A Better AI.

[skankerm]

One semi bug, Ive noticed that when i have leave old improvements checked my workers are flooding my watermills to farms on occasion. there is no need for an irrigation chain for a special resource its just overiding it. and i think sometimes it would override my mines and make windmills. I can't remember exactly what the case was i just know it does it sometimes.

 

[Blake]

Any saves of that would be useful, thanks.

 

[Tiberias]

Thanks for the detailed response on the civ distances. My overall impression of the mod is that it's showing a lot of promise, even if there are still some bugs. The underlying base by way of economy/tech is relatively solid, just a matter of tweaking some of the military algorithms to make for a truly dangerous AI opponent. I think we all think it's getting close, that's why there are so many comments .

 

[vra379971]

I've noticed something in some of my games...

You pillage a Town that takes forever and a half to build to nothing, and you get 100 gold. You pillage a farm that gives you optimal output from the get go...and you get 7 gold. Max. Often less.

Often when I drop back to defend versus an AI stack, I'd LIKE to defend my Towns...but I often can't. Toe to toe, I sometimes can't afford to NOT have some kind of defensive bonus.

So...my reccomendation...

If say...a Longbowman has City Garrison 3, and is...

...standing in a Cottage - Let City Garrision 1 ONLY apply.
...standing in a Village - Let City Garrision 1 and 2 apply.
...standing in a Town - Let City Garrision 1, 2 and 3 apply.

Either way, I'd like to see Cottage/Village/Town get at the very least, a 5-10% permanent defence bonus which will make these cash/production cows more pratical to defend under pressure.

 

[phungus420]

I've noticed something in some of my games...

You pillage a Town that takes forever and a half to build to nothing, and you get 100 gold. You pillage a farm that gives you optimal output from the get go...and you get 7 gold. Max. Often less.

Often when I drop back to defend versus an AI stack, I'd LIKE to defend my Towns...but I often can't. Toe to toe, I sometimes can't afford to NOT have some kind of defensive bonus.

So...my reccomendation...

If say...a Longbowman has City Garrison 3, and is...

...standing in a Cottage - Let City Garrision 1 ONLY apply.
...standing in a Village - Let City Garrision 1 and 2 apply.
...standing in a Town - Let City Garrision 1, 2 and 3 apply.

Either way, I'd like to see Cottage/Village/Town get at the very least, a 5-10% permanent defence bonus which will make these cash/production cows more pratical to defend under pressure.

This is not within the scope of this project. This project only deals with AI, not game mechanics. Though you do bring up a good point with the 100 Gold return on Town pillaging. It is very worth it for an AI or a player (lord knows I did it all the time in MP when I used to play) to rampage around with defensive stacks pillaging towns and villages just to make money and run my foe's economy into the gutter so I could hit them later.

 

[vra379971]

This is not within the scope of this project. This project only deals with AI, not game mechanics.

Oi. I appeared to have missed a paragraph. The reason I'm posting this here is because the AI (as far as I have seen) can't seem to grasp the importance of pillaging Towns over Farms - when they are doing it and when I'm doing it.

Which is why I'm saying providing cottaged tiles a value might improve AI competitiveness as they would be more apt to defend it.

 

[WuphonsReach]

Gotcha on the cultural pressure causing revolts but not flips...

...

So, in my game, Monty is going for a cultural win (and he's only 100 turns from winning). He has 3 cities around 100k (goal is 150k) and all 3 are in the 450-675/turn range. One city is 53 turns away from turning legendary the other is 96 turns and the third was somewhere between.

Sounds like he's got a great plan right?

Muwahahahaha...

1) He doesn't protect the coast. He had (2) clam resources plus he was working 3 other sea tiles. That made it super easy for me to roll up with 2 destroyers, kill his clam boats, and camp the coastline. Which plunges his city into starvation mode and cuts cultural production by about 25% within a few turns.

2) Even though my troops have been dug in on this beach for 20+ turns (7 marines, 5 infantry), he's never moved troops to the city. Nor attacked my stack or attempted to dislodge me. This is his capital that I'm about to take.

