AI Aggression levels (please discuss here).

[kettyo]

phungus420,

AI doesn't take into account which leader you're playing you're just a human player for them.

 

[Wodan]

A thread here made me think that checking agressive AI should make the AI percieve the human player as a warmonger. If I click nothing, the AI should treat me like our diplo shows, and how the AI would treat my chosen civ leader. However if I click agressive AI, the AI should see me and go "****" I started next to Monty. It shouldn't get the basic diplo penalty as is done now, but it should be aware I can and do backstab, like how it is aware of this for Monty.

Basically in terms of the intended "friendly" MP game, if I click agressive AI, I think it's reasonable for the AI to assume that I will play any leader in a similar diplomatic fasion as Alex or Monty. If this is not picked, I think it would be wrong for the AI to just go "Human player = not trustworthy"

Good analogy and point.

Wodan

edit: Kettyo, I think he was using Monty as an example... saying that he thinks the AI should react the same way WE do when we start next to, say, Monty (or Alex or Tokugawa etc)

 

[jkp1187]

In a recent game, I launched an amphibious attack on Brennus in the Industrial age. (Well, I was, he was still in Renaissance...) Anyway, I seized a city on Brennus' west coast with a force of 3 grens, 2 knights, and a trebuchet, then scouted a bit with a caravel and noticed that one city on the other side of the continent was empty save a single galley.

It looked like the AI sent all its defensive units west (which seemed like this is what it did, because he did launch a couple of attacks and managed to pick off one of my knights,) hoping to crush the beachhead. At the same time, it started cranking out catapults from the cities at the far end of the beachhead. This effort was for naught because I was able to overwhelm the enemy thanks to superior firepower and use of naval transport, but in general this was not a bad strategy for the AI to follow. If he'd been up in tech, this would've been more effective. (And FWIW, he did not have a huge army -- Brennus was stagnating because he'd been isolated from all the other civs for most of the game.)

Anyway, good job on teaching the AI how to fight.

 

[Uncle_Joe]

It looked like the AI sent all its defensive units west (which seemed like this is what it did, because he did launch a couple of attacks and managed to pick off one of my knights,) hoping to crush the beachhead. At the same time, it started cranking out catapults from the cities at the far end of the beachhead. This effort was for naught because I was able to overwhelm the enemy thanks to superior firepower and use of naval transport, but in general this was not a bad strategy for the AI to follow. If he'd been up in tech, this would've been more effective. (And FWIW, he did not have a huge army -- Brennus was stagnating because he'd been isolated from all the other civs for most of the game.)

Yes, the AI tends to 'panic build' siege weapons rather than defensive weapons when a city is threatened. I think it should build the best city defender it can rather than a catapult or cannon. Chances are, that unit wont even get a chance to attack if the build is in response to a stack moving adjacent. So, it makes more sense IMO to have a unit that will be harder to dislodge.

I also, I agree that the AI currently has some trouble distributing units. I believe some of that is the AI trying to build a SOD. If the AI has 50 units for 5 cities, it currently seems to be 38-40 in one city (border) and then a handful in a few larger cities and then only 1-2 in 'safer' cities. I believe it would be better off redistributing those forces a bit rather than a single mass.

 

[kettyo]

Good analogy and point.

Wodan

edit: Kettyo, I think he was using Monty as an example... saying that he thinks the AI should react the same way WE do when we start next to, say, Monty (or Alex or Tokugawa etc)

OK sorry my fault then.
I agree but i think AI should handle humans as high threat even without the AggAI option

 

[Uncle_Joe]

War is still the fastest way to win.

Yes and no. Its a very fine line to tread IMO. Sure, most of the games where I've won I've had to take out a nearby opponent or at least weaken them somewhat. And many of the games I've lost have been because an AI on another continent is able to solidify control early on and just become too big to deal with.

But a large part of that is due to the nature of Civ4 itself. There is very little you can do to impact a rival outside of military conflict. If someone has more/better land than you and you are other otherwise equal, then eventually they'll pull ahead unless you do something about it. But in Civ4, almost the only variable to use IS to attack (either to take land from them or at least to wreck their homeland bring them down that way).

