AI Aggression levels (please discuss here).

[Blake]

No the train unit is a fuzzy thing...
Basically at a certain point in the (Vanilla) code for choosing production - after things like basic defensive and economic needs have been taken care of, the AI will have Build Unit % chance of training a unit, otherwise it will try to build a building, and if it can't build a building, it'll then train a unit.
Defense
Settlers Workers
BuildUnit% check:
- Choose a unit
- or Choose a building
Choose a Unit
Choose a Building
Process.

What BetterAI does to the build selection process is adds a lot more context sensitive decision making - rather than just having a simple dice roll decide whether to build a unit or building, the AI will actually build a unit if it feels underdefended, or build an economic building if it'll pay off quickly, or prepare a dagger stack and it'll build a barracks before training significant units... and so on. Only after all that context sensitive stuff is done, does it finally get to the simple BuildUnitProb dice roll.
Once the AI runs out of buildings to build, it goes into unit spam mode regardless - until it hits it's quota of units and uses a process, if one is available.

It is somewhat fair to say that the buildUnitProb is indicative of how large an army the AI would like to keep on hand - especially offensive army above and beyond the basic defensive requirements. The probability itself doesn't mean much - but Isabella's being twice as large as Gandhi's means that Isabella should have twice as many offensive units...

 

[Elandal]

That chart does indeed shed new light to my experiences. Seems I've been hit by closet warriors where I thought I had peaceful neighbourhood.

 

[Wodan]

The only way to survive in 1/25 (short of luck) is to horde military. You cant neglect it even for a little while or else your military rating falls and its tough to catch back up. So, I end up with stacks and stacks and stacks of military units every game.

Have to disagree there. My latest game I started on a peninsula. Continents map. Louis was in the middle, and Alexander down south.

I spread my religion to Louis, and tossed him some bones (extra fish or something, free gift, plus some cash I think). I did the same thing to Alex.

Louis, true to his nature, built a wonder or something and had strong cultural pressure on 2 of Alex' cities.

Quite soon, Alex, of course, declares on Louis. He asks me to help. I tell him to go sit and spin (not in those words of course). Louis asks me to help. I say yes. Not a single unit of Alex' attacked me. (Louis was between me and Alex, remember.)

After a bit, I sue for peace with Alex.

After a bit more, I send Alex some cash to declare peace with Louis.

Okay, repeat that cycle twice more. My dip relations with Louis are huge by this point. (I have a +5 or something for religion, +4 for agreeing to help him in wartime, +4 for good trade relations, +2 for supplying him with resources, yadda yadda.)

After the second time Alex had all but wiped Louis out. I still hadn't fought a single one of his units. Yet, I had just a single unit in every one of my cities, either a warrior or a chariot. Alex had swords and phalanxes all over the place. Anytime I wanted, I could have build some units and saved Louis' bacon. But, I didn't have a problem at all with Alex both wiping him out and also with Alex choking himself with empire maintenance and unit supply costs.

So, I guess I have to take exception to your "only way" phrasing. There is at least one set of circumstances where it can be done. I daresay there are more.

ps by the time Louis was gone, my tech level was way way ahead. I was building Cav and Muskets while Alex had just barely gotten longbows.

Devil's advocate now. Game before this one, with the 1/9 build (I think). Anyway, I did Aggressive AI and got a middle pangaea start. Holy cow was I torn asunder.

I'm saying thats its going to be impossible to program an AI to make 'judgement calls'. In some cases, I absolutely think it would be a good idea to double-team the leader. In others I dont. And I certainly wouldnt want to get to a situation where every time you are about to win, 5 AIs pile on because its the 'smart thing to do'.

Its not really a solvable equation IMO because there isnt even a 'right' answer to be shooting for.

You've got a point, but I think it could be done. And I definitely think that there should be an increased chance of AIs declaring war if you're in the lead and/or about to win.

I agree having ALL of them declare war EVERY time would be bad for the game. There shouldn't be a kingmaker phenomenon going on (AIs with no chance to win knocking down the leader simply so 2nd place could win). Also, peaceful AIs that you have good relations with should stay that way.

