Why does the AI commit suicide?

[Moraelin]

Originally posted by Redstar
"Huh? I've never noticed anything like that. Most of the attacks I've seen were on some city that's near the center of my empire "

not so...the AI will go for weakly defended cities.

you can prove this to yourself simply this way...

Redstar, you may notice that I was saying the same thing myself. Everyone WILL go for a weakly defended city. They will NOT, however, factor in such thing as whether it's easy to reinforce or not. They'll just as gladly go for one of the core cities, that's connected with roads to 20 other cities and to a couple dozen cavalry units, as for one that's unconnected and all the way beyond the mountains and jungles. (Or even worse, which is across water, and would take 10 turns to move troops there by galleon.)

In all fairness, I can't blame it for not factoring in the ease of reinforcing it. It's a pretty fuzzy thing, and you could argue that if it moves swiftly, I may not have the time to reinforce it anyway. Just, well, pointing out that it doesn't.

[gamadict]

Although the AI is almost certainly not doing this, I wonder...

If you want the AI to play somewhat realistically as a nation, it would rarely make sense for them to attack another nation that is overwhelming superior. However, if the AI is playing to try and win a game of Civ3, then maybe, if they are so hopelessly behind, it could be a last-ditch effort to try and knock of a competitor, slow down the leader, get some cities/territory, etc. Now, there are probably better ways to do that then just a head-on suicide assualt. Try and get a diplomatic victory, for instance, or try and get the other civs to gang up on the leaders (Which, if successful, would lead the human civ to complain about all the AI's just being out for him . But if you're that far beneath another civ, especially a human civ, they're probably not going to let you get the UN. Maybe the AI sees a suicide attack as it's best chance, however unlikely, to eventually be able to win...Probably not, though...

[LRotan]

Interesting view, Gamadict. It doesn't explain why in my last game(a good while back) the Iriquois with all of one city and a military consisting of 2 MWs and a spearman declared war simultaneously on two neighboring superpowers. They did ask for my help, but even that doesn't show much sense. There was no way I could help them out as I would have had to travel through every other Civ on the map to get at their enemies. I was the only one they asked for assistance and two turns later they were obliterated.

Still, it puts a new spin on how one might view such behavior. I'm pretty sure I wouldn't have declared war in their shoes, though.

LR

[C Chulainn]

It's not as bad as in Civ2 where I can remember getting a "Sneak Attack by ---" message even when the enemy were reduced to one city and well behind in technology.
Another factor in AI head-against-wall bashing seems to be in negotiation - whether the AI or you offer terms. In the game I'm playing at the moment, as Persia, I eventually crushed the Babylonians - I captured or razed over half their cities including Babylon but they constantly refused to surrender, boasting that they'd crush my armies, when I offered terms. When, however, they finally sued for peace themselves they accepted my counter-offer, which involved giving me all their gold and technology, immediately.

[Moraelin]

Well, strictly speaking, it is "realistic" for a smaller nation to attack a larger one. I seem to remember Germany vs Russia, or Japan vs the USA at Pearl Harbor.

However, in both cases, the plan was a bit more elaborate than that. E.g., Germany counted on the technical inferiority of the Soviet army, air supremacy, and on superior tactics. They also counted on the idea that Stalin would keep his siberian troops in the east, waiting for a Japanese attack.

Now I realize that all that would be impossible to put in a game, or at least impossible without ending with an unwieldy behemoth.

Maybe it would make more sense, though, if a weaker country would seek an alliance BEFORE starting a war, instead of starting the war alone and MAYBE later trying to get some help. Sort of like Germany counted on Soviet help in Poland, on Italian help with France (which didn't help much, but hey), and on Japanese help to keep some of the Soviet troops tied when invading Russia. Basically it would make more sense if two AI countries started the war by making an alliance against me, instead of alone.

