AI's Inexplicable War-Declarations, esp. in Mid-Game

[yimboli]

Originally posted by Common Sensei


Your asking for something beyond the Power graph correct?

Short answer: yes.

Long answer: I was pointing out an inequity... a bias for the AI... a nonlinearity in the game that makes it nearly impossible to play a peaceful civ. The AI has a much better guage on my military than I do on the AI's. Arguably, the AI needs all the help it can get, but this seems like more of a lazy workaround to avoid strengthening the AI. The game would be a *LOT* more enjoyable if these cop outs were avoided, which is a point I think everyone will agree on while Firaxis is pulling its hair out (AI programming is not easy). But if there was just one thing I could ask Firaxis to do, it'd be this: instead of telling the AI which civs are weakest, make them work for it, and make it more chancy. The AI will devote __% of resources to scouting (an overly simplistic view I realize), and that's it. And if it notices a weak civ, well hey score one for the AI. If it doesn't find a weak one, it can build more scouts or just sit tight. That's the way it works with me and all other human players.

Now the power graph does provide this information, which is an argument for another day.

Perhaps this is also tied into zabrak's point about the naturally competitive nature of the game's players?

[allybhoy]

but there are non cheating ways to guage the strength of opposing civs. Opening borders lets you wander through checking their strength. As does sharing a religion you own. Last night I was playing large continents with extra civs and was the only person on the contint to found a religion, Buddhism, and it spread like wildfie, with some help of course. All civs had my religion and I could see almost every city on the continent, and therefore helped myself to the Chinese land because I saw they only had a few spearmen and archers against my macemen

[Quillan]

The only warning I can notice is the switch to Vassalage/Theocracy. It's a fair bet when Alex or Monty adopts those two, he's gearing up for a war. If I look at the civs he can reach, and there isn't anyone who is both weaker and less liked than I am, I figure he's coming for me and prepare.

The only problems with this are they don't always drop those civics once the war is over, and they don't always go after the logical target. In my last game, I was sharing a continent with Ghandi, Alex, Mao and Huayna. India was weak and disliked by Alex (we shared a religion) so I figured when he adopted the warmongering civics he was probably going after Ghandi. It turns out he did in this case, but right after Ghandi he turned on me. I was ready for him, but if I hadn't prepared, he would have crippled me.

[Veritass]

I think the AI just has a better memory than me, and pays more attention to this. It is a computer, after all. If I wanted to go to the trouble to constantly scout through other lands and check religion-spread cities and sail by with my galleys and remember how many troops they had in various cities, I could have this information at the same level the AI does. I just don't take that much time to do so.

[yimboli]

Originally posted by allybhoy
but there are non cheating ways to guage the strength of opposing civs. Opening borders lets you wander through checking their strength. As does sharing a religion you own. Last night I was playing large continents with extra civs and was the only person on the contint to found a religion, Buddhism, and it spread like wildfie, with some help of course. All civs had my religion and I could see almost every city on the continent, and therefore helped myself to the Chinese land because I saw they only had a few spearmen and archers against my macemen

Yes Sir. There are non cheating ways. And while the AI simply _knows_ who is the weakest, I am forced to work for it via the non-cheating ways you mentioned. It's an inequity that gets programmers out of having to script AI that knows how to scout and guage strength.

And all this simply stems from what I believe to be the uselessness of the pacifism civ. Either you keep up military spending to avoid random wars, thereby killing your cash reserves... or you don't keep up military spending and the AI attacks because it *knows* you are weak (regardless of whether or not it has actually *scouted* you to observe this!).

Like I said before, I think pacifism would be a viable civ if the AI didn't automatically realize your weakness: I'm willing to accept a lower defensive posture in exchange for +100% GPP. But not if it means I will be inevitably targetted by AI that doesn't even have to observe me to realize my lackluster defenses.

Perhaps if someone can shed some light on the pacifism civ - provide some examples of when you would use it? some examples of when it has been successful? please take into account my comments on pacifism in these two posts so responses are more meaningful.

I'd appreciate any insight! Thanks!

[Generaldoktor]

Pacifism strikes me as a non-starter too, anytime during the game, but I play with "Raging Barbarians" a lot and tend to custom-group with aggressive AI's in my games. Some guys swear by it (Pacifism.) I recall having a conversation with Coeurdelion about it on another thread. You might ask him; he was on this thread earlier.