However, military AIs, especially on aggressive settings, by all means should attack. They've probably been doing it all game, though, so that's nothing new.

The real question is a "neutral" AI, who has a realistic chance to win if he "changes sides" against you. I definitely would be in favor of the BetterAI programming this to happen.

Its a big complicated ball of wax that I cant see being resolved any time in the near future. My understanding is that the AI team is trying to build towards the 1.0 release. And in that case, I think we are farther from what I consider to be 'fun' than we were a number of weeks back. YMMV.

Thanks for your thoughts on the rest of the stuff. No comment except on this last. To me, I'll repeat that if the BetterAI team reduces the building of huge garrison stacks this will help immensely IMO. I think they've already said they're doing this, so we'll have to wait till the next build and see.

This is something I probably have to work on. I really don't like settling so that I can't secure a culturally owned path to my core, but maybe it's necessary at times.

I think I noted the backyard backfill in "lucky start" category, although after considering Snaaty's strategy this seems doable without reliance on lucky start that gives the backyard to you: settle forward to create the backyard. Again this is something I probably should try.

Give it a try. I do it more often than not. It's hugely powerful. It's my main "counter" to the people who say you HAVE to do an early war to do well on high levels (because you need the territory). One or two well sited cities can cut off galley/coastal approaches, or can block off a neck of land and give you the majority of a continent and/or subcontinents to expand into. One or two cities with higher maintenance early in the game is a paltry price to pay.

Without quoting more, I do agree with you on many points. Your points regarding what smart warring might mean were definitelly good ones. Indeed if understanding the big picture can be broken down into small steps the AI can take, it'll be closer to the goal of being smart.

Yeah, I guess we'll see what Blake and Iustus are able to pull off.

Wodan

 

[Arlborn]

No the train unit is a fuzzy thing...
Basically at a certain point in the (Vanilla) code for choosing production - after things like basic defensive and economic needs have been taken care of, the AI will have Build Unit % chance of training a unit, otherwise it will try to build a building, and if it can't build a building, it'll then train a unit.
Defense
Settlers Workers
BuildUnit% check:
- Choose a unit
- or Choose a building
Choose a Unit
Choose a Building
Process.

What BetterAI does to the build selection process is adds a lot more context sensitive decision making - rather than just having a simple dice roll decide whether to build a unit or building, the AI will actually build a unit if it feels underdefended, or build an economic building if it'll pay off quickly, or prepare a dagger stack and it'll build a barracks before training significant units... and so on. Only after all that context sensitive stuff is done, does it finally get to the simple BuildUnitProb dice roll.
Once the AI runs out of buildings to build, it goes into unit spam mode regardless - until it hits it's quota of units and uses a process, if one is available.

It is somewhat fair to say that the buildUnitProb is indicative of how large an army the AI would like to keep on hand - especially offensive army above and beyond the basic defensive requirements. The probability itself doesn't mean much - but Isabella's being twice as large as Gandhi's means that Isabella should have twice as many offensive units...

It sounds you made a very good way to the AI decide what to build

 

[kettyo]

Blake, got some question about your recent chart.

'Train' is buildunitprob or that something in leaderheadsinfo.xml?
'War' is maxwarrand?
'Backstab' is dogpilewarrand?
'Warpleased' is nowarattitudeprobs?

I think yes but i cannot look at the file now it's not in this computer. (edit: found a copy in temp dir hurray )
Thanks.

 

[Blake]

Correct. BuildUnitProb is in leaderheadsinfo.
And there are of course more factors that go into the AI personality, but that set is somewhat representative.

 

[kettyo]

Correct. BuildUnitProb is in leaderheadsinfo.
And there are of course more factors that go into the AI personality, but that set is somewhat representative.

Your chart is somewhat flawed.
For instance Churchill and Cyrus have 100% nowarattitudeprob for pleased so they should never attack pleased rivals.
They have 70% for cautious so they should also be reluctant to attack their neutral rivals (30% prob).

 

[cf_nz]

And I definitely think that there should be an increased chance of AIs declaring war if you're in the lead and/or about to win.

