a major oversight on their part.

 

[MadDjinn]

"1UPT combined with the map size used in basic CiV simply will not result in a game that´s challenging AI wise" + "it works with Panzer General".

Is the sound of thinking well inside the box to me.

If you can claim that 1UpT allowed strategy in a 'small' map with a very outdated AI style, then it's easy to extend that basic assumption to understand that a larger map can be made up of a series of smaller maps, even overlapping, to have the same 'effectiveness'. Ergo, If it can work with a 'small' map in Panzer General, it can work by extension in Civ 5 if approached in a similar manner.

Do I think that Firaxis has the will to do that, given Sid's commentary on the matter? That remains to be seen, but to claim that it can't work due to the map being larger is rather poor logic.

 

[Babri]

Mind u, AI stands for Artificial Idiot not Artificial Intelligence.
They are going to improve AI + they are making changes that will help AI more than human players. Examples include battle lasting longer & more forgiving to mistakes made by AI. Also they stated that AI will make more balanced army so there is hope that we'll get a good AI eventually.

 

[Optional]

The opening poster is an exception, as there aren't that many players that have a cakewalk on Deity. You can't judge the game and the AI as a whole because of a very small minority for whom the game has now become too easy.

One unit per tile means I have to think more as a player, as unit movement has become much more particular.
I see the AI making tactical mistakes, but that doesn't mean that I can afford making those mistakes, so it doesn't affect the way that I play the game much.

The AI will always remain the AI, and never take into acount the amount of considerations that a human can. The AI would be better if AI designers could work on the AI almost indefinitely, but that's not going to happen.

On the other hand, if you see the AI doing something very stupid and you don't think it's acceptable, you could consider posting this in the bug section. Developers are reading that. They will probably also read your 'the AI is stupid' posts, but that's less constructive.

 

[mitsho]

I hope so. They claim to. We can't really know when we don't know how the battle system is changed.

If you take a look at VEM, where the AI gets an extra sight which already helps them to a degree. These are questions of the combat mechanics. What does it mean that we have a continuous upgrade line for archery units? Melee Naval units may mean more dangerous quick invasions possible. There's really no prediction to make here. Maybe the Spies will tell the AI more on your empire -> planning?

The other problems lie mostly in the AI. If you can make them upgrade their units instead of sitting on their gold, they get more competitive quickly. If you can make them not suicide their ranged units or make them realize that shelling a city without being able to capture it doesn't make much sense. It's the same with City States.

So in the end, there's three sides: Combat System (do they use the units correctly? do they have enough information? Can they focus fire on one unit? Which one?), Economy (can they buy the units they need? do they upgrade enough) and Bugs (the famous shelling without capturing).

 

[nokmirt]

I hope that because Dennis Shirk has knowledge of historical boardgames, he will understand that various things are missing in hex combat within CiV. In CiV you cannot overcomplicate things though. As we know it is not only a wargame.

In many hex strategy games there are rules for several things, supply, retreat, attrition, entrenchments, pursuit, weather, etc. But I feel Shirk is moving in the right direction by helping the AI build better combined arms forces, and upping the HP value of units from 10 to 100. The idea of limited stacking also comes from hex wargames, which has been dicussed as of late. From what I understand this will improve the naval side of the game. Although, I have not heard if any limited stacking rules will be applied for land combat.

I feel that the first step is to help the AI organize its forces. Being in a good position is many times half the battle. The AI also has to learn the concept of rock, paper, scissors better. In Panzer General the AI seems to know this concept quite well. Though many say the AI in that game is less than spectacular. I have the game and it gives me quite a battle. Mostly though, the turn time limit within the game for each scenario, is what gives a great advantage to the AI. The human can learn to beat the AI tactically, but can the human take all the objectives in time. This may be an insane idea, but perhaps wars in CiV can be limited by a number of turns. Also, the civ declaring war is on the offensive, and the civ attacked is defending its territory.

Now if the invading civ cannot take the enemy capital in so many turns then it loses the intiative, and the defending civ can begin a counter offensive, where the attacker then tries to defend his/her capital for a set amount of turns. If no one takes a capital a peace agreement is signed with whatever conditions can be worked out by both sides, just as it is in the game now.

Of course many wars can go on at once. You can pick and choose which ones you want to fight. You could end up in a defensive war against three or more civs at once. Any number of unpredictable situations could occur. What a crazy idea.

