Sirian
Designer, Mohawk Games
Kylearan said:Now this more powerful civ *should* be able to conquer an equally intelligent, but less powerful civ. Which I haven't seen happen so far.
Should it? Are you sure?
There are only two ways to balance offense and defense. You advantage one, or you advantage the other. The scales tip very quickly, so any effort to "balance" these inevitably results in advantage to attacker.
Aeson covered how Civ3 AI fails to defend its cities, but the problem extends beyond even that. In GalCiv, the AI doesn't even leave a token defense. It sends ALL of its fleet in a beeline at the enemy, leaving its homeworlds undefended. The human player can bribe the AI to attack a third party, with his own attack force already in position. The sucker AI sends its ships offworld and the human can hit every world owned by that AI while it has no defense at all, and wipe out a whole civ with little effort in an "Alpha Strike".
The only way to stop this is to send less units out on offense, and keep more at home on defense.
The Civ3 AI is PATHETIC in this regard. It does leave a token defense, but that defense is all static. It will produce new units and suicide them immediately.
I urged Soren to change the AI. Many changes were made.
1. The AI now has three "war states" instead of just one.
-a. Limited War - intent is to damage a rival to slow them down or butter them up for concessions.
-b. Dogpile War - AI asks another civ to join them in a 2v1 vs their enemy.
-c. Total War - AI is intent on conquest and will try hard to take cities.
2. Limited War means the AI is trying to do harm without taking heavy losses, so it will focus mainly on pillaging. To stop pillagers, the defender must go on the attack. It will go for cities if there is a border city very weakly defended, but chiefly it will try to disconnect resources, rip up cottages, etc.
3. Dogpile wars are very opportunistic. The dogpiler will continue its effort if it is making gains. If it is taking a lot of losses and not gaining much, it will tend to stop.
4. Total War is in there, still, but even it is not as aggressive as Civ3 or GalCiv AIs. Even in a Total War, the AIs will keep a stronger defense at home than they do in Civ3 (making them less vulnerable to being wiped out by a simple dogpile.)
5. All the AIs keep two kinds of defense: static city defenses (mostly City Garrison type units, but also a few "counter" units like axe, spear, or machine gun) and mobile counterattack forces. The static defenses do not move, period, but the mobile defenses will try to respond intelligently to invaders. That is, if the invasion is too strong to counter in the field, then the mobile defenses will hide in the cities and help to protect the cities. If the mobile defense can hit isolated pillagers, or has enough force in position to have a solid chance of actually wiping out invading stacks, they will attack, as a group, in force. ... There are still holes in this process, especially early game or for a civ who lacks resources or has had them pillaged, but by and large the AI tries to use strategy now, in choosing if and when to fight. (If you always bring overwhelming force to a fight, you won't see the mobile defense mobilizing very much. They WILL stay in the city, usually, if that is the best option available to them at the time -- and it's something I do, too, when I lack the force to deal with a large invasion. Protecting cities is more urgent than protecting the countryside, in nearly all cases.)
6. The AI on attack has two approaches: Scatter and SoD. The SoD method will use stacks of six or eight or ten or twelve. It will land several galleons of troops in one spot. It will GO FOR CITIES. The scatter approach will try to surround cities, and will pillage a lot more, but will go for a city if it gets enough forces in position. The AI will also use siege units on attack. This has room for more improvement, but they can put a good threat on a single city even if it is very stiffly defended, and they can attack at more than one place at a time with multiple SoDs.
The Civ4 AI, at the big picture level, is not behaving as I hoped. There are two basic reasons for this.
A. There is no variation (or not enough variation) in the ratio of AI units committed to its various military roles. There is offense, mobile defense, and static defense. Currently, the AIs are more or less following a One Size Fits All set of unit ratios. This is set at a level where the AIs keep strong defenses and rarely take any major risks. Thus one civ's offense is never a match for another civ's total defense. This is chiefly due to the fact that the AI ON DEFENSE was made significantly more capable!