3) My transports have been undisturbed while ferrying troops. I'm headed back for my 3rd trip (at around 12 turns round-trip), ferrying troops to this beachhead. That convoy has 1 battleship, 5 destroyers and 4 transports.

4) The troop strength in the city is 1 machine gun, 6 rifleman. And nothing else. There are other cities within a few turns of movement (age of railroads) with half a dozen troops. He could've easily bumped that defender count up to 2 dozen troops plus half a dozen suicidal siege weapons and driven me off the beach.

A few other notes:

- This is warlord difficulty. Huge, fractal, marathon.
- My production is about 3x more then his (I have assembly lines and have upgraded production across most of my empire in the last 200 turns). Back 200 turns ago, it was much closer. 1476 vs 545.
- His crop yield is half of mine, but that used to be almost even. 1470 vs 792. Back when it was even, I started going after the weaker of the 2 continents and have wiped out 2.5 civs on it. Monty, OTOH, has been stalemated.
- My (tiny) vassal on the continent with him has been at war with him for about 100 turns now (maybe 150). Even though my vassal has lost half his power, Monty still hasn't taken any cities. My vassal has a tiny tech edge because of techs that I gifted him.
- Monty's culture graph started accellerating like crazy about 200 turns ago. You can see that he's used 3 great artists as culture bombs in his primary cities. His overall culture is easily 6x mine.
- Monty's GNP is 1115 mine is only 725
- Land area is 490,000 (Monty) vs 630,000 (me)
- He has 31 cities. 20 of them are size 10 or larger. 13 cities lie on the coast.
- At one point he had 2 vassals.
- My german vassal on the continent has only 6 cities (sized 2 to 12). I'm amazed that my vassal still exists.
- He has about 27 strategic resources, 25 luxury, 29 food resources (11 are seafood).

So... cultural victory condition definitely works (yay Blake!). It's very noticeable on the culture graph. But it's going to be a short-lived attempt. I've got 24 troops on the beach already and another dozen will be arriving in another 10 turns.

My empire has 49 cities. Research slider is at 60% w/ 85/turn profit. My navy consists of 63 destroyers, 17 subs, 11 battleships and 5 transports. Ground troops are 86 infantry, 34 marines, 9 SAMs, 18 MGs, 10 arty, 9 calvary.

Bit of a shame that I'm going to wipe out Monty's cultural centers. Seems a bit odd to see Monty go for a cultural victory instead of trying to dominate the world.

 

[Wodan]

Thanks for the detailed response on the civ distances. My overall impression of the mod is that it's showing a lot of promise, even if there are still some bugs. The underlying base by way of economy/tech is relatively solid, just a matter of tweaking some of the military algorithms to make for a truly dangerous AI opponent. I think we all think it's getting close, that's why there are so many comments .

100% agreed.

 

[vra379971]

Yes, it is getting close. Though, when it does happen, I'm unsure how many of us will be celebrating

 

[HardCoder]

Thanks for the detailed response on the civ distances. My overall impression of the mod is that it's showing a lot of promise, even if there are still some bugs. The underlying base by way of economy/tech is relatively solid, just a matter of tweaking some of the military algorithms to make for a truly dangerous AI opponent. I think we all think it's getting close, that's why there are so many comments .

No, two or three people could work on this for a year and there would be arbitrarily more work left than was done.

The big gains from the current work are just making the AI waste less of its resources and taking advantage of typical mistakes that players make. That will probably make it impossible to win at the highest levels in the future, or perhaps even now. But as for doing things that are clever, or even implementing things that are fairly clearcut but not trivial (like finding sites for specialized cities and building them), that's a whole different ballgame.

The one thing I'm concerned about is that when the AI becomes more efficient at closing loopholes in the game, the winning style of play may become formulaic, like it is at the highest difficulty levels.

Making diplomacy more nuanced and balanced, and giving different leaders truly different personalities, might help with that. Perhaps, for example, if you seem to be pursuing a builder strategy, other builder leaders might tend to favor you considerably more than they currently do, and help you considerably against the "warmongers," as long as you don't attack one of them. Likewise if you have the same religion as a deeply religious leader - the leader should prefer you greatly to leaders following other religions, and somewhat to leaders who have no religion or aren't religious.

