AI Aggression levels (please discuss here).

[Blake]

1/30 Report:
Well I can say I'm glad I tried it. IMO, HUGE strides have been made in alleviating some of the problems that I routinely encountered in the earlier builds. I still believe that there is probably a need for some further tweaking, but I have to admit that I was very favorably impressed in the near two games I played out with 1/30.

That's good to hear. And it's good that changes are working as intended .

1) Some of the AIs are still perhaps overbuilding units. The attached screenshot is from my first game. The Ottomans had almost 50 units stacked in a nearby border city (including over 30 Catapults and Trebs which had been converted to Cannon by the time of the screenshot). To me, that is still too many for that time in the game. It takes a LOT of hammers to build that huge force and it literally sat there until the Modern Age when it was converted into Artillery. That is a huge expense to be sitting on that long. I hestitate to want to see it totally changed but perhaps redistributing it a bit so its not all sitting in one huge stack? In my second game, there have been a lot more smaller wars which generally has kept the SODs reined in a bit. I havent seen stacks bigger than 12 or so in that game and it feels pretty 'correct'.

From looking at the minimap it appears that Mehmed and Isabella are best-buddies. He's basically put nearly his ENTIRE army where he feels the most hostility. I guess he has about 10 cities, so that's only 5 units per city (plus 1-2 normal garrisons) so maybe 7 units per city. That's not a huge army and would be costing him probably only 5% of his GNP, while somewhat expensive in terms of hammers the fact is there isn't really that much to build and the AI is fairly efficient about building up infrastructure quickly - over the course of a game it's really not much to train 7 units per city. So while it's a big SoD I wouldn't call it evidence of overspam.

2) Sometimes the AIs dont know when to call it quits in a war.

Yeah no kidding. The entire war-motivation thing needs a major overhaul. It's not going to happen in 1.0, for now the logic will remain "I attack you because I hate you".

3) The building of much larger number of units is slowing tech progression down to the point where no one even got close to launching the Space Ship by game's end. I think 2 Civs had built the Casings and one was working on Thrusters by 2050. At some point, I think the AIs need to release that enough units is enough and focus on teching them rather than adding to them (new types such as aircraft notwithstanding).

FYIW in my first Warlords (2.00) game under similar settings I launched in about 2048 - tech pace can be pretty slow sometimes. It is worth noting that commerce is somewhat undervalued in the 1/30 build so the AI's are getting less commerce than they should be (that also extends to getting more hammers and thus completing infrastructure sooner and spamming units sooner which means more money tied up in expenses...).

Overall:

Like I said above, I'm actually very favorably impressed. The AIs certainly played a much 'tighter' game in this build. The biggest complaint (early wars of extermination by massed units) seems to be corrected. I think with a few more tweaks and bug fixes, this AI is pretty close realizing what I believe were the original goals for BetterAI. I think it still might be a bit too militaristic (not necessarily warlike, but arms racing), but its feeling like its moving in the right direction towards more balanced play again.

Great job on the latest build!

You're welcome. Balance is always at the forefront of my mind. But understand that a few builds back I had written a COMPLETELY new system for allocating defenders (and attackers for that matter) smartly and a lot of supporting code had to be changed to work with it and then more tweaking had to be done etc etc (the AI is a verrrry complex system it's no co-incidence that so few have risen to the challenge). With an ambitious feature always comes some kinks to work out.... so I'm saying I never lost sight of the goals, but actually just side effects of ambitious changes - and not unexpected side effects (there has to be some bad builds to make the really good builds possible...).

I'm happy enough with it's performance over all that I don't plan to add any new major features for 1.00 - the focus will be on some fine-tuning, this means the next builds will just keep getting better, but it also does mean that for now the AI will keep it's current "fundamental" limitations in diplomacy, war-making and so on.

 

[Mannu]

Playing a game with 1/30 and I can echo some of the feedback that it does indeed seem to be the best build yet. As for AI agression, here's what's happened:

Isabella started the game between Genghis and Shaka, gone by 1000 BC, I don't think she even built a 2nd city.

Hannibal next to Genghis and Shaka, gone by 0 AD.

Oddly, these two are bosom buddies, sharing buddhism.

I'm on an island by myself (now) and decided on a cultural victory (playing Tokugawa). Not the best leader for it, but I thought it setup well with the number of religions I was able to get early and the isolated island. Ragnar had other ideas, he backstabbed me at friendly status landing a 2 galleons of knights on my shores. I can probably fight him off, but at this point it looks like I made a serious miscalculation going for a cultural victory with those other leaders in the game.

All in all it's been the most fun of any Civ game I played. There have been some questionable decisions (not all by me, lol).
1) I don't think that the AI responds to a pillaging stack very well. I sent about 12 units, including 4 longbows into Genghis territory (shaka is between us, I have just one city on that land mass). Instead of attacking it like a city with siege units, he threw units one or two or 3 at a time which I was easily able to fight off.
2) Once this stack was decimated I decided to throw the remainder (about 5 units) at his capitol and declare peace, I actually managed to take it with some lucky dice rolls, but for several turns he had the opportunity to reinforce the city under attack, even just one more unit would have saved the city. He also could have switched to slavery and rushed some defensive units. I should have razed it, but I didn't and he retook it a few turns later...

