A Better AI.

[Zoolooman]

What wont be changing is the AI mounting a proper defense. In 2.08 the AI's are VERY much pinata's - bursting with goodies to be liberated with even trivial military investment. So while the unit spam will be toned down, the Pinata's will not be coming back.

I'm not asking for piñatas.

Don't discount what I'm about to say. I know a wrote a long and loud post. Ignore it. Read this one instead.

I'm asking for a different design philosophy. There are many kinds of AI. Some AIs play to not lose. Others play to obstruct or harass the human player. A few very good AIs play to win. These are all great goals for particular types of games. I don't think any of them is appropriate for Civilization.

AI civilizations should play to be interesting components in the game. What do I mean? I mean that the AI should seek to provide situations with interesting, challenging, and non-trivial decisions for the player.

Defender-spamming is playing to win, playing to not lose, and playing to obstruct a common human strategy. Defender-spamming does not create choices. It limits them. It makes conquest/domination victories tedious and boring, it forces the human to participate in the unit-spam game, and it blatantly reduces the military options for small civilizations. Besides that, unit spam in general causes performance issues in the Modern Era.

I recommended a unit cap because it is an easy and effective solution. 7 to 8 defenders per a city is far more than the AI ever produced with the default code; in general, they never produced more than 20 units in an offensive stack either. Limited stacks are also manageable and affordable for the AI, even under the Pacifism civic. If you can find a way to convince the AI to limit itself naturally instead of artificially, then that'd be best. But until then, I offer my opinions and solutions honestly and openly. I'm here to have fun, and I have more fun playing (and sometimes losing to) the regular AI on Emperor than I have playing (and always peacefully defeating) the defender-spamming AI on Prince.

 

[HardCoder]

The governor now kicks ass. I can turn it on and with very few exceptions it does a fine job, especially once I know how it works. Unless I need to avoid its zealous pursuit of growth it almost always does a better job automagically than I would do without significant study.

The governor is the single most important reason I'd hate to go back to 2.08.

The least awesome things about BetterAI

Too many units = too much tedium. Why not figure out how to have the AI attack and defend with fewer upgraded units than more old units. As other posters have said, moving large stacks of units around the map is not really the way the user interface was set up. You can say that "that's how it is at deity" but, really, who cares, except the people who like the glacial pace of deity games with all the attendant micromanagement. I'm all for satisfying that itch for other players, but please satisfy mine for a more strategic orientation also.

Weird resource distribution. Starting locations can be awesome while every other spot in the neighborhood is a bunch of dry yellow tiles or one of those special Warlords-style Gobi deserts. I don't know if enhancing the starting locations is contributing to deficits elsewhere in the maps, but I think something is out of whack. Capturing an civ's capital is like capturing 4 normal cities because of all the goodies around it. Because of the limits on coastal resources, and the consequent lack of "seafood madness," coastal locations don't have the utility that they used to. Maybe I am just imagining this change - but I used to be able to build high production port cities, and now they seem to be just "ehh." I rarely build a Heroic Epic + West Point on the coast anymore.

 

[Elandal]

The build governor is for me the biggest thing as well. It allows me to concentrate on bigger matters than managing some insignificant cities - I still handle key cities myself of course, whether for better or worse (sometimes I think the governor would do better than I do).

The large numbers of units is a harder one. First, I like the fact that WW is a big issue. It discourages the rampage somewhat. However, the UI simply can't cope with the large stacks needed now - just selecting a single unit from the stack by clicking it from the plotlist takes a few seconds. Also, building that big stacks takes a lot of time, which means that the stacks will have shorter useful lifespan (before obsolence).

 

[Arlborn]

So many complaining about too much defenders, OK there IS too much defenders(as Blake said), and OK its default number for all AIs and get the personality thing off(Blake said he will try to fixx it). But I think that the biggest problem(for me) is when an AI build a huge offensive stack and JUST DONT ATTACK! Its lovely to see it atacking, gives such a challange to the game, specially when you think you already won the game and then, poff! 30 units around your cities! But when they change their mind and dont use the stack, they just leave it stting around comsuming money and DONT even use it in a good way to defend cities because in the AI's mind it is suppose to attack not defend(its true, right?).
I want the AI to ACTUALLY disband in mass! If you changed your idea and wont go to war anymore, what you do with all those city riders?? 30 units or so sucking your money! I disband most of them, specially all without promotions! But the AI just WONT do it! ANd then they sit there with a stack of useless units sucking their money for a LONG time in the game, even when they are already obsolete(specially with the new handicap, less money for update.