If it becomes exceedingly difficult to attack a neighbor, then eventually the game comes down to who has the best start area. Obviously that is overstating it a bit, but it is the logical endline if conquest becomes TOO hard.

Personally, I think 1/30 strikes a pretty good balance at the moment. It could still use some tweaks to force distribution and whatnot, but the cost of warfare seems pretty appropriate in most cases. Yes, I've seen some 'death spirals', but I've also seen some quick blitzkrieg-style conquests.

2.08 has the AIs too weak and I think the 1/16 build was just the opposite. The AIs were overbuilding units to the exclusion of much else in many case. I still see that occasionally with 1/30 (AIs seem to get caught in some 'keep up with the Joneses' loop), but by and large it seems much better.

 

[Elandal]

Problem in AI considering the human a high threat is that only 95% of the human players are backstabbers who'd sell their grannys for the right price..

I would rather the backstabbers play with Aggressive AI (the warmonger setting) so that those who intend to play more peacefully can do so (by not ticking Aggressive AI) without the AI massing units on the border "because the human is a high threat".


I'm currently playing Hannibal (07-01-30 + handicaps, fractal, large, epic, prince) and had a most unusual start. No, having double-stone + marble with fish + double-clam feeding the city isn't unusual (used to be, but seems marble and stone are very common these days). Settled on plains hill stone (one turn to move) and decided that this capitol is clearly a production powerhouse. In next three or some turns I find London..
I almost never go to early war, but here I had London right beside me. So I teched to BW and found copper in my capitol (would've been out of the fat cross if I hadn't moved settler initially). Some 20 axes later (this is why I rarely go to early wars - the number of axes required is too high compared to expansion by settlers), I had five cities without building a single settler. Also built Stonehenge, Great Wall, Oracle (-> CoL), and Pyramids (high production capitol with marble and stone - almost never build a wonder these days as the AI builds them quite early).
Shaka (who was the third civ on this continent - if this had been a pangaea I might have quit already, but I got continents-type map so I was happy) took the last English city, opportunistic as he is.


I was arguing for the point that the human shouldn't be treated as a backstabber, and here I am: going for the first civ I see.. However, this is unusual for me. Maybe one start out of 20 I go to war with axes, and generally I play some diplomacy to make sure that my target (who isn't usually pleased at all with the diplomacy I've played - some trade embargo and so on penalties) isn't too much liked by my friends. Which means that while I'm not Gandhi, I doubt I'm any more of a warmonger than Mansa is, and certainly nowhere in the always-war-Monty/Ragnar range.


By this time, I settled to consolidate my holdings and tech up. I had some cats from the last stages of English Wars, and some veteran axes (the survivors). No iron, no horses, no ivory - only copper (Shaka had horses, iron, and copper). There were no backlands to settle, the only method of expansion would be war with Shaka. Certainly this game was different from what I usually get: I was given almost no choice in methods of expansion. What I did have was tech lead.
So went and declared on Shaka. Took half his empire (got up to maces and trebs during the war, upgrading my veteran axes as I went), and sued for peace when he got maces. Teched up to grens, and declared again with the intent to get the whole continent in this war.


Shaka has had reasonable defenses for most parts. Mixed stacks, half a dozen defenders per city, small stacks attacking my sod as it marches. His problem is mainly that I have more advanced units, but there's another problem: he's got cats. Lots of them. Where a city has half a dozen defenders, it also has half a dozen cats. And no cat ever has attacked my sod. If they had, even my advanced stack would've been halted - collateral damage kills any tech advantage and any stack when applied in high enough amounts (you can stop an infantry-sod with catapults and finish it off with knights - although sieges in that stack would be a bit harder).

I would say that the AI defenses are right currently - except for the siege units. Either use them or don't build them, but having cats sitting in a city, waiting to be destroyed, is not useful. And I think replacing the cats with city defense units would be wrong, getting those cats out with other active defense units to smash the offensive sod would be the right thing to do. In this case I probably would be only slowed down due to technologically superior army, but if we were at tech parity, it'd be enough to stop the invasion completely. Even without cats, the only thing that kept my sod moving at times was the Medic3 unit..