[zeace]

I like pacifism in some cases. If I've just eliminted/subjugated all the civs on my continent, and I have railroads around most of my coasts, it's actually quite simple to defend yourself with a small number of units. Assuming that the other continents are still sorting themselves out, there will be easier, local targets for everyone and they'll leave you alone.

I especially like this strategy if I've been neglecting my GPs. Sometimes I'll only have created a couple of GPs by this point, so I can quickly crank out another 4-6 in a few dozen turns. This can really help get you ready for the inevitable D-Day invasion you need to start planning.

[allybhoy]

Originally posted by yimboli


Yes Sir. There are non cheating ways. And while the AI simply _knows_ who is the weakest, I am forced to work for it via the non-cheating ways you mentioned. It's an inequity that gets programmers out of having to script AI that knows how to scout and guage strength.

And all this simply stems from what I believe to be the uselessness of the pacifism civ. Either you keep up military spending to avoid random wars, thereby killing your cash reserves... or you don't keep up military spending and the AI attacks because it *knows* you are weak (regardless of whether or not it has actually *scouted* you to observe this!).

Like I said before, I think pacifism would be a viable civ if the AI didn't automatically realize your weakness: I'm willing to accept a lower defensive posture in exchange for +100% GPP. But not if it means I will be inevitably targetted by AI that doesn't even have to observe me to realize my lackluster defenses.

Perhaps if someone can shed some light on the pacifism civ - provide some examples of when you would use it? some examples of when it has been successful? please take into account my comments on pacifism in these two posts so responses are more meaningful.

I'd appreciate any insight! Thanks!

OK mate fair enough. I do not believe that the AI cheats, just that it can calculate faster than a human can. I have no evidence of this, just a feeling that I get less wars if I have only my own religions and closed borders.

So if we could work on that assumption firstly. Close borders and garrison most units in cities with foreign religions and cities near borders that the enemy can see into, make him think that you are strong. Even if this assumptions are wrong, the advice below should still work

Overall best civics to run for great people are Mercantilism, Caste System and Pacifism. Meaning you are not penalised too much for closed borders by having mercantilism that gives you extra trade routes and Caste gives your specialists a bonus

Plan B, money money, money, found an early religion, spread it like wildfire, build a shrine and bathe in cash. Thats what happened to me last game and I am running pacifism, 80% science and have 170 gold extra per turn, and still have the biggest and most advanced military on my continent. Build the spiral minuret and colossus for more cash.

Play on archipeligo, destroy any near neighbours and give yourself space to run pacifism. Then your enemy can only attack by boat and you can defend yourself easily enough.

Play a spiritual civ, you will be attacked at some point, but you can change quickly to slavery and police state?? and whip and draft a military out to defend yourself next turn

Sorry I've rushed this a bit, and spelling may not be clever, but I have to rush to work now

[yimboli]

Interesting. I don't think we should have to work off any assumptions regarding the AI's omniscience. There should be someone around here that can answer this question - does the AI know when my cities are poorly defended even when it can't see them? Besides the power graph of course, because I can see that too... =)

[allybhoy]

hope that helped. If you really want to know if the AI knows and acts on what it shouldn't see, you should maybe ask the question on the modders forum. They are the dudes that go in and mess with the game code.

[Yosho]

Originally posted by johnmcd


In addition to that, it seems impossible to issue your own ulitmatums or have your previous conduct considered in terms of following through.

No, you can issue your own ultimatums. If you're in a position where you're ready and able to crush the AI, the AI usually will give you all of his money or something like that. Vice versa, it seems that the AI will usually only demand something of you if either he thinks he could crush you, or you've annoyed him, or both.

Not only that, but if you ask for certain things of your friends as "gifts" (like declaring war on someone, or ending trade with someone, or changing religion or civics), it apparenlty dosn't hurt your relations even if they say "no", and sometimes they will surprise you and say "yes". It's only when you demand money or a tech or a city that it seems that it hurts your diplmoatically, but even there you can intimidate someone much weaker then you into giving you cash, although it does hurt diplomatically; I do it fairly often.

[Chaos Theory]

It's quite annoying that the AI can declare war, then refuse to talk for dozens of turns, forcing a human player to build up substantial amounts of war weariness without recourse, short of obliterating the offending civ.

[Azuarc]

For that matter, it seems highly unrealistic to me that the first sign I see of enemy aggression is troops making landfall next to my city. In real life, such a thing would never happen. Are there any examples of a country being completely caught with their pants down in the face of a massive naval invasion?

