The key to AI greatness

[alexman]

Ever wonder why AI England is always laughable, while AI Egypt often dominates? Why Zululand rarely gets past the Middle Ages? Well, it’s more than just traits and UU. Ever wonder why the AI sometimes doesn’t have a marketplace in the industrial era, or why it doesn’t have factories well into the modern era? The answer is the build strategy that the AI follows in its cities.

Each civilization has a “build-often” and a “build-never” list. You can find this list under the civilizations tab in the editor. The build-never list is always empty, but the build-often list is as follows:

Civilization  Off Def Arty Setl Work Nav Air Grth Prod Hap Sci Wlth Trad Expl Cult
----------------------------------------------------------------------------------
Rome (4)       X   X                          X    X
Egypt (3)                                     X    X                           X
Greece (3)                            X                     X        X         X
Babylon (4)                                                 X                  X
Germany (5)    X                                            X                  X
Russia (4)                                X   X             X
China (2)                                     X    X        X
America (3)                               X   X    X                           X
Japan (4)      X                      X                 X
France (1)                                              X            X         X
India (1)                                     X                 X    X         X
Persia (4)     X                                                X    X
Aztecs (4)     X                                        X
Zulu (5)       X
Iroquois (2)                                  X         X                 X
England (3)                           X                         X    X         X

Off: Offensive Units
Def: Defensive Units
Arty: Ground Artillery
Setl: Settlers
Work: Workers
Nav: Naval Units
Air: Air Units
Grth: Growth Improvements
Prod: Production Improvements
Hap: Happiness Improvements
Sci: Science Improvements
Wlth: Wealth!
Trad: Trade Improvements
Expl: Exploration Units
Cult: Culture Improvements

X = build often

The number in brackets after the civilization name is its aggression level.

Looking at this list, it’s easy to explain why each AI behaves the way it does in the game. For example, England not only has bad traits and UU, but it also has Wealth and Naval Units in its build-often list. Every human player knows that a Navy is not really worth building on most maps, and that you should never build Wealth, even when you have nothing else to build! (It’s much more efficient to build a unit in a core city and disband it in a corrupt city than having the core city produce Wealth and rush-building in the corrupt city). Another example: Rome and Zululand produce units like crazy, at the expense of any kind of cultural improvements. They sometimes do well at the beginning of the game, but never last. Also, why don’t expansionist AI build scouts? Well, none of them except the Iroquois have the Explore Units in their build-often list. Why does France outperform their traits and UU? They are one of the few AI to have trade (marketplaces!) on their list, and don’t waste time building extra units.

Keep in mind that this list doesn’t mean that the AI doesn’t build other things. It builds other things, just at a lower priority.

I think the game could present a better challenge if the AI had build priorities closer to what we humans have found optimal. These build priorities are, of course, trait-dependent. Here’s a list of build priorities and aggression levels that I propose to make the AI better:

Civilization  Off Def Arty Setl Work Nav Air Grth Prod Hap Sci Wlth Trad Expl Cult
----------------------------------------------------------------------------------
Rome (4)       X                                   X            O    X         X
Egypt (3)                                      X   X    X       O    X         
Greece (2)                                         X        X   O    X         X
Babylon (1)                                    X   X    X   X   O    X         
Germany (5)    X                                   X        X   O    X         X
Russia (3)                                         X        X   O    X    X    
China (4)      X                               X   X            O    X         X
America (3)                               X    X   X            O    X    X    X
Japan (4)      X                                   X    X       O    X         
France (1)                                     X   X            O    X         X
India (1)                                      X   X    X       O    X         
Persia (4)     X                               X   X        X   O    X         
Aztecs (4)     X                                   X    X       O    X         
Zulu (5)       X                                   X            O    X    X    X
Iroquois (4)   X                                   X    X       O    X    X    
England (2)                           X            X            O    X    X    X

X = build often
O = build never

These are the build priorities and aggression that I would choose if I were playing those civs. I am not the best civ3 player in the world, so feel free to make suggestions on how to change these.

My philosophy was:

• Trade is a top priority for everyone. Everyone should build marketplaces ASAP.
• Some form of cultural build is a priority for everyone. Scientific civs get scientific buildings, Religious get happiness buildings, and everyone else gets generic the cultural flag checked.
• Nobody ever builds wealth.
• Production (factories) is a top priority for everyone. Subsequent builds go faster after factories.
• Industrious and builder civs get growth (granaries) as a priority.
• Militaristic and good offensive UU civs (Persia, Iroquois) get offensive units as a priority, and have higher aggression.
• Builder civs have lower aggression.
• Expansionist civs make exploration units a priority.