The diplomacy modifiers themselves are not particularly balanced. The permanent -6 for declaring war is silly. The lack of modifiers for behavior like "you attacked your friend(s)" (not "my friend") is also frustrating. Other AIs should become suspicious of Napoleon when they observe that he attacked someone he was friendly with.

Vassals who have converted to your religion should be much less likely to return to their "own" than they currently are.

Vassals who are cautious or better, or at least pleased or better, should also be extremely reluctant to trade away important techs. I often don't give important techs to my vassals because I worry that some competitor will wind up with them.

 

[kettyo]

There are so many areas to develop i think it could help to give our thoughts about the priorities in AI development.
AFAIK right now there are only Blake and Iustus working on this project (quite hard) so it's probably better to focus on just some areas and if they seem ok then move forward.

My priorities are:
-combat AI
-warmaking logic (more of a decision and less random)
-AI going for domination
-others

 

[Todd Hawks]

My priority is a mostly bug-free version. Still waiting for that one. Maybe you should stop introducing new features for a month or so and focus on making the stuff you added so far work as intended.
Don't get me wrong, I like what you are doing with the AI but the quality of the releases could be better.

 

[jan33]

I think the changes to the AI so far only make aggressive warmongering even more essential. Not good in promoting different ways to play the game.

nah u can still live ur building life but got to keep an eye on ur military,in my eyes warmongers are often loosing if they go a bit to far...

 

[Wodan]

My priority is a mostly bug-free version. Still waiting for that one. Maybe you should stop introducing new features for a month or so and focus on making the stuff you added so far work as intended.
Don't get me wrong, I like what you are doing with the AI but the quality of the releases could be better.

They've already said that is exactly what they're doing. It sounds like you're thinking these interim releases are final versions... they're not. The way I see it is this is a beta test, open to those who are interested and willing to devote their time and to provide feedback. Along with that territory comes some amount of inconvenience to the beta testers, dealing with bugs and things not quite perfect. That's precisely the point of a beta test.

Anyone who feels the inconvenience of beta testing is not worth it (to them) may of course simply wait for the release. We're still in v.0.8 or something. Wait for v.1.0.

Wodan

 

[Pfeffersack]

They've already said that is exactly what they're doing. It sounds like you're thinking these interim releases are final versions... they're not. The way I see it is this is a beta test, open to those who are interested and willing to devote their time and to provide feedback. Along with that territory comes some amount of inconvenience to the beta testers, dealing with bugs and things not quite perfect. That's precisely the point of a beta test.

Anyone who feels the inconvenience of beta testing is not worth it (to them) may of course simply wait for the release. We're still in v.0.8 or something. Wait for v.1.0.

Wodan

Exactly.And the sooner new features (here in fact improvements in different areas of the AI) get implemented, the more time is to make them work correct with the help of the many people who use and test this modification.I don't think the final result will be better if new things have to wait until other issues are "perfect" (aside from the fact that adding new things often makes balancing old stuff again necessary) or if Blakes team has to do the testing partly by themselves.

Some of the builds might have made the AI worse in certain aspects or situations temporarily (compared to older builds or sometimes even stock AI), but overall -at least for my style of playing- each build is still a clear improvement over the stock AI and I prefer(ed) playing with this AI all the time.

 

[player1 fanatic]

My priority is a mostly bug-free version. Still waiting for that one. Maybe you should stop introducing new features for a month or so and focus on making the stuff you added so far work as intended.
Don't get me wrong, I like what you are doing with the AI but the quality of the releases could be better.

Agreed,
There will alway be new ideas for implementation. But every new idea needs a week or two to get it right, so it comes to endless loop bugfixing old ideas adding new ideas, bugfixing...

Best to cut at one point, and do only bugfixes and introduce experimental relase as seperate download.

 

[phungus420]

My priority is a mostly bug-free version. Still waiting for that one. Maybe you should stop introducing new features for a month or so and focus on making the stuff you added so far work as intended.
Don't get me wrong, I like what you are doing with the AI but the quality of the releases could be better.