I also have a huge tech lead over everyone. I don't recall this kind of tech lead on prince before.. Maybe it's because of the map? Monty and Cyrus (with two cities, vassall of Monty) on one continent, Asoka and Tokugawa on another, Catherine and Mansa on yet another. So there certainly hasn't been much techtrading between AIs until Optics. Also getting Pyramids (and thus denying them on some AI) has probably helped.
I was able to delay Liberalism to grab Steam Power with it.. Basically the game has been over for a while - taking half the Zuluan Empire would've given me enough land to win cruise to space. I'm guessing I'll finish off Shaka then decide whether there's any point in continuing the game anymore, but considering Monty decided to declare war (with a caravel, not being able to send out any real units) I'm a bit pissed off and probably will continue just to stomp on him. The reason I'm even thinking about that is that I can run most of my cities with the build governor and thus am not bogged down into too much MM

This game feels like at least a level easier than a couple of halfgames I've played with current BetterAI. Maybe early war makes a game easy, maybe wonders made it easy, maybe the map made it easy - too many random issues to say what it was.

I've had some trouble with a mod losing settings (easily restored by setting them again then quicksave -> quickload) - not really important as those aren't gameplay affecting settings but rather interface settings, just annoying as it sometimes happens more than once when playing a single turn. I don't think this is related to BetterAI in any way though, the UI mods have felt flaky most of the time.

I'll attach two saves (one from the beginning and another from when Monty declared war while I'm finishing off Shaka).

 

[Mannu]

I can confirm the problem with siege units not being used to counter-attack a stack outside the city. I've seen this as well. This one change, would I think make conquest for the human player considerably harder, but I don't think it will cause an arms race.

 

[Uncle_Joe]

Yep, I think thats the way to go at this point...help the AIs 'fight smarter, not harder'...ie, I dont think they really need hordes more units than they have now, but they need to make better use of them. As indicated above, 'defensive' siege needs to sally out and attack or else its useless. And a huge SOD on the borders probably isnt as beneficial is having slightly more defenses in more areas.

 

[Arkatakor]

errr... can i have an exact link for downloading this mod? Also, do i need the expansion for it?

 

[Uncle_Joe]

http://sourceforge.net/project/showfiles.php?group_id=178407

There are versions for both Warlords and Vanilla Civ4.

 

[Quagga]

A side comment on the siege units: The scorned Charge promotion has become useful in the 01-30 build.

I've found that mounted units with Charge can really chew up those tons of catapults or cannons now found in AI cities. Sure, they're pretty easy to beat anyway, but with Charge, the horsey suffers much less damage in the process. And given that you know each city will have several siege units to take out, it makes more sense than before to use this promotion.

But the big benefit is when the AI gets Machine Guns. These guys are extremely difficult to take out with any gunpowder unit. Yes, if you have enough of your own siege you can use them, but I often run out before I've dealt with the MGs. This is where Cavalry with Charge really shine!

 

[Nor Me]

In the game I had with the 1/25 build on a world map, Qin managed to destroy the Koreans and Mongols, make vassals of the Indians, Persians, Arabs and Egyptians and conquer most of Russia. Saladin later broke away after refusing a resource demand but that wasn't the most sensible of decisions and Toku signed on to replace him.

Qin ended up getting half of the UN votes for a diplomatic victory just from him and his vassals. This wasn't quite enough to win but if I hadn't been in as good a position to intervene earlier , he would have.

True, he was falling behind in science. He would have been better off sometimes trying to trade with his vassals instead of blindly gifting them tech.

The point is that Aggression can pay off for an AI in the right position.

So my vote goes for 1/30 being a little too soft.

 

[Elhoim]

OK sorry my fault then.
I agree but i think AI should handle humans as high threat even without the AggAI option

I don´t think so. I think it should consider the human player a high threat when he is amassing forces, especially near it´s borders.

 

[Roland Johansen]

Partial Game report:

I've started a game with the 30-1 build.

