War AIs

[EsoEs]

So I know we don't have access to a lot of the AI code, but Im wondering if we have access to the code that relates to AI denouncements. We have all observed how the war AIs like Greece, Aztecs, to a certain extent Japan etc do not usually do as well as some of the more peaceful builder AIs. I think one of the reasons behind this is their tendency to over denounce. I just started a game as Rome and before even reaching classical, the Aztecs have denounced every other player I've met so far (about 4). Their start lends themself to this a bit but still, theres absolutely no way that they are ever going to be a threat past the Medieval era. They don't have a single ally, and are basically just painting a huge target on their backs. I'd like to see if we have access to this code, to see if we can refine it a bit to make them a bit more selective. Pick a target, don't mass denounce. This would, imo, make these kinds of civs much more competitive as they would focus more on their target, and receive less AI hate. Thoughts?

[rhettrongun]

Each leader has ai flavors for it, so a simple xml mod can be used to reduce the tendency that a leader makes denouncements. Maybe I should add that to include this in the unofficial patch is not the best idea imo. Having the denouncement happy leaders adds to the variety of the game, and for me this is good.

[gdwitt]

I select Montezuma and Rome when I am playing higher levels because they take the early pressure off my back.
On emperor or above, the AI seems to gang up on the human player unless their is an outlier civ that is causing trouble.
I just stopped playing a game where I was denounced and then attacked by every civ but Rome when I refused to attack Rome. I held out for 120 turns, but it wasn't fun and lost the opportunity to get wonders. I had 3 chokepoints and citadels.
At the end, it was only the tenacious Ottomans that refused to back down from war. I don't know what it was with them.

I wonder if we can defuse the AI tendency to gang-up all at once? Or add more ways to create positive relations with AI civs to prevent mass denouncements.

[Sneaks]

The AI should gang up all at once. In attacking a player put into a very weak position, they not only give themselves a chance at military expansion, they better relations with the other attackers. Mass denouncements only happen occasionally in my game to me, and almost always when I go ahead and denounce someone or expand my borders too fast. In both cases, I have done things that diplomatically seem threatening to those around me.

[EsoEs]

Originally Posted by Sneaks

The AI should gang up all at once. In attacking a player put into a very weak position, they not only give themselves a chance at military expansion, they better relations with the other attackers. Mass denouncements only happen occasionally in my game to me, and almost always when I go ahead and denounce someone or expand my borders too fast. In both cases, I have done things that diplomatically seem threatening to those around me.

This is kind of my point, we can control the circumstances around our denouncements somewhat, but certain AIs cannot help themselves, and just denounce everyone around them for the reason that they are around them. They see the other civs as being threatening merely because they start less than 20 tiles away. If we (and by that I mean the programmers ) have access to both the code and some spare time I wouldnt mind them taking a look at this and possibly just tone it down slightly. Although like gdwitt said it could be that this feature is a necessary evil of the higher difficulties, easy way to make friends, distract other AIs, and possibly pick up some easy war gains early.

[GamerKG]

Originally Posted by EsoEs

This is kind of my point, we can control the circumstances around our denouncements somewhat, but certain AIs cannot help themselves, and just denounce everyone around them for the reason that they are around them. They see the other civs as being threatening merely because they start less than 20 tiles away. If we (and by that I mean the programmers ) have access to both the code and some spare time I wouldnt mind them taking a look at this and possibly just tone it down slightly. Although like gdwitt said it could be that this feature is a necessary evil of the higher difficulties, easy way to make friends, distract other AIs, and possibly pick up some easy war gains early.

I would certainly see people as being a target/foe if they started less than 20 tiles away depending on my playstyle! I have never seen this uber-denouncing going on, but I do hate how every AI declares war on you really early when they:

1. Are too far away to fight properly
2. Are too far away to properly consolidate their new holdings and assimilate into the empire
3. Can't possibly afford the happiness for my cities. I mean seriously?
4. Must have far better things to do than declare war.

And by really early I mean before swordsmen/horses. Just archers and spearmen rush. Like what the hell?

[Zaldron]

Originally Posted by GamerKG

I would certainly see people as being a target/foe if they started less than 20 tiles away depending on my playstyle! I have never seen this uber-denouncing going on, but I do hate how every AI declares war on you really early when they:

<snip>
3. Can't possibly afford the happiness for my cities. I mean seriously?

And by really early I mean before swordsmen/horses. Just archers and spearmen rush. Like what the hell?

Actually most likely given that you have 1-3 happiness resources they don't your cities will be break-even or a net gain at that stage in the game.

[GamerKG]

Originally Posted by Zaldron

Actually most likely given that you have 1-3 happiness resources they don't your cities will be break-even or a net gain at that stage in the game.

I can barely keep myself afloat, how the hell can they when they don't even know how to upgrade their land with workers?

And as soon as any of those cities gains a person, its happiness hell for a while.

[tlaurila]

I've found myself the target off ganging up, but usually only if I try to be friends with everyone. The logic is that if you have no enemies, then you don't really have friends, either. I do like that, it forces you to pick a side, choose who your friends are, or risk being the target everyone loves to hate.

So you need to pay attention to the diplomatic spread, identify those you want to be friends with, and then face the fact that you need to denounce common enemies and refrain from provoking your friends. It isn't always as easily done as it's said. One thing I've found is that it's always easier to have a friend who's not an immediate neighbor. That "covet lands you own" can be very hard to overcome.