[Edit: removed useless bic file]

[DeepO]

Alexman, I don't know if I agree. Sure, sometimes I too find it annoying that certain civs do not produce well, don't trade well, or build loads of navy when there is no sea around. But, that also gives a wider spread in the game... the English will be better on archipelago maps then the zulu will, because they build more navy. In other maps the Zulu will be better, as they can take out large parts of the world before the others have a chance of bracing themselves.

Maybe making production a must build for all is a good idea if you look at it from one point of view, but in a mixed AI-map, having the difference will mean that certain civs will get a much bigger boost, and therefore will get more advantage then others. Just by destroying ancient civs (like the Romans or Zulu) they grow big, and menacing. If each civ would have the same production tastes, it would mean that all would be relatively equal in power, and therefore all keep about the same size. I think that in order to get better AI resistance, certain civs should not be given late game traits like trade or production.

Even the wealth option has some advantage: it will mean that certain civs will have more cash, cash they can spend on upgrades(!), techs, or making deals. Because this will always coincide with more trade buildings, it can become very powerful in comparison to other AIs. Putting all wealth to 'never build' seems a very poor decision IMHO, sometimes I use it too (never to rush buildings in corrupt cities, that is always done with spare cash floating around). I'm sure the AI will find some use for it as well.

No thanks, I appreciate the effort, but I think that having such diverse AIs will make it more enjoyable than the blendness you are proposing. Certainly if this means that certain AIs will belong to certain eras (the Zulu are ancient not industrial, while the Americans are exactly the opposite). I think better results will come from chosing the AIs carefully when playing, and so far the scenarios I've seen do that excellently: modern wars mean industrial civs, ancient wars mean ancient civs.

DeepO

[edited: too many typos]

[lockstep]

Originally posted by alexman
Every human player knows that a Navy is not really worth building on most maps, and that you should never build Wealth, even when you have nothing else to build! (It’s much more efficient to build a unit in a core city and disband it in a corrupt city than having the core city produce Wealth and rush-building in the corrupt city).

I'm fairly sure that the AI's (who LOVE to keep their obsolete units around) don't use the 'build unit and disband it elsewhere' strategy. So 'building' wealth sometimes may be a good idea for them.

Otherwise, your changes sound pretty interesting.

[jshelr]

Very good work. Explains a lot. And the proposal might just represent a material improvement, although you have to think this is so fundamental that it could require lots of effort to balance it right. Thanks

[DeepO]

No... looking at it again, the only improvement I see you are proposing is giving the Persians a scientific need, instead of the Chinese. And even that is debatable: the Persians were, irl, not very scientific minded, while the Chinese were.

Maybe some shifting between the Culture things can happen, but I guess this was also thought out with respect to culturally linked starting positions. And the defense of the romans is most likely to get them to build even more legions, removing that would probably mean less attacks (how strange it might sound).

No, going over the list again, I don't see any need for changes. What should be changed (but I heard this was in the 1.29f patch) that all civs would like building improvements a bit more, which means that civs with production flagged would only build even more then that. I don't know how much of this is true, I have yet to play a 1.29 game into the industrial age...

DeepO

[Soren Johnson]

A couple notes:

Trade = harbors/marketplaces/airports
Wealth = marketplaces/banks/workers, not just "Wealth"

Secondly, I would recommend rarely using the Never flags. If you test it a lot and it seems to work, maybe, but otherwise be careful...

[DeepO]

Thanks for the explanation, Soren. Did the general tendency for building factories went up in the latest patch, or did I dream that on up?

DeepO

[FrustratedPoet]

Originally posted by Soren Johnson Firaxis
A couple notes:

Trade = harbors/marketplaces/airports
Wealth = marketplaces/banks/workers, not just "Wealth"

Secondly, I would recommend rarely using the Never flags. If you test it a lot and it seems to work, maybe, but otherwise be careful...

Thanks for that ... maybe we should put all AI Civs on 'build often' for both Trade AND Wealth then. That should ensure they always have a lot of money, which will hopefully make them stronger.

We have to be careful not to put too many things on 'build often', though, because it would make very little difference then.