Aggressive AI, huge fractal map, epic speed, conquest is the only victory condition, emperor level. I'm quite comfortable at playing emperor level as I normally play at higher levels with the unmodded AI.

I was first attacked by Mehmed in 1990 BC, I got peace a few hundred years and many deaths (mostly his) later. Then I got attacked by Mao Zedong around 1300 BC. Again a few hundred years and many deaths later I got peace again. Mao marched on with his left over troops to attack Wang Kon. Then Alexander declared war around 700 BC. I can't get peace with him yet.

Then 385BC, disaster year. Mao asked me to join his war against Wang Kon (yeah, right, as if I didn't have enough on my hands) and then made peace the same year and attacked me the next turn with the stack that was moving through my lands towards Wang Kon. In that same turn (385BC), Hannibal attacked.

So I've been the victim of 5 out of 6 war declarations in my neighbourhood (I know 6 AI's by now).
Now I'm at war with 3 emperor level opponents, 2 of which are stronger in the power graph. It could become my shortest game yet. But I'm not giving up yet. But I really need peace.

Next to that, there is no copper in my neighbourhood (say within 15 tiles of my capital, haven't explored any further) and so I've only been able to build axeman since 385BC when I connected an iron mine with my fifth city.

I'm not really complaining about this game. I mean, I did want aggressive AI's but I really hope that they start being aggressive towards oneanother instead of towards me. I'm not that weak in the power graph, there are other targets. Just bad luck, I guess.

One points of concern. Usually, I can barely keep up with the AI expansion rate at emperor level. But this game, I'm the one who's expanding the fastest. The AI should realize that there is a lot of open territory around it on huge maps and it should claim more of it or else it will fall behind later in the game (if I get there ).

Okay, now lets see if I can survive a three front war...

 

[Roland Johansen]

(continuation from post no 155, just above)

Ok, it's not 425 AD and I still alive

I've just been declared war upon for the tenth time, soon I'll lose count Why do they all pick on me? I'm not the weakest, at least not according to the power graph and there is someone else with a different religion (allthough they don't all know him). I'm actually hoping that the religion of my periodical enemies spreads to my cities, but for some reason all the other religions in the world do spread to my cities but not the state religion of four of my neighbours. I have 3 different religions in my lands and 3 of my 10 cities don't even have a religion. Just bad luck I guess. But I'm not going to switch to another religion than that one or I shall surely be the victim of even more war declarations.

Now, why am I still alive in an emperor level game where the AI seems to think I'm the anti-christ and I need to be destroyed?

The major reason is that they seem to attack with mostly spearmen, some axemen and archers and a few swordsmen. Now, I know, I do have some chariots so a stack of only axemen would not work that well but a stack of 6 spearmen, 2 axemen and 1 swordsman is a bit weird in my opinion. The lack of swordsmen might be related to getting iron relativery recently. But the huge amount of spearmen compared to the axemen is weird.

My guess is that the AI compares the 100% bonus vs mounted to the 50% vs melee and thinks 100% > 50% thus spearmen are better than axemen (apparently the bonus is more important than the 4 vs 5 strength difference). Something simple like that would explain the love for spearmen over axemen. Maybe the code should look at city defenders and realise that mounted troops don't get a defensive bonus and are thus bad city defenders. Therefore spearmen are not needed a lot for city attack duties. The spearmen are clearly chosen as city attackers as they even have the city attack promotion.

The other reason is that while they are building all these units and using them as cannonfodder against my defending city garrison II archers and combat I, shock axemen, I'm also expanding a bit. We're living on a huge world together and there's more than enough space for all of us. But it seems that the AI doesn't realize this and just builds very few settlers compared to the number of units. Now I know, it's the aggressive AI setting, but still some sensible expansion would be good even for the uberaggressive homocidal AI.

I'll play some more tomorrow and see how much longer I will survive the onslaught. I already have 2 great generals from defending and the third one is more than halfway.