It could also help if the game had non-aggression pacts. Or even better, downright buying someone's promise that they won't attack you in the next 20 turns. Sort of "OK, I want to invade Austria. How much does it cost me to secure your non-interfering?" That could be a way to counter MPP's, at least for a while.

[ancient]

Originally posted by Moraelin
They also counted on the idea that Stalin would keep his siberian troops in the east, waiting for a Japanese attack.

boy that was a mistake, some inteligence could have easily figured out th ey knew japan wasnt planning on attacking them..

[GeorgeWang]

I don't think the A.I. plays to win... I've played awhile and the general A.I. strategy seems to break into two parts.

1. Gangbang the Wimp

2. Screw the Leader

Now, in every game, who's usually the leader? The human player. By the same token, the Wimp (Weakest Military Unit-Wise) is again usually the human leader who can actually realize a Motorized Infantry is 20 times more effective than 20 warriors, not the other way around like the A.I. sometimes believes.

So, no, by playing like a human, you will have the A.I. treat you differently from every other civilization.

[narmox]

Originally posted by Moraelin
Best of luck to you, then. But I'm pretty sure that there's NOTHING you can do to remove an UU until that civ had its golden age. So if the Greeks keep missing on building Commercial or Scientiffic world wonders, they'll still have hoplites in their list all the way into the 20'th century. If the Romans didn't have a war yet, or don't have the right wonders, they'll still have their Legionaries in the list, no matter what.

What you say!! I've had my golden age early enough (as Romans, waging a war with legions), and I still can build legionaries (IIRC, haven't played for more than a week). I think a civ can always build its UU once they get it, even when it's hopelessly out of date.

[Moraelin]

Actually, most likely what you have there is a unit that doesn't upgrade. (Or you don't have the resources for any upgrade.) Most UU's were shipped with their upgrade broken like that, and some stayed like that even after the patch.

So if a unit doesn't have an upgrade it stays in your list.

If it has an upgrade, but you don't have the resources for that, it stays in your list.

If it's an UU and you didn't have your golden age yet, it stays in your list no matter what.

[TrainWreck20]

I think the AI 'suicide' is partially in the programming to keep the game from getting boring. The most interesting games I have had have been the ones in which an opponent declared war prior to rail. Although I was not too worried about the long term military threat, at least one other AI would join in from the opposite direction. This at least posed some interesting logisitics and entertaining warfare.

I have found that island games are too easy for this reason and large continents where you are close contact with as many civs as possible much more entertaining.

I think the AI is set up so that the attack from the weaker civ will cause you to move your forces enough to make you look like a good target for a second (or third) civ. Some of the early wars can get involved enough to stunt your development and give a different civ an advantage. To be fair, I have seen the AIs do this to each other, not just the human.

Regarding the AI knowing the location of every unit, that is annoying, but since I consider ir cheating (not AI coding), I don't feel bad for abusing it . You can make a fleet (Galleys, anyway) full of units sail back and forth endlessly while you prepare to destroy it by using a couple of your own galleys 'open' and 'close' a passage along a far away coast. Somewhat monotonous, but can give you time to build your settlers and take the land yourself ....

[Syzyon]

For me, the fact that the AI can see inside all your cities becomes a really big problem. Not because it becomes harder, the game actually becomes easier.
During a war my border cities are naturally more guarded than cities inside my empire. But apparently the AI feels that instead of attacking cities on the border which it can attack and take on the same turn, it sends for example 15 cavalry past my border city towards some distant city it will never reach. Within two turns ZOC and fast units will have slaughtered them since they can't use roads and I can send 2000 tanks and the same number of artillery across 2000 kilometres in an instant "due the unlimited efficiency of railroad...(?)" So the AI seldom actually capture any of my cities...
Takes away some of the challenge methinks.

[FNBrown]

I've seen the unreasonable "suicide attack" a few times, too. Typically, it's against a small, lightly defended and remote city.

