Without cheats, will the AI be able to press the offensive?

[korn469]

It appears from all of the changes in Civ3 that the balance has tilted more in favor of the defender. Probably the biggest advantage the defender has is the use of roads, why the attacker has to plod along at normal rates. This is compounded by low movement rates. The defender also has other advantages, including forced labor and the draft, so if they are in a pinch they can mobilize a few extra troops.

Without the ability to use road or rails, the howitzer blitzkriegs of Civ2 are a thing of the past. Infantry on the offensive poses little threat, and the defender can easily use their transportation network to bring troops to the front and engage enemy troops in preselected kill zones.

These challenges certainly won't overwhelm a crafty human player, however the AI could certainly have problems with this.

I would say that a human would go on the offensive like this:

*Gain air superiority
*Cut holes in the defenders transportation network
*Bombard entrenched defenders with air power
*Bring up fast land units to seize the objectives
*Use paratroops when needed

Will the AI know to do that? My worst fear is that the AI sends army after army with cutting off my ability to reinforce my front, and then ends up losing its massive armies.

So in your opinion will the AI be an Atilla the Hun or will it be a George Pickett (Pickett's charge)? What does the AI need to know for it to be successful on the attack?

[TechWins]

If the game is to be potent, the AI will have to be able to handle the battle situation. Which, in all honesty, I think it will be able to do quite efficiently.

Also, all of your military plans makes me want to play the game so bad.

[Boliao]

I think at diety or emperor levels where the AI cheats, they might travel on rail or road as per normal, that might improve movement rate and blitz might be possible. Maybe ...??

[Mongoloid Cow]

Destroy the roads connecting their important resources with their cities, so they cannot replace the units you destroy. And the attacker evens the score. A successful attack in the modern age would require simultaneously conducting air raids, destroying resource and transport roads, and using tanks to take out the easy objectives, and using them to sheild your artillery from the enemy (remember they can capture catapults, cannons, etc.) Doing too much or too little would give the defenders the advantage. And failing to have enough defense in the conquered cities could see a successful counter-attack.

[korn469]

Mongoloid Cow

Destroy the roads connecting their important resources with their cities, so they cannot replace the units you destroy.

they might not be able to rebuild the more sophisticated units in the cities the attacker besieges, but they can use the draft to quickly get cheap defenders and then they can also bring up reinforcements close to the battle on their road/rail system

A successful attack in the modern age would require simultaneously conducting air raids, destroying resource and transport roads, and using tanks to take out the easy objectives,

i agree about the air raids and destroying the transportation network, but before airpower i think the attacker will have trouble

first if both sides have equal tech, then i don't see how an attacker could cut the defenders transportation network without a fair amount of cunning, something which i am afraid the AI will lack

using them to sheild your artillery from the enemy (remember they can capture catapults, cannons, etc.)

if the defender has bombard units, especially bombard units on fortified hills, then they will have a range advantage against other bombard units of the same type, plus until howitzers i'm sure catapults, cannons, and artillary have a movement of 1, giving the defender the first shot...plus as long as the defender has their bombard units stacked with normal combat units they are safe from capture

And failing to have enough defense in the conquered cities could see a successful counter-attack.

exactly!

all in all pulling off an attack against a defender of equal strength with take skill, will the AI be up to the task?

[Adm.Naismith]

Originally posted by Mongoloid Cow
Destroy the roads connecting their important resources with their cities, so they cannot replace the units you destroy. And the attacker evens the score.

You are suggesting smart (modern) tactics that can work very well, if no units overpowered will result in unbalanced battle and shortcut tactics to the victory.

The point of korn is that this advanced tactics are way too smart for any know game A.I., hence he doubt that Civ III AI can manage them as an attacker.
I suppose the AI will be quite succesful until someone discover its "pattern" and develops workaround tactics.

CTP II syndrome anyone?

[cort]

Originally posted by korn469

Will the AI know to do that? My worst fear is that the AI sends army after army with cutting off my ability to reinforce my front, and then ends up losing its massive armies.

So in your opinion will the AI be an Atilla the Hun or will it be a George Pickett (Pickett's charge)? What does the AI need to know for it to be successful on the attack?

i am pretty sure that the AI will NOT BE CLEVER ENOUGH to perform such coordinated attacks.

it will cheat at higher difficulty levels; produce units at lower costs, probably transport them to fronts more easily than human player, even spawn units and all of those will contribute to a more challanging gameplay for the human player but they are all defensive issues.. since cheating will be much more obvious during an attack, i suggest not to expect much from AI assaults..

not to blame Fraxis though, current technology level does not permit that. think of the chess game; compared to civ, it is a far far simpler game and it takes super-duper computers to challange a human player..