 

[MadDjinn]

minor note: "I hope that because Dennis Shirk has knowledge of historical boardgames,"

-- Ed Beach is the name you're looking for, Dennis is the producer.

 

[CTH]

I have the game and it gives me quite a battle. Mostly though, the turn time limit within the game for each scenario, is what gives a great advantage to the AI. The human can learn to beat the AI tactically, but can the human take all the objectives in time. This may be an insane idea, but perhaps wars in CiV can be limited by a number of turns.

War weariness... More or less a time limit of your attack, take too long time and your empire will be so unhappy that it wont be able to produce any new units or buildings.

There is already the happiness penalty of captured cities but proper war weariness would slow down a attacking human more (I assume that the AI would suffer less of this due to happiness bonuses)

 

[nokmirt]

minor note: "I hope that because Dennis Shirk has knowledge of historical boardgames,"

-- Ed Beach is the name you're looking for, Dennis is the producer.

Yes, your quite right its Ed Beach I meant.

War weariness... More or less a time limit of your attack, take too long time and your empire will be so unhappy that it wont be able to produce any new units or buildings.

There is already the happiness penalty of captured cities but proper war weariness would slow down a attacking human more (I assume that the AI would suffer less of this due to happiness bonuses)

I gather war weariness could begin to affect production and happiness as the war progresses. War after all does test the will of people. This could simulate a nation's resource capability for fighting a war, and its peoples will to continue that fight. This could go along with a certain number of turns to gain an objective, such as the other nation's capital. Perhaps you get a gold or some other bonus for gaining an enemy capital in so many turns. Perhaps the entire issue should be decided on war weariness, and the war weariness scaled depending on the difficulty level the human is playing on. The AI should get a higher war wearinees rating than the human, due to its disadvantage. I don't know? Molding these ideas into a viable system for fighting wars will make an interesting topic to dicuss.

See, we need a system that will help the AI gain some advantage without having to give it huge bonuses to make up for its shortcomings against a human. We know the human can easily outsmart the AI. So the question is how do we balance things in a different way? If anyone has any other ideas on this I'd be glad to hear about them. I need to think more on this as well.

 

[Pax_Romanus]

I kinda made this topic after being frustrated with the AI. It sent an invasion force of three pikemen and an archer through marshland to take out my army of 8 camel archers.And in a different game, it sent ONE pikeman to take out six knights. I know the AI isn't ever going to be on par with a human's strategic capacity, I am just praying that something is done so that it'l make less dumb moves and not suicide its units ALL the time.
P.S. Optional, I'm not yet at Diety; I've been upping the difficulty of my games lately, I've just noticed that the AI keeps making the same bad moves on all levels. It's not a bug, it's just a real lack of unit coordination. Although, the aztecs did suprise me last night. Instead of going for the lands I had taken from his neighbour, Monty circled around the long way and tried to attack my homelands. Only problem was, he DOW'ed me while his forces were about five turns from my borders, so I was able to rush-buy a few camel archers and station them at a chokepoint to defend myself. If he'd waited a few extra turns, I would have been defensless, and Monty might have taken a few cities. THAT'S the kind of thing I want to see improved; it's basic strategy to attack at your enemies weak point. The AI already targets "weak" civs, maybe with espionage they'll start attacking weak flanks more?

 

[nokmirt]

I kinda made this topic after being frustrated with the AI. It sent an invasion force of three pikemen and an archer through marshland to take out my army of 8 camel archers.And in a different game, it sent ONE pikeman to take out six knights. I know the AI isn't ever going to be on par with a human's strategic capacity, I am just praying that something is done so that it'l make less dumb moves and not suicide its units ALL the time.
P.S. Optional, I'm not yet at Diety; I've been upping the difficulty of my games lately, I've just noticed that the AI keeps making the same bad moves on all levels. It's not a bug, it's just a real lack of unit coordination. Although, the aztecs did suprise me last night. Instead of going for the lands I had taken from his neighbour, Monty circled around the long way and tried to attack my homelands. Only problem was, he DOW'ed me while his forces were about five turns from my borders, so I was able to rush-buy a few camel archers and station them at a chokepoint to defend myself. If he'd waited a few extra turns, I would have been defensless, and Monty might have taken a few cities. THAT'S the kind of thing I want to see improved; it's basic strategy to attack at your enemies weak point. The AI already targets "weak" civs, maybe with espionage they'll start attacking weak flanks more?