I think the problem is that the AI doesn't take into account the "righteous anger and thirst for vengeance" factor that seems to motivate me to crush his puny little nation in retaliation for his offense. The initial attack is almost always a rousing success for the AI civ.. and sometimes, it can take years of game time to retaliate. France hit one of my remote settlements in a "huge" map game, and it took me forty turns to assemble troops and transportation to get them overseas.

Of course, by the time I got there, I had developed air units, and could have just airlifted my assets...

Nonetheless, I had my revenge. I doubt the computer really figured on me having the patience and resolve to mount such a retaliation.

[Grrr]

Originally posted by Dan Baker
What in the world does the computer factor in making peace? I've seen the most stuborn peace negotions when a country was on the verge of desctruction. Clearly, every thing in the game has a monetary 'price' tag and the computer computes how valuable something is to you. It also knows how valuable something is to it. Apparantly, peace to avoid destruction isn't that high on the list.

Talking of peace agreements, next time, they winge for peace, demand 999999999 gold per turn, I bet they'll take it, along with anything else you demand.

See the following thread
http://apolyton.net/forums/showthrea...threadid=39124

I agree with the AI that it is often worthwhile to attack small undefendable cities, but they should RAZE them!
I also don't like the AI building cities next to my colonies and watching them disappear off the face of the earth.

[Paul_1868]

I notice smaller civs that do the suicide wars are always attacking one of the game leaders. Maybe smaller civilizations attack much larger civilization toward the end because it’s the civs last change to obtain victory, even if it is a far shot.

[opaque]

By the same token, the Wimp (Weakest Military Unit-Wise) is again usually the human leader who can actually realize a Motorized Infantry is 20 times more effective than 20 warriors, not the other way around like the A.I. sometimes believes.

Probably one of the causes of the "spearmen beats tank" combat problems people are talking about. They could well have done that so there was less chance of the AI getting obliterated when it attempts to attack more advanced units with ancient ones, thinking the quantity of their units will lead to a victory.

[Barnacle Bill]

Originally posted by Paul_1868
I notice smaller civs that do the suicide wars are always attacking one of the game leaders. Maybe smaller civilizations attack much larger civilization toward the end because it’s the civs last change to obtain victory, even if it is a far shot.

Actually, this is a rather common symptom in almost all 4X games - the AI is programmed to thing like a gamer instead of thinking like the leader of a country. If a gamer can't immediately pursue the victory conditions, he tries to prevent his opponent from reaching them. If one player looks about to win, all the other players tend to gang up on him. The AI in single-player 4X games is almost always programmed to gang up on the human player if he/she is the leader. In real life, of course, governments don't do that because the "game" never ends as long as you continue to survive. Instead, they play to maximize their own board position within the constraints imposed by domestic politics (for example, it is hard to ally with traditional enemies or regimes your people think to be evil, or against countries for which your people have strong sympathies). This realpolitik results in countries that are weak allying themselves with the strong, not attacking them suicidally because "they are winning". It may, however, allow peaceful bad relations (dislike, rhetorical opposition, refusal to trade) with the strong powers they don't like because of ideological/culteral differences and domestic political concerns, especially if either the strong power in question has no reputation for launching wars of conquest or the weak nation has a powerful protector (or is far enough away from the strong that the strong has difficulty projecting sufficient power way over there). The game which, IMO, comes closest to doing this sort of thing realistically is EU2 - not perfect, but far ahead of any other I can think of.

[Grrr]

You are right about the AI's "LAST DITCH EFFORT" to win, and it would probably be better if you had all the time in the world, like SMAC, err sorry, planet. Well if someone was to make a mod that from say 1950, turns are a month each, this will reduce the scientific burden of the 4turn limit too.

By the way, has anyone seen the AI raze a city?!?!

[zorbop]

the ai has to declare war if they ow you money every turn but they don't have it.

and the ai has razed a couple of my cities

[Grrr]

Originally posted by zorbop
the ai has to declare war if they ow you money every turn but they don't have it.

That sucks!, there should be an option to renogtiate, but anyway, aren't improvements sold to make do?!