Now, I dont know exactely if and how the AI disband units, but for my impression they just dont do it near enough when they got a attack stack that wont attack. I already saw AIs mounthing huge stacks for the whole game and never getting in war..


Just my opnion, maybe Blake can tell me more about it?

 

[Elhoim]

Since most people agree that the governor is working great, and the combat part of the project is kinda controversial, would it be possible to make the mod into modules?

Not right now, but it would be nice to have the option for:

- City Governor and Worker AI Module.
- Combat AI Module.
- Both Modules.

Thanks!

 

[jkp1187]

Does it happen? Yes it does. But my sense of it is that the AI is struggling more now when contending with larger number of units before than it did previously (at least when not using Aggressive AI).

You know, I wonder how Aggressive-AI works on Noble with Better AI. Has anyone out there been testing this on Noble?

I have not been playing with it and, so far, I have NOT noticed any significant defender spam going on in AI cities (except in the capitals, which always seem to have ~10-15 units.) I have not had much trouble keeping ahead of the AI in terms of tech lately, but then until my first game after 2.08, I never really had trouble with that on Noble (and, honestly, I have had some very lucky starting positions in my last couple of games.)

 

[Iustus]

Had an 'Assert' pop up in my current game.

Here was the message:

File: .\Cvteam.cpp
Line: 1612
Expression: iCount>Setatwarcount(bIgnoreMinors)

Saves are attached.

Hope it helps.

I posted the same Assert a few pages back and there are a few save games with it.

I dont know if that will help you or not I believe it is the exact same error message.

If that was the post you were talking about (direct link to post), you did not actually include the save file in your post, although you said you did.

Do you still happen to have the save? I am still trying to get a reproducable case of this.

-Iustus

 

[Wodan]

Guys, here's a bug report.

I don't know if this is a problem with Better AI or with the game itself. I'm sure you can tell, though.

Stack with mixed Commando and non-Commando units uses the Commando "Go To" path algorithm, when it should use the more restrictive non-Commando algorithm.

I first noticed it when in an ally's territory and trying to swing around to the other side... the comp wanted me to go through enemy territory (taking 3 turns) rather than use my ally's roads to get around in only 1 turn.

Wodan

 

[Uncle_Joe]

If that was the post you were talking about (direct link to post), you did not actually include the save file in your post, although you said you did.

Do you still happen to have the save? I am still trying to get a reproducable case of this.

Hmmm, perhaps I didnt attach them correctly? No idea. But unfortunately, no, I dont have those saves anymore.

For future reference though, will saves still help you if I playing with modded graphics and handicaps file? Would you be able to load the game and get the info you'd need?

Sorry for the lack of a save...

 

[Iustus]

Hmmm, perhaps I didnt attach them correctly? No idea. But unfortunately, no, I dont have those saves anymore.

For future reference though, will saves still help you if I playing with modded graphics and handicaps file? Would you be able to load the game and get the info you'd need?

Sorry for the lack of a save...

if you are playing with a special mod, the best thing is to post the mod you are using as well, if possible. you can remove the betterAI dll from the file you upload, that should make it's size small enough. If it is too large, removing any purely cosmetic changes should be fine as well.

If you are playing with the changes in CustomAssets rather than loading as a mod, these things are slightly less necessary. Part of it comes down to reproducability. If I can run your save without your mod, and still reproduce the problem, then I do not need your mod. Whether or not I need your mod to reproduce it comes down to the specific issue, in this case, probably not, but in the future, who knows.

If you are playing as a mod, then I have to load the game with a mod of the same name. I do not believe I have to have exactly the same mod (just the same name) to get the save to load, but I may need it to reproduce whatever the bug is that you are reporting. It is infinitely easier to fix a bug if I can see it myself. It is analogous to a car mechanic trying to fix a problem if you just give him a sound recording of the noise your car made vs bringing the car in to his shop and letting him open the hood and actually see what is going on. Sure some problems can be solved by listening to the sound, but it is a whole lot harder to do, and more prone to wrong guesses.