Yes it is essential we have a better AI. The AI needs to have a bigger resolve it gives up way too easy.

 

[Optional]

...the AI... it sent ONE pikeman to take out six knights.

This is a surprise for me. What the AI normally does quite well in Civ 5 is holding off an attack until it's got a nice number of units. Often you'll see a couple of enemy units not coming nearer to you, as they're waiting for backup.
This is a huge improvement over Civ III, where the AI always attacked piecemeal. It was standard in Civ III for the AI to have a galley sail over to drop off two archers (defense 1), while you had knights (attack 4) waiting for them.
Civ III had units with seperate attack and defense values. The AI had a problem with that. For instance, it always suicided its longbows (attack 4, but defense 1). I believe this is the main reason why units now have only one value for strength, and no seperate attack and defense anymore - the AI finds this easier to handle.

You'll still see the AI badly exposing its units. But you'll have a tough time if you're getting rushed by an Alexander or Darius at turn 60 or thereabouts on Immortal or Deity. The AI on these levels expands well, moves through the tech tree quickly, can build units quickly and a human player badly needs the AI's tactical mistakes.

I'm personally more concerned about how the AI acts diplomatically. what I find a real annoyance is that some civ with high boldness can attack me again and again and again, while never standing a chance. You could say, wipe them out if they're no match for you, but if the peace deal they're offering is more interesting to me than their lands, I'll take the peace deal.
I see the AI doing a lot of questionable things in diplomacy, and I'm used to better from the Civ series, while the tactical AI wasn't better before, so I'm mild in my judgment about that.

 

[nokmirt]

The AI has trouble taking cities, especially from the human player. Several times when it could have taken a city, it attacks with 1 or 2 units, instead of with 5, 6, or more even, it has on hand. If the AI had attacked with all its strength it could be far more decisive. Cities would fall at a far greater rate.

The AI does defend cities a bit better, but tends to use to many ranged units and not enough infantry. On top of that it sends in its horse to be massacred, for no purpose. It is usually also a foregone conclusion that the city attacked by a human will fall. So playing defensively is not the best choice for the AI. If it is forced to do so it must make the human pay much more than it does now.

When the AI attacks out it the open it usually has a decent force at the beginning of the war. Once that initial force is defeated by the human, or even partially defeated, the offensive begins to fall apart. The AI seems to retreat and give up too easy and many times for the rest of the war it fights on the defensive. Which is a sure strategy for defeat and loss of cities. I think even worse than Civ III does, which I just played recently. The AI in that game comes to fight it does not give in so easily as CiV's AI does. Civ III AI also lands troops amphibiously at different points, causing you to split your defenses. (CiV's AI could learn a trick or two about naval warfare from the Civ III AI, that's for sure!) It fights with tons more backbone than CiV AI. They need to make CiV AI more vicious. When those dumbbell AI civs declare war, they should fight that war as if they are angry. Not as if they are at Sunday school.

P.S.- One time the poor stupid AI attacked a CS. It never took it and tried well beyond 100 turns. It could have done so if it actually attacked with all available forces, instead of these tiny turn after turn pin prick assaults. To me that is just simply bad programming, and there is no excuse for that AI to be that helpless in a premiere game like CiV.

Furthermore, the games difficulty levels should not be required to give the AI bonuses to make up for its shortcomings. A Diety player should not have to rely on AI tactical mistakes to win, although the AI is never going to be nearly as good as a human. Instead a new system of war should be developed including war weariness, and so many turns to gain objectives, such as capitals. Each harder level would make war weariness more severe, and gaining objectives would have to be done in less turns. There also should not be an AI in each game two tech eras ahead of everyone else on the map. I could see some advanced tech over their neighbors, but to me its more interesting if the civs are more or less equal in this regard.

 

[mo123567]

I wonder if the AI is really being improved that much or if the rules are being changed to help the AI?
I don't really mind rule changes as long as they don't sacrifice fun in order to put a band-aid over an AI issue.