I have always believed that the ratios need to be more flexible. Sometimes the AI should commit more of its total force composition to offense. (A human certainly will do that, but a human can also more readily tell when he is or isn't vulnerable to a dogpile/backstab/etc.) If an AI took more risks, now and then, they'd more often be able to overwhelm a target, and if they fail, they'd be highly vulnerable to a third party. (Getting this to calibrate and balance correctly won't be easy, but I hope I get the chance to try.)
B. With AI competent now (finally) on defense, they also need to be a bit more competent on offense. Not tactically (some room for improvement is there, of course) but strategically.
DIPLOMACY IS THE OVERARCHING ISSUE OF SINGLE PLAYER.
The current diplomacy model works out pretty well when there are an equal number of Peaceful Builder and Raging Warmonger AIs in the game. If the situation works out with three civs on one side and three on the other, and regular wars between one or more civs from each side, then the game works out pretty well. However, only a minority of AI personalities are really warmongers, so if there is an imbalance, usually the warmongers are too few and cannot form dogpiles, and fall behind for having fewer trade partners. I have even been in games with NO warmongers at all. Peaceniks don't tend to start wars, but prefer to build.
The diplomatic balance is not what it could be. And the AIs lack the ability to coordinate more effectively -- with the human or with one another.
Combine these two issues (lack of variance in AI unit mission ratios, plus lack of cohesion and balance in the "diplomatic blocs") and the result is that (on average) AI wars vs AIs are both too few and too ineffectual. This CAN be remedied but it is not a small task -- and some eggs will have to be broken to make the omelette.
These issues were understood at release time. However, they are "next generation" issues. These are emerging from Civ4's AI. The problems of the Civ3 AI have largely been addressed. Every major weakness has been shored up, and considerably so in some cases, such as AI competence on defense. The AI is more competent on attack as well, but its defense competence shot up by orders of magnitude, leaving the AI-on-AI wars in a new and strange place.
You can see how challenging it is to work on AI, though. Even when you fix ten or twenty things, there is another layer beneath those that will come to the surface. Beneath that is another layer, and another. We are making huge progress (in my opinion) but there is no doubt that more can be done.
Players grew fat and lazy off the Civ3 AI. In my opinion. They got used to being able to pit one AI vs another at will, for pennies. They got used to AIs who would gas themselves in wars with other AIs, leaving the player able to cruise in like the Big Bad Wolf, huff and puff and blow their house down. Well, no more. The AIs are competent on defense now, and that is as it should be. Competent does not mean unbeatable, though.
Civ4 warmongering is more than possible. The gold gained from repeated taking of cities can fund a lot of research, a lot of cities. Rivals wounded by your efforts are, as Aeson puts it, knocked out of the game, more or less. Warmongering will no longer snowball, though, and if that is what you want from it, then go back to Civ2 or Civ3 -- or mod Civ4 to make war The Only Right Choice once again, by making it (once again) in to a cakewalk.
War was the fortunate son of old. Seems that some people liked it that way, but when you put people like Aeson and myself, who grew bored with the Civ3 AI being too easy to beat, in to advisory roles, the AI -WILL- become harder to beat. (That was kind of the point, you know?)
There was no shortage of warmongers on the testing team. As Aeson also eloquently described, the problems that folks are complaining about in this thread are less about the war system and more about the AI. Well, I've described the two main reasons why the AI is falling down on the job. I've also described many of the areas where it is getting the job done.
Will Civ4 be even stronger and more fun after the first expansion? I hope so. Meanwhile, some situations (maps, civs involved, settings, etc) are more well balanced than others. If the powder keg of the AIs can be sparked off by getting them started on some warring vs one another, then the diplomatic system will tend to foster additional wars. Find ways to ignite the powder yourself, and you can get in to situations where the game is playing out the way it should be. If I'm in position to improve the ratios and redress the weak spots, I'll do it. That's all I can say for the moment.
Civ long and prosper.
- Sirian