Is it just me? The comp-controlled aggressive opp can't play

[Freddz]

Well, it seems the opinions on why the aggressive A.I. fall behind differ quite a lot. Some think the aggressive ones attack randomly and too often, others, like me, feel that the aggr. A.I. actually doesn't put enough pressure on you and very very rarely has any ally when they become involved in war, which makes one fear them less.

[Shaka II]

Originally posted by Freddz
Well, it seems the opinions on why the aggressive A.I. fall behind differ quite a lot. Some think the aggressive ones attack randomly and too often, others, like me, feel that the aggr. A.I. actually doesn't put enough pressure on you and very very rarely has any ally when they become involved in war, which makes one fear them less.

Handel has been playing with aggressive AI on, so the accentuated, eratic behavior may indicate part of the problem of why the aggressive attribute is not implemented effectively enough.

FootieMad is also correct in that pangea offers the aggressive AI a better map to excercise that attribute. Pangea is the map I play for that reason and offers that "Land war on Asia". But even on pangea, the aggressive AIs fall behind, or are usually not aggressive in the right way to achieve dominance.

Aggression has to be smart aggression, not eratic aggression. Making sure that a war makes sense, not just attacking because it's their nature. In Civ4, it seems that the peace mongers are the bigger threat.

I have seen Napoleon do quite well in a game, but usually Louis XIV offers a stronger challenge. Also a leader like Caesar with expansive/organized will creep up and achieve great strength in the end game, one reason why I like to play as Romans, along with getting Praetorians.

But of the games so far, Washington, seems like the strongest player, mainly from a research perspective. I still haven't played against them all. Elizabeth has strong science, but is usually not that strong a military power.

[johnmcd]

I feel the aggressive AIs should go through phases of unit building and attacking and other phases of digging in and building up. I guess they used to get a kick out of stealing techs on taking a city which compensated for low research budgets.

I do agree the aggressive AIs do seem to perform very poorly on average though, but once in a while very well. I suspect it depends on how well their first war went really. They always do better than the isolationist Japanese though, that has to be the single most useless AI behaviour.

[Freddz]

One must hope that Firaxis really balances up the different A.I. opponents in future patches, or at least in an expansion.
Will this mean something that many of us feel this way, or will it be of no concern to Firaxis? If they fix it, when do you think we can hope for a fix and what kind of fix?

[Bluefusion]

I disagree with people who say pillaging is worthless. Pillaging can really screw over your production and food when you're trying to pump out units to defend. Plus, in the early game when you do not have many workers, a nasty pillage could send you a dozen turns behind in land improvements.

And how many times have the AI pillaged a vital resource you've needed to produce soldiers? For instance, they always tend to go for the iron or copper when you need them for axemen.

Edit: I won a game by constantly airstriking a guy's aluminum supply late in the game. (of course, using nukes might help explain the victory as well :-) )

However, I think the problem with AIs is they tend to ONLY pillage and fail to muster powerful forces to attack cities.

[AuraSeer]

Originally posted by Bluefusion
And how many times have the AI pillaged a vital resource you've needed to produce soldiers? For instance, they always tend to go for the iron or copper when you need them for axemen.

I've seen it happen once or twice, but only because a mine was on the first square inside my border. The AI seems to pillage the closest improvements it encounters, even if that means destroying the nearby cottage instead of moving one more square to destroy an important strategic resource. (This happened early in my most recent game. When the AI horse archer stopped to destroy a cottage, that gave me just enough time to move spearmen in and defend my only metal mine, on the next square over.)

A related issue is that the AI will sometimes stop to pillage every improvement it steps on, even if it's on its way somewhere. I've seen a tough AI stack advance toward an under-defended border city, only to be delayed for three turns by stopping to pillage the farms it happened to cross. That stupidity allowed me time to move in defenders from elsewhere, and fight off the stack when it finally attacked the city.

[Shaka II]

Originally posted by Bluefusion
I disagree with people who say pillaging is worthless. Pillaging can really screw over your production and food when you're trying to pump out units to defend. Plus, in the early game when you do not have many workers, a nasty pillage could send you a dozen turns behind in land improvements.

And how many times have the AI pillaged a vital resource you've needed to produce soldiers? For instance, they always tend to go for the iron or copper when you need them for axemen.

Edit: I won a game by constantly airstriking a guy's aluminum supply late in the game. (of course, using nukes might help explain the victory as well :-) )

However, I think the problem with AIs is they tend to ONLY pillage and fail to muster powerful forces to attack cities.