I agree having ALL of them declare war EVERY time would be bad for the game. There shouldn't be a kingmaker phenomenon going on (AIs with no chance to win knocking down the leader simply so 2nd place could win). Also, peaceful AIs that you have good relations with should stay that way.

However, military AIs, especially on aggressive settings, by all means should attack. They've probably been doing it all game, though, so that's nothing new.

The real question is a "neutral" AI, who has a realistic chance to win if he "changes sides" against you. I definitely would be in favor of the BetterAI programming this to happen.

That sounds pretty good, I was mildly concerned from your first mention of it that you would advocate ganging up on the leader or backstabing friends to win. Whilst a lot of human players may well do that (I tend not to) I don't know thats it's a good idea gameplay wise.

I have no issue with your enemies (regardless of position) or neutral civs (in a position to win) attacking the leader if they are about to win (I'm not sure about en-mass attacks though, all against one; maybe friendly civs should be more willing to help out, suppport a friendly victory, even without direct benefit). Friends should stay as friends, perhaps try to ally with a leading friendly Civ (note I always play perm alliances).

 

[kettyo]

So anyone who won't like occasional attack on friends should set nowaratitudeprob values for pleased and friendly to 100% for all leaders as far as i understand.

 

[kettyo]

Because of this backstabbing tendencies AI's maybe should keep a bit more units in friendly borders not just the 1-2 we see now in the 1/25 build. (Noble*)

 

[kettyo]

To the idea of AI's being obstacles in the game and how the game was intended to be played.

As a fact, as Civ4 came out AI's are really obstacles.
But Civ is a game series and it never been this way AFAIK.

Just an example:
I checked back Civ1 (Civnet actually) recently and wasn't surprised that the Russians (a Civ1 warmonger) wiped out 2 civs until 1880 BC !
And they don't play like an 'obstacle' at all
They are dominating the game. Actually they would have won by domination if that was possible in Civ1.
Still very likely they will won by spaceship. They have half the world.

And that is common in Civ1.

Also don't forget the massive wars of Civ2.
There sometimes esp. when you're weak you just can't make peace with an AI. "We've decided to rid the world of your worthless civilization" Are you remember?
You can't even beg

In Master of Orion (which is actually also a Civ just in space) races are going for victory desperately and there are multitudes of wars in the game.

So both thoughts are valid:
1.) Firaxis/Sid changed their usual gamestyle for Civ4 to lean it towards a restricted AI game
2.) They didn't have the time or enthusiasm to adapt a similar AI behaviour to Civ4 it was in previous games.

I think 2.) might be closer to truth but actually it doesn't matter that much.
We grown up on Civ and MoO and an AI which could give us feelings of the old competitive AI's is a good thing.

To tell the truth when Civ4 came out i was disappointed how insanely pacifist the whole game was.
It felt very wrong to me.
Not a single war in an entire game or only some marginal ones.
Now we are closing to a state where it resembles the old days.

That's my point of view.

 

[Blake]

Your chart is somewhat flawed.
For instance Churchill and Cyrus have 100% nowarattitudeprob for pleased so they should never attack pleased rivals.
They have 70% for cautious so they should also be reluctant to attack their neutral rivals (30% prob).

I have corrected the entries. Although it doesn't change my points any (other than the specifics).

In Master of Orion (which is actually also a Civ just in space) races are going for victory desperately and there are multitudes of wars in the game.

Master of Orion has excellent diplomatic concepts (some flawed ones too, like the "Enemy of my Enemy") but on the whole it's a good model for 4X diplomacy and warmaking. I especially like the turf wars / cold war concept. That you can kill a unit and say "What'ya gonna do about it, huh?". It's a very nice flexible system.
And it's also nice that the AI's have diverse personalities rather than being tied to a specific leader.

 

[cf_nz]

Not a single war in an entire game or only some marginal ones.
....
To the idea of AI's being obstacles in the game and how the game was intended to be played.

Yeah, no wars are boring. Though I have had a number of BetterAI games like that (not the most recent build though).