Most of what the AI team (blake and Iustus) have added seems to work well. The AI does in deed whip quite effectively, for economics, and also on defense. The AI will whip up longbows like crazy if a city comes under threat. The New build governor actually works, and I find very few bugs now. Mainly all the issues are with the warmaking aspects. I see these not really as bugs, but more like things that can be improved. It'd be awesome to see an AI actually get a domination victory. Lord knows I didn't expect them to code a cultural victory, and that's in the game now for those of us with this mod. That adds a lot more depth, then the original 1 dimensional Spaceship default AI.

With Warmongering, Kettyo pretty much hit on the main points. If the AI ever stands a chance of growing a third head, and achieving Domination victories then it's warmaking decisions need to improve. (Which Blake said warring across the map has been stabbed through the heart hopefully )

In order the AI needs help with:

1)Warmaking decisions (who to fight, and when)
a) The AI picks some of the weirdest rivals to go after. Why troll around across the map, when you hate the guy next door?
b) When a warmonger like Genghis gets rolling (conquers or vassilates a rival) he should also start becoming even more agressive, so that like the cultural victory, the AI starts on a path toward victory if it "sees" that it is possible.

2)More logical unit actions.
a) Like pillaging Strategic Resources (when apropriate). Or pillaging for Cash.
b) Be more willing to move the stack after taking a city. Watching a Dagger stack that could roll over a rival sit in a recently aquired city for 10 turns so the rival can build his defenses is heartbreaking.
c) Use units in the field more (Especially Seige Engines). I've often seen the AI with a decent stack that could decimate a rivals, just sit still, even though with a couple suicide cats the stack would be so soft the rest of the units would get great XP mopping up, and would also lessen the rivals warring capacity. This is particularly obvious against Barbs, when a City has 1 Archer, a Spear and a Chariot in the city , but for some reason wol't send the chariot to kill the lone axe that's taring up all the workers hard work, and sending the civs economy back litterally 100s of turns sometimes.

That's pretty much all I see right now. Honestly, the AI does pretty dang well at rushing units when apropriate, and even appears to start harrasment wars early on. It just doesn't know how to do harrasment wars all that well yet, but that's what this mod is all about. It's the coolest thing going in civ IV right now. And that's why we're on page 60 still giving insight and checking for more updates. Most of us really hope to eventually see a rival AI that can run the gambit for victories against us, and we hope to get busted down a difficulty level in the process.

 

[Quagga]

My priority is a mostly bug-free version. Still waiting for that one.

I'm still waiting for a bug-free version of Windows... 13 years and counting. I'd rather play the Better AI now than wait for the perfected version.

 

[HardCoder]

My priority is a mostly bug-free version. Still waiting for that one. Maybe you should stop introducing new features for a month or so and focus on making the stuff you added so far work as intended.
Don't get me wrong, I like what you are doing with the AI but the quality of the releases could be better.

I don't think that's really a major issue, and as a lifelong software developer, I think you have to take it or leave it. I haven't seen quality problems in the releases, which are, basically, glorified test builds of an experiment in progress. This is not "the game," as in, what you get out of the box. Should this be integrated into and released with a future version of Civilization 4, the QA would take months after feature freeze.

The project is there for you to "play" with, which is appropriate when you consider that you are "playing" a game. It's a bummer if it crashes or units behave weirdly, but it's not as if that sort of thing is unknown in the off the shelf version of the game. But more to the point, if bugs in a game bother you that much, you have set the bar in your life pretty low ....

 

[HardCoder]

Hey, has there been some change in the barb behavior? I've seen some spectacular rushes when the barb behavior changes from "stumbling around in the woods" to "rushing headlong at my cities and improvements." Like 8 or 10 barbs coming from all directions at once, on Prince level in ordinary playing conditions (no raging barbs). I don't remember it being like this in previous builds. There's this sudden huge wave of barbs and then they go away completely.

It's actually pretty difficult to defend against. I'd rather have a stack of 8-10 units coming from Shaka in one direction than the same number of axes and warriors coming from every which way not caring whether they live or die.