The AI seems to be pretty good at pillaging in my opinion. It certainly is treated with high priority. Like you say, they tend to go for the strategic resources like oil, and aluminum, uranium. Usually I find out that they strafed my oil fields, when I get a message saying "you can no longer continue to build modern armor, what would you like to build now?"

Mech Infantry and artillery.

[Pyrodrew]

Aggresive AIs should, imho, be a lot more opportunistic, striking whoever is weak.

That's fine in the beginning, but as the game progresses the AIs should also make sure a player doesn't get a huge lead. In Civ3, once you became top dog the AIs usually feared you too much and fought their insignificant wars while you coasted to victory. The AI should play to win, not play for 3rd place or 4th place or "good enough" place. This sometimes means making allies to fight the top dog or for non-aggressive AIs economic tactics.

[pauli]

...i'd rather not go back to civ2 style "everybody vs the guy in the lead (if he's human)," thanks.

[marvinkosh]

There are any number of human strategies which would make the AI much more competent in war. The problem is encoding them in such a way that the AI is able to:

1) Produce the right units to take the objective. This means you need recent intelligence on that location and cities nearby, so you know what you will be facing.

2) Plan a route to the objective. This may include sending several batches of troops overseas, and in this case it is important to secure the coastline before commencing the attack.

3) Remember that units which are assigned to an objective are not to be used for anything else on the way, unless the survival of the unit team is a stake.

4) Recognise when the operation has succeeded or failed, and have a contingency plan for each outcome. For example, if you fail to take that first coastal city, you want to get any survivors to safety using your transports, and prepare a new assault if you are able to. If on the other hand you succeed, you have perhaps dealt your opponent enough damage that they will cave in to your demands. On the off-chance that you have not, you should mobilise enough units to defend the newly-taken city against a counter-attack, and also make sure you are ready for a counter-offensive elsewhere.

[Pyrodrew]

Originally posted by pauli
...i'd rather not go back to civ2 style "everybody vs the guy in the lead (if he's human)," thanks.

It's not about everyone vs. the guy in the lead if he's human. It's about playing to win regardless if the lead is human or not. There are victory conditions and the AI should: 1>Aim to achieve these victory conditions so it's wins; 2>Aim to stop others from obtaining these victory conditions so it doesn't lose. By not having the AI defend these victory conditions from opponents it's like a football team with everyone on offense and no defense.

And "everybody kill themselves while the guy (human or not) in lead always wins" is hardly a better solution. It's a strategy game where only 1 player can win. It's not the Sims.

[ChaotikVisions]

Really the biggest flaw in the AI currently is it doesn't know how to use siege engines. It doesn't seem all that different from the Civ3 AI. Often it will trickle units in and either bash themselves to deaths in cities or split up trying to pillage. Either tactic results in their death. I've even seen the AI do this when it is the one to declare war. When it does actually get siege units it often doesn't wait until defense is completely bombed down.

So assuming the player has a decent garrison its still easy for them to beat an attacking army. I've also seen the AI place too much importance on its capitol. It will leave 10-15 units in it while its other cities are slowly captured by me. Chances are it couldn't actually destroy my attacking force completely but it could atleast inflict some damage. By the time I get around to clearing out their capitol all those units are useless because its just a matter of time until I wear them down.

I have occasionally seen the AI launch good attacks with lots of units accompanied by siege units. The only problem is its a rare occurence that seems to take the AI a long while to build up to. Unfortunately these problems with the AIs tactics mean it will almost never conquer neighbors fully or quickly so an aggressive human player can easily out-conquer and end up on top power wise.

[Freddz]

First of all, the strategy discussions here for A.I. performance in war have developed towards two goals as I see it:

A/ Improving the warmongering so aggressive A.I.s become as fearsome as peaceful ones.
B) Improving combat pressure. As I said a number of times, the Civ-series is largely about pressure and war-pressure aren't nearly as great in Civ 4 as in the previous games in the series and this must definately be improved.

I'd like to see this thread as a thread to help Firaxis find easy solutions to improve warmongering in human vs computer games, without unbalancing human versus human games. Also as a thread to promote the importance of feeling pressure in war. Which is lacking many times now.

Can we find easy, really helpful suggestions or insights to improve warring for the A.I. as I am sceptic that we can get Firaxis to do large-scale improvements on the strategically thinking engine.

Our chance of actually getting something done would be to present clear insights here together with potential fixes in the simplest possible manner. For example, I believe an A.I. cheat is easier to get Firaxis to buy into when it comes to aggressive A.I. than that of spending weeks or months of rethinking A.I. strategies and then programming the fixes sensibly without programming flaws added as well.
Some of that "The A.I. shouldn't cheat" gotta go I'm afraid. But Firaxis should also be ready to reprogram some changes to A.I. behaiviour as well. At least in an expansion. Without that, I won't buy an expansion.