Manila?
Pearl Harbor?

You can argue Pearl Harbor if you like, but the Spanish were sitting ducks when the U.S. showed up in the Spanish-American war, and it completely ended any war in the Pacific theater before it started.

In addition to that, it seems impossible to issue your own ulitmatums or have your previous conduct considered in terms of following through.

The worst part about threats is not being able to say "well, what about this instead?" Typically a Civ demands a particular tech I have because they think they can either get it from me for free or they can't get it at all. I promptly trade them something they have - not necessarily something I want, but I'm usually pretty free-dealing with most techs - and still get the diplo penalty even though ultimately I gave them what they wanted.

I reject almost all AI demands, since if a stronger opponent demands something, giving it to them just puts off my doom, and if a weaker opponent demands something, giving them something would just help them challenge me.

-1 vs +1 and some bonus to "Our trade relations have always been fair and forthright" is sometimes worth it.

It certainly slows you down less to build more units than it does to deal with the annoyance of an invasion, having to build all the improvements etc. again.

The computer doesn't seem to take into account any tech disparity, or the fact that you can kill huge numbers of inferior units with your human intelligence and small but technologically advanced army.

Sadly, sometimes having to take either action is enough of a deterrence from your gameplan, especially if you were going for cultural or space race. Usually the civ that does it to you is also not the one that is worried that he won't win merely because you do.

The tech thing is the part that slays me, though. I have 2 riflemen in each city. I get attacked by macemen and knights. My riflemen are dug in with city defender, and I've got a few quick cavalry I throw together to mop up the marauders...yeah, GG...

1. You are ahead of a powerful AI.
2. Based on unit count (the power graph) you are weak.
3. You get attacked.
Where is the surprise in this pattern?
The fact that this pattern irritates builders means the game designers are doing their job. It was not designed as a noncompetitive game.

Right, cause nukes are real dominant too. =p

It always irritates me because sometimes while I may be in no position to wage war, I am well within my rights to be able to guard against it, and the AI doesn't care. One civ comes in and raises hell so that a third party can win. It's almost like your opponents conspire against you. (Granted, sometimes they DO, but I mean beyond what is explicitly intended.)

In a recent game I played, Isabella was dead last in all things but military power, and nowhere near me either geographically or in terms of tech. But she had tons of power, somehow, because she had tons of piddly units. Knock, knock, it's Santa Claus! Never mind that Isabella had really no reason to want to attack me over the other civs -- I was farther away, more powerful, and probably less diplomatically offensive than the other Civs. (Nobody was using her one religion.) She did enough harm to throw a wrench in *my* game, but as far as I saw things, if I went on autopilot, Louis was gonna win the game anyway.

It's like playing Risk with someone who knows they're going out, and decides they're gonna turn in their cards just to kill off the rivals of the leading dude so the game ends quicker and they can do something else.

It's quite annoying that the AI can declare war, then refuse to talk for dozens of turns, forcing a human player to build up substantial amounts of war weariness without recourse, short of obliterating the offending civ.

I had this happen in another game the other day. I got bored of my island position after conquering my half of the globe, so I decided to pick on someone on the other side. Problem is, they're all religiously-aligned. So I picked a fight with Japan, and 20 turns later, I get side-swiped by Rome. Japan finally calls it quits, which I agree to since I'm dealing with the more powerful Rome now. Just as Rome is about to leave things as the status quo, England goes to war with me. For a period of about 150 turns, I was not at peace at all. That did REAL wonders for my economy.

[Generaldoktor]

These AI are not going to act like people. They act in a way suggestive, rather than imitative, at least for another ten years before we have real AI. (And it will probably go in the then-current President Bush's warplanes before they put it in games . )

Point being, if your gameplan is thrown off that bad by a technically inferior foe launching a strategy-less parry, there is something wrong with your gameplan. The AI has to have some proactive responses to deal with the dominant human. Deal with it. Most people in these threads complain the AI is too wimpy; many fleeing to the new MP community as a result.

Oh, and I used to play "Risk" that way you describe too.

[Blake]

Originally posted by yimboli
Interesting. I don't think we should have to work off any assumptions regarding the AI's omniscience. There should be someone around here that can answer this question - does the AI know when my cities are poorly defended even when it can't see them? Besides the power graph of course, because I can see that too... =)

The AI does "maphack", it at least sometimes targets weakly defended cities deep inside borders, reloading and buffing up the garrison causes the unit to switch targets.