-Iustus

 

[Iustus]

Guys, here's a bug report.

I don't know if this is a problem with Better AI or with the game itself. I'm sure you can tell, though.

Stack with mixed Commando and non-Commando units uses the Commando "Go To" path algorithm, when it should use the more restrictive non-Commando algorithm.

I first noticed it when in an ally's territory and trying to swing around to the other side... the comp wanted me to go through enemy territory (taking 3 turns) rather than use my ally's roads to get around in only 1 turn.

Wodan

Unfortunately, this is one of the few things that is in the game itself. The pathing code is all propriatary firaxis stuff, we have absolutely no control over it.

The one thing we can do is to either discourage or prevent the AI from grouping commando units with non-commando units, but other than that, there is nothing we can do. We give the pathing code a start and destination, and it tells us the path to take.

There might be a few cases where we could catch this (we can tell the pathing algorithm to path only outside enemy territory), but I believe this is a strict requirement. When we specify such, every plot must be outside enemy territory, so it would not find shorter paths that are part inside, part outside enemy territory. Such a fix would cause a slowdown every time you had a group with commando units, so I am reluctant to make this half-fix.

Please report this bug on the bug reports forum so Firaxis can look into fixing it in the future.

-Iustus

 

[Iustus]

What is a dogpile war?

The AI has three basic types of wars when it is the agressor:
(1) total war
(2) limited war
(3) dogpile war

(1) total war means the AI is more or less trying to take as many cities as possible, and going all out

(2) limited war means pillaging is more likely, less overall production will be going to the war effort, and also means the AI will charge less for peace. While still focused at taking cities, it is not so much the primary goal.

(3) dogpile war means the AI went to war when someone else was already at war, adding in to attack a weak opponent. In such a war, an AI will seek peace if the others at war also seek peace, not wanting to stay in a war without one's allies.

The final type of war is a defensive war (when someone else declared war on the AI).

The type of war affects both the production choices of an AI and also the price for peace.

-Iustus

 

[Iustus]

Since most people agree that the governor is working great, and the combat part of the project is kinda controversial, would it be possible to make the mod into modules?

Not right now, but it would be nice to have the option for:

- City Governor and Worker AI Module.
- Combat AI Module.
- Both Modules.

Thanks!

I understand what you are saying here, but I think the real goal is to just get everything tuned right. I think we can all agree that there are some problems with the current 'unit spam', but that is more on the order of a bug, rather than a design flaw.

Maintaining such a split would multiply the number of builds we have to support from two to six! It would be entirely possible for their to be bugs in one or the other, but not in all of them.

Such a division does not fit naturally into how the code is organized, and I am not even sure the division you describe is what you really want. Does the better explore code go with the combat mod? What about great merchant trade missions? These are both in the 'unit' code, which some might argue is your so called combat part.

Perhaps what you really mean is the code the AI uses to decide what to put into the build queue? That would be a bit easier to split off, but it would be a hassle to go back and add back in the old version of the function, and build and maintain and bugfix two different versions.

Blake is working on reducing the number of units that are built, using some new information I have made available to him (how much gold is needed to upgrade every unit, and what percentage of our total costs are unit costs).

I also suspect an old bug may have resurfaced where an AI never thinks it has enough defenders so it continually spams them. If a defender is built, but then is attached to an attack stack, it is no longer counted as a defender. The code that decides what to build does one check to see if a defender is needed, if so, it builds one. Then once its built, that unit checks to see if it needs to be a defender, if not, then it joins an attack stack. Because of the way these things are set up, its really easy to get out of sync here, and constantly build defenders, think you dont need them, set them off on other things, and build more defenders. If you are running chipotle, you can hold down the ctrl key and hover over a city to take a look at the situation. You are looking to count the number of groups on city defense.

Hopefully this should be resolved in the next build.

-Iustus

 

[Uncle_Joe]

Hopefully this should be resolved in the next build.

Which, of course, begs the question of when we can hope to see the next build.

I actually reverted back to 2.08 to play a few games 1) for comparison and 2) because quite frankly, the 1/16 build sucked all of the fun out of the game IMO (and no, not because I like to easily conquer the AIs either...).