Fix examples that are fairly simple to execute:

1) A somewhat reduced production cost at building military units for aggressive opponents would improve the Aggressive A.I. loser problem, as some observations points at that if the A.I. is given enough time he can actually launch some well done attacks now and then. That with a small bonus to economy to take care of those extra units.
The A.I. has some set requirements for it to launch an attack to a city , and not only stop at pillaging. Those requirements are easier achieved when having some more units. Is this a possible solution to anything? One good thing with this idea is that the aggressive A.I. actually becomes a better builder with this(less time spent to build units is more time spent building other things.)

2) Another fix could be to lessen the defence gained from culture. That way the A.I. would not be in such desperate need of several siege which it appearantly doesn't use to its best advantage now.

Any other suggestions to simple fixes added to those sharp analysis of A.I. behaiviour presented in this thread? Of course, we still need to find out the things that are causing this mess in A.I. tactics.
Okay, some responses to thoughts here:

Originally posted by Pauli
...i'd rather not go back to civ2 style "everybody vs the guy in the lead (if he's human)," thanks.

Well, something has to be done in the end game as it is now. I agree with the posters who feel that it is a problem the A.I. doesn't play to win. It would also make the end game more interesting if you don't lead all the way through it without hardly anyone declering war to you, which is the way it stands now. That's more terrible to me. Last game I played and led towards the end I longed for an A.I. to war on me for me to feel excitement. It never came, and the end game excitement never came either.

Originally posted by Shaka II
FootieMad is also correct in that pangea offers the aggressive AI a better map to excercise that attribute. Pangea is the map I play for that reason and offers that "Land war on Asia". But even on pangea, the aggressive AIs fall behind, or are usually not aggressive in the right way to achieve dominance.

Playing the Earth maps some, I have noticed that Genghis Khan are always last or in the bottom. The starting location may not be the best of the Earth map starting locations(I don't know), but the opportunity for land war should be great yet the mongols invariably fails. The civs I attack always drop towards the bottom as well, which really isn't a great grade for A.I. performance in war. So A.I war behaiviour has to improve, not just with cheats. But which improvements can realistically be done? The half-rookie human player is at the moment tons better than the A.I. at conquering cities.

Originally posted by ChaotikVisions
Really the biggest flaw in the AI currently is it doesn't know how to use siege engines.

I agree that this is a major flaw that needs improvement. Another great flaw is that I am rarely at war with several opponents, as most neighbours are of the peaceful, non-opportunistic sort.

It doesn't seem all that different from the Civ3 AI. Often it will trickle units in and either bash themselves to deaths in cities or split up trying to pillage. Either tactic results in their death. I've even seen the AI do this when it is the one to declare war. When it does actually get siege units it often doesn't wait until defense is completely bombed down.

True. I rarely ever see them use siege weapons to bombard. The attack of the A.I. must improve, no doubt about it.

I've also seen the AI place too much importance on its capitol. It will leave 10-15 units in it while its other cities are slowly captured by me.

I also see this all the time as well. This is a thing that needs careful balancing though, as moving too many units from the capital would make it too easy for the human player to deal a deadly stroke to the A.I that it can't come back from.

All in all, I become more and more certain that the (aggressive?) A.I. need a small boast to unit production to give a challange. That together with several of the ideas presented in this thread about diplomatic aggression and improving A.I. behaiviour with siege weapons and pillaging could do something good I think.

But I'm unclear whether pillaging behaiviour really needs improvement though, or if it is actually a fact that when the A.I. hasn't got a half-decent chance of taking a city, it pillages to do something good. Maybe this is MORE of "a lack of units"-problem. So with more units, the A.I. would more often see that it has a chance of conquering cities. Let's face it, even a well-programmed A.I. needs more units than a human player to be a threat.
Raising the difficulty level won't solve these problems since the A.I. may do very well in research and other areas.

[Bluefusion]

Don't forget, AIs cannot always go on a pillage frenzy. They're shooting themselves in the foot if they intend to capture a city whose land they pillaged just turns ago.

And...I actually see this very often.

[Freddz]

Originally posted by Bluefusion
Don't forget, AIs cannot always go on a pillage frenzy. They're shooting themselves in the foot if they intend to capture a city whose land they pillaged just turns ago.

And...I actually see this very often.

The first goal should be to capture, not to pillage. Are you saying it is the other way around at times?
Or are they waiting for re-inforcements?