[Rosacrux]

Rabish, even a entry-level chess programm can beat the **** out of a mediocre human chess player. An advanced propramm can beat a master or be beaten by him.
It takes a Deep Thought/ Big Blue computer to beat Casparov - you know he is not excactly the average human player

[Grumbold]

I expect the AI to be better in defence because it has all the advantages you describe. The AI does not have to cope with an instant attack on any part of its empire reachable by (rail)road so it can devote its forces to the area under attack. It has also become much much more difficult to wipe out a player early so the AI should have a much closer parity in number of cities to the human player. This bodes well for its defensive potential.

Going on the offense is a completely different matter. I suspect its strategy will involve building up a huge quantity of troops and then unleashing them toward one of your cities. If that city proves too well defended then they can at least have a whale of a time turning the surrounding tiles back to wilderness. The minute you try and counterattack them the defensive advantage is in their favour. If they are vaguely clever they might fortify defensive units around your city and march the rest on toward the next one or send up more and more artillery to pound you with. I don't expect the AI to be offensively brilliant, but I do expect them to be able to cause lots of devastation and inconvenience.

[Alex]

if both sides have equal tech, then i don't see how an attacker could cut the defenders transportation network without a fair amount of cunning, something which i am afraid the AI will lack

Precisely. That's why espionage options would be so important. With a well organized intelligence work, the attacker could cut the defenders transportation network (using sabotage techniques, for instance), and the AI could be programmed to use such means whenever it attacks.

If you object to this saying that there are no more "spy units" on the game, I reply saying that those actions could be performed inside the screen containing the espionage options, a la CTP and its use of Public Works. Of course, those actions would have to be somewhat limited or countered by the AI, otherwise it would be easy enough for the human player to beat the AI on the intel department.

[DaveV]

A lot depends on the rules on movement after disembarking from a boat. In Civ 1 (IIRC), units couldn't disembark and attack, so naval invasion was difficult. In Civ 2, disembarkation only counted for one movement point, and the most effective AI attacks were naval landings by fast units to take a lightly defended coastal city. Putting a couple knights on a boat, landing them and attacking a city shouldn't be beyond the AI's capability. If the AI can actually establish a bridgehead and exploit their victory, I'll be impressed.

[korn469]

Rosacrux is correct

the best chess playing computer in the world will probably get to the point where it can beat a grand master the majority of the time, but that is more of a hypothetical situation

how good will the AI be at coordinating attacks?

lets look at the AI military strengths from SMAC

*built large number of units
*would advance in force
*usually maintained a good mix of units (but usually focused too much on infantry)

and its weaknesses

*its large stacks would die to collateral damage
*very poor at launching amphibious attacks: it would build units with the marine special ability, but i never saw it actually attack with those units from ships
*could not use airpower (this was its worst flaw in SMAC)
*would almost always attack the closest base
*had no concept of chokepoints, so it would continue to send large numbers of units on the quickest path and wouldn't change paths even after horrendous losses
*it was not able to identify weak spots in the defense
*would engage in wars across the map, yet it couldn't mount an effective intercontinental war
*it only had a limited grasp of key objectives it needed to sieze in order to win the war

Alexnm

besides the AI maybe not being able to handle the spy options, as far as i know, the sabotage option only disrupts what is being produced, and that a player won't be able to sabotage enemy tile improvements

so before airpower, it will take a great amount of skill to overcome a well defended civ

[HalfLotus]

The chess AI comparison isn't fair. Civ has many more systems involved, and many more move/strategy options.
Firaxis hasnt said a damn thing about the AI. Hopefully they're keeping it as a pleasant surprise. They have been saying that single player is the focus of Civ. Let's hope they walk the walk.

[Steve Clark]

Originally posted by DaveV
A lot depends on the rules on movement after disembarking from a boat. In Civ 1 (IIRC), units couldn't disembark and attack, so naval invasion was difficult. In Civ 2, disembarkation only counted for one movement point, and the most effective AI attacks were naval landings by fast units to take a lightly defended coastal city. Putting a couple knights on a boat, landing them and attacking a city shouldn't be beyond the AI's capability. If the AI can actually establish a bridgehead and exploit their victory, I'll be impressed.

So will I. I have more hope that it can acheive a non-military win better than trying to manage dozens of units in a coordinated campaign across a varied terrain against opponents that are not going to sit there simply defending. But then again, all it matters to some folks is whether the Knight is shown with the proper chainmail.

Excellent thread, good to see you again DaveV.

[joer]

One of the biggest advantages of the AI in the past (I am positive about SMAC, suspect about Civ II) was complete information. It knew where all of my units where, how big my defenses were, etc.