 

[Iustus]

(continuation from post no 155, just above)

Ok, it's not 425 AD and I still alive

I've just been declared war upon for the tenth time, soon I'll lose count Why do they all pick on me? I'm not the weakest, at least not according to the power graph and there is someone else with a different religion (allthough they don't all know him). I'm actually hoping that the religion of my periodical enemies spreads to my cities, but for some reason all the other religions in the world do spread to my cities but not the state religion of four of my neighbours. I have 3 different religions in my lands and 3 of my 10 cities don't even have a religion. Just bad luck I guess. But I'm not going to switch to another religion than that one or I shall surely be the victim of even more war declarations.

Now, why am I still alive in an emperor level game where the AI seems to think I'm the anti-christ and I need to be destroyed?

The major reason is that they seem to attack with mostly spearmen, some axemen and archers and a few swordsmen. Now, I know, I do have some chariots so a stack of only axemen would not work that well but a stack of 6 spearmen, 2 axemen and 1 swordsman is a bit weird in my opinion. The lack of swordsmen might be related to getting iron relativery recently. But the huge amount of spearmen compared to the axemen is weird.

My guess is that the AI compares the 100% bonus vs mounted to the 50% vs melee and thinks 100% > 50% thus spearmen are better than axemen (apparently the bonus is more important than the 4 vs 5 strength difference). Something simple like that would explain the love for spearmen over axemen. Maybe the code should look at city defenders and realise that mounted troops don't get a defensive bonus and are thus bad city defenders. Therefore spearmen are not needed a lot for city attack duties. The spearmen are clearly chosen as city attackers as they even have the city attack promotion.

The other reason is that while they are building all these units and using them as cannonfodder against my defending city garrison II archers and combat I, shock axemen, I'm also expanding a bit. We're living on a huge world together and there's more than enough space for all of us. But it seems that the AI doesn't realize this and just builds very few settlers compared to the number of units. Now I know, it's the aggressive AI setting, but still some sensible expansion would be good even for the uberaggressive homocidal AI.

I'll play some more tomorrow and see how much longer I will survive the onslaught. I already have 2 great generals from defending and the third one is more than halfway.

The next build will have some changes in when AIs build settlers, which should belp this situation.

-Iustus

 

[Roland Johansen]

The next build will have some changes in when AIs build settlers, which should belp this situation.

-Iustus

Ok, that sounds good. I also seemed to notice this behaviour in previous builds so it's probably not something very new. I'm now expanding twice as fast as my opponents (10 cities versus 5 or 4) and I'm hindered in my expansion speed by the constant war and the jungles surrounding my starting position and of course the always limiting city upkeep cost.

And what about the issue of the abudance of AI spearmen in city attack groups? I could send savegames if needed, but it is just a description of general behaviour.

 

[darrelljs]

I've just been declared war upon for the tenth time, soon I'll lose count Why do they all pick on me? I'm not the weakest, at least not according to the power graph and there is someone else with a different religion (allthough they don't all know him).

I've noticed something similar with this build...top tercile in power but a lot of ganging up on me. My guess was that the AI has been bribing other AIs to go to war against me, based on the timing of the declarations. Is that possible in your case? I found I couldn't bribe anyone to join in on my side, so I've modified my "avoid adopting a religion" diplomatic strategy (which coupled with a strong military used to allow me to avoid wars altogether) to one where I try to adopt the religion of a few hopefully stalwart allies. It seems to have worked out better, but I don't have a very large sample size so it could be coincedence.

Darrell

 

[Iustus]

I've noticed something similar with this build...top tercile in power but a lot of ganging up on me. My guess was that the AI has been bribing other AIs to go to war against me, based on the timing of the declarations. Is that possible in your case? I found I couldn't bribe anyone to join in on my side, so I've modified my "avoid adopting a religion" diplomatic strategy (which coupled with a strong military used to allow me to avoid wars altogether) to one where I try to adopt the religion of a few hopefully stalwart allies. It seems to have worked out better, but I don't have a very large sample size so it could be coincedence.

Darrell

Sharing borders is also a pretty prominent factor. If you turn on chipotle, you can see how the AI feels about going to war with each civ, to see if you really are on the top of the list, or if you are just unlucky in the random rolls.

-Iustus