I'm warming more to the idea of AI's being more like opponents than obstacles; Blakes 'friendly multiplayer' decription sounds like a good compromise. My concern was that a 'win or else' human attitude would cause AI's to loose their personality (for me leader personality is a key Civ concept).

 

[kettyo]

Master of Orion has excellent diplomatic concepts (some flawed ones too, like the "Enemy of my Enemy") but on the whole it's a good model for 4X diplomacy and warmaking. I especially like the turf wars / cold war concept. That you can kill a unit and say "What'ya gonna do about it, huh?". It's a very nice flexible system.
And it's also nice that the AI's have diverse personalities rather than being tied to a specific leader.

I like the preemptive strike concept best.

When they see you're gathering ships in a nearby system and either they tell you to stop it or there'll be war (and will) or they don't say anything just take the first shot at you telling something like "Don't take us for fools! Your massing of fleets in XY star system is an act of war. Now pray to whenever gods you believe..." after it.

I also like the event system especially that sometimes you get into some totally unplanned and unwanted wars because of it messing things up a lot and making it very hard or impossible to be in good terms with everybody while they kill each other with diligence (which could be done in MoO2).

Also sometimes a leader got killed by an assasin thus the 'personality' of a rival changed. Nice idea.

 

[kettyo]

My concern was that a 'win or else' human attitude would cause AI's to loose their personality (for me leader personality is a key Civ concept).

Sure it is.
They must try to win according to their personality.

 

[cf_nz]

When they see you're gathering ships in a nearby system and either they tell you to stop it or there'll be war (and will) or they don't say anything just take the first shot at you telling something like "Don't take us for fools! Your massing of fleets in XY star system is an act of war. Now prey to whenever gods you believe..." after it.

I'd love to see that in a future Civ. It annoys me right now that there is nothing diplomatic you can to if you see an attack coming. I posted a thread about it some time ago. If you strike first you get 'you declared war on us/our friend' penaties; or you wait and let them invade; or bribe someone to attack them first. None of the scenarios are particularly great. I hope Firaxis puts some serious effort into diplomacy for the future expansion/game.

 

[scu98rkr]

Completely out side the scope of this project but it would be cool if a CIV is stagnating it could swap leaders. Either if a warmonger has over expanded and needs to catch up technologically or if a peaceful leader isnt using his technological advantage militarily.

 

[kettyo]

Completely out side the scope of this project but it would be cool if a CIV is stagnating it could swap leaders. Either if a warmonger has over expanded and needs to catch up technologically or if a peaceful leader isnt using his technological advantage militarily.

Yes would be great but surely impossible through this mod.
And would be really annoying when you are dispatched

 

[Arlborn]

Yes would be great but surely impossible through this mod.
And would be really annoying when you are dispatched

haha would be fun

What a off-topic! Post it in General section of CIV4, who knows

 

[Quagga]

Here are the "Bastard Charts" to show which leaders are the biggest scumbags, according to their probability of training units, their probability of declaring war, their probability of shamelessly dogpiling and finally their probability of declaring war at Pleased - the Total column is basically their "Bastard Rating" - the higher the total, the more likely they'll cause trouble. For all numbers, bigger means badder.

The chart is cool, but the column headed as "Backstab" is really meant to be "Dogpile"? That's what you say in the paragraph above. I would think that "backstab" refers to the probability of DoW when Pleased.

I've been having a very entertaining game starting with the 01-25 build and now using the 01-29. As I noted somewhere, Stalin has been in a rolling war through the whole game. I made peace with him for a while -- he declared on Monty then -- but now he's back at me. Mehmed decided to join in the fray, for no reason other than dogpiling, near as I can tell.

Until Mehmed attacked, I had a scout in his borders. Most of his cities had only a catapult as a defender. However, he has come after me with three stacks of elephants, knights, crossbows, and pikes. I lost a city that I captured from Washington (a Stalin vassal) -- a city with no buffer tiles. I should be able to repel him before losing anything else. The wildcard is Shaka, who is Friendly with me, but won't join in any reindeer games -- er, I mean, wars.