 

[vonbach]

The AI has trouble taking cities, especially from the human player. Several times when it could have taken a city, it attacks with 1 or 2 units, instead of with 5, 6, or more even, it has on hand. If the AI had attacked with all its strength it could be far more decisive. Cities would fall at a far greater rate.

I don't know I've been attacked by mobs of units and rarely do I have troubles defending myself.
The AI just isn't that good at sieges. Most of the time I just destroy their 1 catapult and it all falls apart.

 

[nokmirt]

I don't know I've been attacked by mobs of units and rarely do I have troubles defending myself.
The AI just isn't that good at sieges. Most of the time I just destroy their 1 catapult and it all falls apart.

If it would attack the city walls with all available units it would take the city. I have never seen it do this with enough resolve to finish the job. I have been surrounded several times and rarely has the AI used its available strength to wear that city down and capture it. Unless it was a size 1 city, and I could not get defensive units there fast enough. Even then it took the AI 4 turns to take a size 1 city, when it should have been captured in 1 turn. So, yes I agree not only is the AI terrible at sieges, but simply terrible at combat period.

The question is, will this truly be addressed in the expansion?

 

[Pizzaspy]

A question that has been raised by an actual game designer (and already repeatedly mentioned), and the basic answer is that 1UPT combined with the map size used in basic CiV simply will not result in a game that´s challenging AI wise. (In other words, the AI used with CiV can´t cope in a strategy game of this size; it works with Panzer General, it does not work with Civ, which uses a different map size.) The only way around this is if the community comes up with enough detailed-map-scenarios (sized like The Hundred Years War, for example) to make for any kind of interesting game. That CiV designers did not realize this (indeed in the early stages of development of CiV) is a major oversight on their part.

Sorry, but this isn't true. Civ V has awful tactics AI, but it isn't a given just because of 1UPT.

There are games published by companies like Matrix Games with much more complex rules, and much bigger maps than civ, that have excellent AI.

If there was an oversight it was underestimating how much work it would be to get right, but they can get it right.

Here is hoping they take the time to do it!

 

[Optional]

Civ III AI also lands troops amphibiously at different points, causing you to split your defenses. (CiV's AI could learn a trick or two about naval warfare from the Civ III AI, that's for sure!)

The AI in Civ III always knows exactly where you have your units. If there is some point in your empire weakly defended, then that's where it'll land its troops, guaranteed. But this gets predictable. Yes, you need to hold some units back to deal with these landings, but a couple of fast units will do, the AI doesn't attack with big forces amphibiously.
The trickiest points to defend against the AI in Civ III are islands. The AI loves to target those, and it's more difficult to get your units there. But the amphibious AI in Civ III is not a real force. It has merely been designed to keep the human player busy, I remember having read the AI designer himself saying something of that order.
Apart from the amphibious aspect, the human can build its tactics upon the knowledge that the AI will always go for the weakest point. The whole 'bikini babe tactic' was based upon this. The AI would try to move around a wall of your own strong units, just to get to some undefended worker behind it.

The AI in Civ 5, yes, it's certainly hit and miss, you're not wrong. I often see hilarious things, like an AI having a City State down to the last hitpoint, and each turn again firing on it with a large battery of ranged units. Oh AI, if you would only take one little puny melee unit over as well, that town would be yours, but with only archers, no, not in a 1000 years you will take it!
I've seen both England and Arabia being guilty of this, probably not coincidentally both civs with a ranged unique unit. There's probably something in the build flavour of these civs that makes them produce lots of ranged units.
The AI will also easily split its troops, like half a force will charge to one of your towns and the other half will go to a City State you're allied with and neither of those units are getting anywhere, because they're not attacking in a concentrated manner.

I don't think the rule changes that are being proposed now are primarily meant to help the AI. With longer health bars battles will last longer, but it'll become easier to save your units by retreating them in time. Will the AI learn to retreat its unit at the right moment? Big questionmark here. Maybe they will, it doesn't seem unprogrammable, but I've gotta see it first.
Embarked units being able to defend themselves will be good for the AI, but the current system is silly anyway. Embarked units are becoming defenseless and can't even be stacked with military vessels. The AI is going completely wrong with this, but also as a human player I'm not pleased. If I'm getting a military unit from an island CS I can't get it safely to my lands. Building 6 frigates and having them form an escort would be the only safe way, but that would be bonkers. No, that change is very welcome, for both the AI and the human player.