 

[jkp1187]

(All games played on Tectonics map, noble level, small map, 8 civs, normal speed).

I agree with the war motivation observations. In a recent game, everyone dogpiled on the Vikings (who were living alone on their own continent,) and would not quit. Repeated amphibious attacks were made, a couple of cities taken and re-taken by the disparate allies, but Ragnar kept holding on. At one point, I delcared war on him just to gain some diplomatic points (I did nothing for the war -- I focused on a space race win.)

In addition, I shared a continent with the Incans in the same game. Although I was clearly posing a major threat, and had expanded over more territory than they, Huyana built a surprising number of galleons. What for, I've no idea; it made it that much easier to take him down, though. (On the other hand, the map was heavily sea-based, so it wasn't a completely stupid idea to have galleons on hand, especially because Saladin was located on a nearby continent and it seemed that he and Huyana were constantly squabbling with each other.)

I also noticed that the AI doesn't seem to expand using settlers as aggressively as it probably should (although that just may be the fact that I've "outgrown" Noble level play and find it easy to out-expand them now.)

For what it's worth, I have played two games to completion using 1/30 -- one to a space race win (launch around 2006), another to a cultural win (around 1990). In general, the military stacks are balanced (and I note that in the Inca game referenced above, Huyana managed to get Military Tradition before me and successfully used cavalry for defense of cities in lieu of musketmen, forcing me to call off my first attack on him until I got riflemen.)

Oh -- one more comment -- In the culture-win game, I joined in an attack by my ally, Tokugawa, on Korea. Toku had some strong stacks of rifles/cannons (although Wang Kong kept chewing them up,) so I sent in a couple of longbowmen and pikemen on a pillaging mission. I had them all over the place, all the way up to Seoul, pillaging towns and what not. Even though Wang Kon had a ton of Artly and SAM Inf, he didn't venture out to kill my pathetic pillagers! That certainly wasn't a good AI decision....

 

[Uncle_Joe]

From looking at the minimap it appears that Mehmed and Isabella are best-buddies. He's basically put nearly his ENTIRE army where he feels the most hostility. I guess he has about 10 cities, so that's only 5 units per city (plus 1-2 normal garrisons) so maybe 7 units per city.

Funny you should mention that. Mehmed and Izzy WERE best-buddies for most of the game. But being that back-stabbing biatch that she is, Izzy eventually DoW'ed him and invaded with a HUGE army. Mehmed had the highest 'power', but she certainly did a number on him. He lost a few cities pretty early on, but eventually took some back when that stack moved out. I guess thats what prompted me to wonder if perhaps that stack would have been better off spread out a bit rather than piled up on the frontier between me and him (given that I was no threat at all...in near last place and constantly embroiled on the other front with the berserk Mongols).

But what I really enjoyed about that was that Izzy back-stabbed someone ELSE for a change. And it was interesting that Mehmed was both higher in military, friendly with her, and cruising to the win. He ended up losing to the Mongols (who won on score), almost undoubtedly as a result of Isabella's attack.

 

[Arlborn]

But what I really enjoyed about that was that Izzy back-stabbed someone ELSE for a change. And it was interesting that Mehmed was both higher in military, friendly with her, and cruising to the win. He ended up losing to the Mongols (who won on score), almost undoubtedly as a result of Isabella's attack.

THE MONGOLS WON? WOW...

Was it Genghis?

That by itself is an improviment for me..

 

[Uncle_Joe]

THE MONGOLS WON? WOW...

Was it Genghis?

Ayup. And this is despite me nuking the snot out of them as quickly as I could build ICBMs. I think I must have fired at least 6-8 at him over the course of our final wars.

I knew I was out of that game fairly early on. The Mongol kept attacking me about every 15-20 turns without fail and that just sapped my strength way too much. I fought the Chinese once or twice and the Germans (who were otherwise friendly) were vassals to the Mongols so I fought them all the time too. Mehmed pretty much kept to himself and exploited the large land area he had been dealt and I thought he was likely to win when Izzy made that attack. She got close to winning on points, but I think she suffered too much in her attack on Mehmed. Still, it was far better odds of her to win that way than simply sitting back and letting Mehmed cruise to victory.

But yep, good ole' Genghis won that game, mostly on land area methinx since his tech was pretty backwards compared to the Chinese and the Ottos. If there had been no time limit, the Chinese were ahead in the Space Race and would have eventually taken the win (assuming nothing unfortunate happened to them).

 

[Arlborn]

Ayup. And this is despite me nuking the snot out of them as quickly as I could build ICBMs. I think I must have fired at least 6-8 at him over the course of our final wars.