 

[Roland Johansen]

Blake is working on reducing the number of units that are built, using some new information I have made available to him (how much gold is needed to upgrade every unit, and what percentage of our total costs are unit costs).

I also suspect an old bug may have resurfaced where an AI never thinks it has enough defenders so it continually spams them. If a defender is built, but then is attached to an attack stack, it is no longer counted as a defender. The code that decides what to build does one check to see if a defender is needed, if so, it builds one. Then once its built, that unit checks to see if it needs to be a defender, if not, then it joins an attack stack. Because of the way these things are set up, its really easy to get out of sync here, and constantly build defenders, think you dont need them, set them off on other things, and build more defenders. If you are running chipotle, you can hold down the ctrl key and hover over a city to take a look at the situation. You are looking to count the number of groups on city defense.

Hopefully this should be resolved in the next build.

-Iustus

I'm one of the guys who doesn't mind big AI defence stacks, but I do worry about the AI upgrade and upkeep costs. So I think it is a great idea to take those costs into account. I hope that a nutcase like Montezuma is allowed to spend a greater percentage of its income on unit upkeep as a peacelover like Ghandi. I personally like different leader styles (as long as none of them is really inferior or artificially stupid).

The bug you describe (if present in the code) would be nasty. I hope you can fix it (or that it isn't there at all).

In my present game, I did notice a large stack of longbowmen with city defence upgrades accompanying a huge attack stack of the AI. The total number of units used in the attack must have been something like 60-80 (split in several stacks). But there were about 20 city defence upgraded longbowmen units in the attack stacks and also about 10 pikemen (who are also better used defensively in general with a few to defend the attack stack). I'm happy to say that artillery (hwacha's and trebuchets) and macemen were the largest part of the stack.

I was playing an aggressive AI game, so maybe this was a total war strategy. It would explain the lack of defenders in its cities. Knowing that the AI (sometimes) uses such a total war strategy with a very limited number of defenders in its cities could lead to exploits by the human player (with very limited I mean 1-2 defenders in border cities).

 

[Xira]

The problem with making the AI even 'better' is that it further accentuates flaws and imbalances in the game.

Notably and most annoying for me: Warmongering is the best tragedy. It's annoying that at higher difficulty levels it is the ONLY way to win. You must warmonger, at least at first, or lose. Builder, much less perfectionist builder, has utterly no chance in this game after a certain difficulty level.

I suppose part of that IS the ai. It's hard to outbuild it. The nature of the hard difficulty 'advantages' the AI gets is also directly aimed at helping it outbuild you, rather than outfight you. Their crossbow men don't get stronger, thus making it a good idea to defend rather than attack. Rather, they get more things, thus makeing them have more crossbow men, thus making it impertive that you stop them from making them in the first place.

By giving the AI more things rather than better things you force players further and further into the only viable playstyle against an enemy that will outbuild you 100% of the time...That of warmonger.

I would personally like to see a rebalance patch that both improves AI and alters the NATURE of the bonuses the AI gets on the diffrent diff levels. I can already win on diety, I'd like to be able to use a diffrent stragety once in a while though.

 

[Uncle_Joe]

I would personally like to see a rebalance patch that both improves AI and alters the NATURE of the bonuses the AI gets on the diffrent diff levels.

There is a suggest 'handicaps' file on the Better AI site already. Its not 'official', but neither is the Better AI.

I personally use my own handicaps and it really isnt hard to custom tailor them to your own personal tastes.

 

[Harrier]

There is a suggest 'handicaps' file on the Better AI site already. Its not 'official', but neither is the Better AI.

I personally use my own handicaps and it really isnt hard to custom tailor them to your own personal tastes.

Then why do you not write a tutorial. Because it is hard if you do not know where to start.

 

[Uncle_Joe]

All the variables are posted on the Better AI site under 'Documentation'.

 

[kniteowl]

Question does the AI know how to properly use their UUs?

Like the French Musketeer Instead of just defending cities like normal muskets, their better off used in pillaging raids when in war.

or the Inca Quencha, When your near HC... You know he's gonna come with a SOD or his UUs and rush you, Same for the Aztecs Jaguars.

How about the Cho-ku-nu? Does the AI know that their UU Crossbowmen is capable of collateral Damage?