That became pretty stupid when certain tactics (hiding in fungus) and unit special abilities gave the advantage of making a unit invisible -- when the AI still always knew where it was.

However, the SMAC AI was too stupid to take full advantage of that knowledge. Since it was incapable to coordinate attacks, it could not direct its forces towards cities which were nearly undefended. For me, the result was usually that all my big center cities were defended by one of the puniest, weakest unit possible -- I just counted on the fact that the AI would never get near my capital.

So...I think a big question is whether or not the AI will have complete game information again, and if it's capable of drawing advantage out of it.

Even without too sophisticated tactics a la chop'n-drop, fast surprise attacks with combined forces against selected cities would mean BIG trouble for me. I am not sure if I can afford to defend every city with several of the latest defenders.

[Crouchback]

I have absolutely no hope of the AI managing complex combined attacks later on in the game, which has nothing to do with the changes in cheating but everything to do with the stupidity of silicon. That is why I'm glad to see the position of defenders strengthened, to help the AI stop me.

But, what about the earlier game when all the AI has to combine is catapults/cannons and PBI? That stage of AI warmongering might improve dramatically under the new system.

David

[Grumbold]

Originally posted by joer
So...I think a big question is whether or not the AI will have complete game information again, and if it's capable of drawing advantage out of it.

In a fog of war game this is certainly one of the best advantages to be had... unless the AI is so stupid you can lure it into attacking by emptying a city hundreds of miles from the border, then restock it a couple of turns later after you've massacred his invasion force.

Even without too sophisticated tactics a la chop'n-drop, fast surprise attacks with combined forces against selected cities would mean BIG trouble for me. I am not sure if I can afford to defend every city with several of the latest defenders.

It seems like really fast attacks will be impossible until at least the flight era. That is going to give me a hard enough time on the offensive so I don't hold out much hope for the AI

[albiedamned]

Originally posted by Crouchback
I have absolutely no hope of the AI managing complex combined attacks later on in the game, which has nothing to do with the changes in cheating but everything to do with the stupidity of silicon. That is why I'm glad to see the position of defenders strengthened, to help the AI stop me.

But, what about the earlier game when all the AI has to combine is catapults/cannons and PBI? That stage of AI warmongering might improve dramatically under the new system.

David

Pardon my ignorance, but what is PBI?

My opinion on this whole debate is that we should not raise our expectations too high regarding the AI's ability to coordinate an attack. There's never been a game that had an AI that could strategize like a human. Nothing against the coders at Firaxis, but if they had achieved such a breakthrough I'm sure they would be bragging about it.

Having said that, I agree that defenders in general will have an easier time in Civ3 due to some of the changes (attackers can't use road networks, bombardment units can be captured etc.). I believe Firaxis realized how easy it was for human players to outsmart and demolish the AI. Since they realized it would be nearly impossible to give the AI the same strategic ability to attack, the next best option would be to enhance the defenders to make attacking harder for the human player.

So the AI probably still won't be able to attack us effectively, but we'll have a harder time attacking it. The overall effect is a game where war is just not as attractive an option. As a builder, I'm all in favor of that! The original Civ's were never intended to be war games, but many played them as such because it was so easy to beat the AI at war. Now I think the game will be much more about true civilization building, and not just civilization conquering.

[old dog]

I don't expect the AI to be able to strategize like an experienced human. However, it is possible for the AI to have a "bank" of plausible strategies, and it is possible for the AI to vary amongst these strategies so as not to be too predictable.

Most of the worst problems arise from a)no strategy at all, just sending individual units forth to attack the nearest city, and b)having a single, predictable strategy which can be fully exploited by the experienced player.

One would hope that, if the strategy worked for the AI a few turns back, the odds of that same strategy being repeated would rise. If the strategy failed, the odds of repeating would drop.

The claim that city governors "learn" from your actions gives me hope that the AI military strategy will rise to at least this level. Add in the "tilted playing field" of an AI advantage (at higher levels) in production and tech development, and you have yourself a very competitive game.

[korn469]

i want to point out that one of the major complaints with CtP:2 was the fact that the AI didn't pose a threat, if the AI cannot attack effectively, then unless like Steve Clark suggestion firaxis makes the AI better able to win in non-military ways the game will lack any challenge

even if you can't crack the AI's defenses it won't matter because every victory will be a spaceship victory

i know the AI can't be a military genius, but i expect that with superior forces that it should be a threat

if i could open up the AI and find out which military leader they based it on it had better not be Sargent Schultz from Hogan's Heroes!

however his famous quote does apply to the AI
I know nothing, nothing.