I knew I was out of that game fairly early on. The Mongol kept attacking me about every 15-20 turns without fail and that just sapped my strength way too much. I fought the Chinese once or twice and the Germans (who were otherwise friendly) were vassals to the Mongols so I fought them all the time too. Mehmed pretty much kept to himself and exploited the large land area he had been dealt and I thought he was likely to win when Izzy made that attack. She got close to winning on points, but I think she suffered too much in her attack on Mehmed. Still, it was far better odds of her to win that way than simply sitting back and letting Mehmed cruise to victory.

But yep, good ole' Genghis won that game, mostly on land area methinx since his tech was pretty backwards compared to the Chinese and the Ottos. If there had been no time limit, the Chinese were ahead in the Space Race and would have eventually taken the win (assuming nothing unfortunate happened to them).

Must have been a hell of a game!

I just met the others AIs in my game, but cant play much heh! Really anxious hehehe.

 

[Dagta]

Playing with the Jan 30th build:

I've found that every game the AI attacks me first and early in the game before I really have a good means to defend myself. The first game it happened I was extremely surprised as I knew I was second in military power. It seems to not matter if I am weak or strong, far away or close, with a religion or without, the strongest AI attacks me first, in the BC era. Possibly the only way for me to prevent this is to build nothing but military units for the first 4000 years.

Anyone else seeing this? I was playing as Ghandhi in all of my games since the 1/30 release.

 

[Dagta]

I just tried a game without using the Better AI mod and the same thing happened again. Perhaps it relates to how I play as Gandhi. The AI just hates me when I'm him.

 

[Wodan]

Yes, I was going to say something like that. To the effect of, "I haven't been experiencing that... maybe it's just you."

Wodan

 

[Uncle_Joe]

EVERYBODY hates Gandhi!

 

[Tiberias]

Not sure why they would intrinsically hate Ghandi . . . For what it's worth, I've been getting attacked right off the bat unless I make Archery a priority and get some defenders out. Otherwise, it's virtually guaranteed that I'll have Axemen on my doorstep before I can say "they have Bronze-working?" Also, certain leaders are virtually guaranteed to get things going early, especially Ragnar, Montezuma, and Genghis. Gettting an early religion and opening up borders with neighbors also seems to help--I think they're slightly less likely to attack if you can get Open Borders early, so Writing is a fairly important tech.

 

[Xira]

I think the most useful thing the better AI guy can do, other than calling everyone wusses for not going into a pure warmonger stragety . is to define for himself exactly what he wants the "AI" to do. Win 100% of the time on emperor? Immortal? Warlord?

Might I suggest you start then to employ the tactics of big blue, that of looking a certain number of turns ahead? After all this is very close to a perfect information game. You can define what is known each turn and plot a course to achive maximum victory over the largest # of possible moves.

What is it you want your better AI to do? Provide more of a challange for a human than Immortal? Just make a new diff slider adjustment then... Provide more of a human-like opponet? They still act live civs, not like people playing a game.

Ultimatly you have to decide at what point you are willing to sacrafice the "Civ-ness" of the AI for raw efficiency.(more lateR)

 

[Xira]

In the ideal world for me, it would be possible to win by space race on Deity, if the AI's spend enough resources fighting and (somehow) you manage to stay out of the crosshairs. I have absolutely no problem at all with tech stagnation caused by large-scale conflicts, it gives builders the chance to leap ahead.

I think the point some of the others are trying to make is that the current AI implementation goes too far in the other direction. Instead of making war cost 'just right' you made war cost 'way too much'. Being on either side of it.

If a single average war means you lose so much tech that you can't catch up and some other civ comes along with upgraded units and crushes you then ALL war is too expensive.

If ALL war is too expensive then the "one true path" becomes a diplomacy/cultural win. All you have succeeded in doing is switching the one true path because you tiled the meter too far in the other direction vis-a-vis war.

That is one of my goals, war should be hell for those waging it and for those it is being waged on. I'll ease up on that for the normal AI setting, but under Aggressive AI, WAR IS HELL.

IMHO if you are trying to prove a historical point there's just no helping you.

 

[Xira]

The game is decided too early, the rest either being slow stagnation to death or a cruise for victory. There's no balance, no competitive endgame.

IMHO 100% agree. Early game is too important in the entire Civ series. 50-90% of the techs, choices, land, and diplo is pointless, depending on skill level of participants.

You want a real challange Blake? Try making the endgame important.

 

[Xira]

To me thats the point, if we want the AI to be better it has to do the unexpected.

Although Ghandi etc are meant to be more peaceful and lean towards building / culture. They need to be able to go to war if the opportunity is right.

This is what the random personalities setting should be for.

 

[kettyo]

If ALL war is too expensive then the "one true path" becomes a diplomacy/cultural win. All you have succeeded in doing is switching the one true path because you tiled the meter too far in the other direction vis-a-vis war.

War is still the fastest way to win.

 

[kettyo]

Ultimatly you have to decide at what point you are willing to sacrafice the "Civ-ness" of the AI for raw efficiency.(more lateR)

I don't understand.
If you meant some simulation of emotions is also a necessity beside effectiveness i agree with that.

 

[kettyo]

EVERYBODY hates Gandhi!

Yes with basepeaceweight of 10 it's very likely that most will

 

[phungus420]

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"