I really like the fact that you risk getting everyone ganging up on you if you try to play all sides.

That said, AIs trying hard to avoid fighting wars on all fronts simultaneously would be good. That is, that the AI wouldn't go declaring wars when already engaged in one, especially if the active war is with someone the civ has a border with.

[Sneaks]

The AI is in fact less likely to engage in a second war if engaged in a current war, unless it really likes its current position.

[Txurce]

Yes, some AI are serial DoW'ers, but concurrent DoW's are pretty much limited to late-game bids by powerful AI's that don't know they can't invade another continent, and therefore think they can win by Conquest.

[EsoEs]

I wasn't talking so much about DoWs as I was the tendency for the more aggressive AIs to over denounce. This will lead to ganging up scenarios, but the problem (if there is one) lies with denouncements.

[Txurce]

The AI thinning out its own ranks usually makes for a more competitive game. Compulsive denouncers like Greece and the Aztecs tend to get hit as a result, but 1) they have their share of success anyway, and 2) there's probably no better system than the present one, where AI's have personalities, but with some RNG variance that keeps the outlier roster shifting a little bit each game.

[tlaurila]

Originally Posted by Sneaks

The AI is in fact less likely to engage in a second war if engaged in a current war, unless it really likes its current position.

Less likely, sure, but still have seen it do idiotic DoWs actively opening up a second front on itself. So could be less yet.

[Sneaks]

I am EXTREMELY hesitant to lower the AI's willingness to war anymore than it currently stands. It is pretty easy to survive peacefully for a very long time in VEM, and while some AIs might get themselves in trouble, I don't think this is a bad thing. Even stupid AIs create odd entangling alliances and such, which make the game more entertaining for me.

EsoEs, in regards to the over-aggressive war-morongers like Monty, this is due to Thal spiking certain values in their leader XMLs. I would agree that it might be a bit too much as being an to everyone from turn 2 tends to be a bit TOO self destructive.

[Dunkah]

Slightly off topic... but I would like to see the Ai's have a "roleplay" personality, besides just playing to win. I would like to see it set up so that maybe 50% of the Civs are roleplaying and 50% are trying to win with vastly different personality settings.

Either that or give CS's real diplomacy options so that they can be the roleplayers.

That would give a huge dynamic and real reasons to actually denounce rather than just denouncing for the sake of pissing somone off.

Denouncements should be linked to something. You did x so I am denouncing you. Then other leaders who didn't like x or liked the civ that denounced could jump on board.

Some day...

[EsoEs]

That would give a huge dynamic and real reasons to actually denounce rather than just denouncing for the sake of pissing somone off.

I never denounce to piss off an AI, since they are robots after all ^^ I denounce certain AIs in order to gain more friendly points with the other AIs, specifically the AIs that are also denouncing this player.

EsoEs, in regards to the over-aggressive war-morongers like Monty, this is due to Thal spiking certain values in their leader XMLs. I would agree that it might be a bit too much as being an to everyone from turn 2 tends to be a bit TOO self destructive.

Sneaks Im glad you agree with me. I know its nice to have a wide range of personalities in the game, but I do think that the extreme war mongers are very often self destructive, not necessarily in their DoWs, but certainly in their denouncements.

[Txurce]

Originally Posted by EsoEs

Sneaks Im glad you agree with me. I know its nice to have a wide range of personalities in the game, but I do think that the extreme war mongers are very often self destructive, not necessarily in their DoWs, but certainly in their denouncements.

Thal will probably answer more fully, but I think this is the only way to have as much "personality" as we have. The extreme warmongers are self-destructive to a degree, but again, the biggest offenders in my games are Greece and the Aztecs, and they thrive a reasonable amount of the time.

If there were going to be an adjustment, I would have a civ like the Mongols be less aggressive (because they can't walk the walk), and a civ like Persia (for example) more aggressive, because it can back it up from the start.

[Thalassicus]

It's possible to change an AI's overall willingness to denounce, but not which targets they select (targeting is done in the game core).

One thing I've noticed is the earlier one AI meets another AI in the first 50 turns, the more likely friendly relations occur. When I did some testing where every leader starts in contact with everyone else, they all became great friends for about 100 turns, and no AIs did their normal suicidal rushes. This gave militaristic AIs time to expand in the early game -- and they did much better. After this initial round of universal DoFs expired, borders were closer to one another, relations soured, and militaristic AIs made successful war campaigns conquering their enemies.

So, I've been considering giving AIs contact with their immediate neighbors at the start of the game. I'm going to wait until the next beta cycle starts in a week, however. Right now I'm focusing on bugfixes in preparation for beta's move to a production release next Wednesday.

[Txurce]

I've also noted the early neighbor attitude, which often includes the human player. I actually found those games less interesting, but appreciated them as part of the variety that the present balance has. This variety also includes mass denouncers, which in turn leads to them sometimes paying a price. Unless one civ in particular is always getting squashed as a result, I don't see why it's a problem, any more than the early neighbor pattern from the other end.

I've liked probably all our experiments regarding AI diplomacy, but I don't see why you'd feel establishing immediate contact is worth a beta test. This change seemingly has other implications (OBs, etc) that would consequently distort the human-AI relationship. (I would dislike it even more if it included the human player.) From your description, it sounds like you're treating a debatable problem (I see it as a good thing) with a sledgehammer.