[lockstep]

A question to Soren:

In korn's blitz mod, some improvements were added that yielded only culture. The AI's actually chose to build some of them. How did they decide on that? Does every improvement that yields culture fall in the 'culture' category?

[alexman]

Soren, thanks for the clarification. It changes things a lot.

DeepO, good comments. I agree that this change is not for everyone.

If you like to see the AI build weird things just to give each civ a different flavor and to mirror their historical path, then definitely this change is not for you.

But if you prefer that the AI plays more like a human, and thus creates a greater challenge to the human, I think these changes would get you closer to that.

On top of that, I would argue that geography is the single largest factor in creating weak and powerful civs. It is very rare to get equal starting positions, so you will always get smaller civs eaten up by larger ones, with the larger ones becoming more powerful. But what if the large powerful civ has dumb build priorities? The answer is that he will be no challenge for the human. Builder and warmonger alike.

Making the Zulus start attacking everyone without any infrastructure is cute, but it is also suicide for them. They are like advanced barbarians. They hurt themselves as well as any neighboring, more balanced civs. Eventually the more balanced civ will win and will grow, but the real winner is the human player that stayed out of this mess.

As for wealth, you're right (especially after reading Soren's comments). It shouldn't be a build-never item, but it certainly shouldn't be a build-often item either.

[Iskandar Reza]

Make them stronger? But I'm having so much fun beating them up.

[Solver]

Interesting ideas, but as Soren shows, setting Wealth to never will ruin the AI even more. But really, I've noticed that I just outproduce the AI after Industrialization, because it doesn't have much factories - very interesting ideas to mess with.

[DeepO]

I agree that geographical reasons have a large impact on how a certain Civ turns out, but my intuition tells me that in order to get killer AIs you will need the diversity on top of a good starting position. Sure, in certain games these two will be countering each other, but I fear that in those games were for instance the Germans get a good starting position, the original .bic will make sure it becomes a killer AI, while your changes will negate it in large as others are able to catch up.

This is my biggest concern, not the historical correctness. However, as you said, the Zulus are cute, as you'd expect them to be advanced barbarians There is probably only one chance they have at becoming a late superpower, and that is to be so aggresive in the early game that they outpower the closest neighbours many times. They do build culture and factories, just not as often as the Germans.

Certainly, as I don't know the specifics, I can only guess at this... it is a gut feeling at best

DeepO

[Soren Johnson]

Originally posted by lockstep
A question to Soren:

In korn's blitz mod, some improvements were added that yielded only culture. The AI's actually chose to build some of them. How did they decide on that? Does every improvement that yields culture fall in the 'culture' category?

At the beginning of the game, the AI preprocesses every unit/building to put them into multiple categories, such as trade, wealth, and culture. Thus, a cultural bank would be listed as both a cultural and a wealth building.

btw, these difference were meant to give the AI's personality... adding to the fun, not the difficulty of the game. If you want to build a killer AI, messing with these figures is probably the best place to start.

oh, and the AI does place a higher priority on shield-increasing building (known as factories and power plants to human...)

[DeepO]

Thanks Soren

DeepO

[lockstep]

Thanks for the clarification, Soren. Mod-makers will be glad to know for sure that the AI can actually cope with additional buildings.

BTW, korn also added a 'mill', a medieval improvement (available with engineering) that increased shield-production by 25%. Result: The AI's LOVED to build mills (the mod was based on v1.17).

[Solver]

Thanks Soren. I'm really thinking to change priorities now... yes, I want the AI to build more factories.

While personalities clearly add up to the fun, sometimes I prefer challenge, for challenge is fun.

[alexman]

Originally posted by Soren Johnson Firaxis


btw, these difference were meant to give the AI's personality... adding to the fun, not the difficulty of the game. If you want to build a killer AI, messing with these figures is probably the best place to start.

Thank you!! Making a killer AI was the point of this thread before DeepO started bashing the idea!

[lockstep]

Good news indeed for the deity players who are obsessed with creating killer AI's.

[Arrian]

Wow, so in the standard bic, only 3 out of 16 civs will build marketplaces & banks "often."

I think the real key is the combo of trade & wealth, to make sure they build those marketplaces. The marketplace is one of the single best improvements in the game.

Questions:

- do courthouses fall under "production" ?
- do markets also fall under "happiness" ?
- does "expl" mean scouts and explorers? If so, that should be a "never build" for all civs. Or does this mean civs will build regular units which are told to explore?

I think all civs need to "build often" Wealth, Trade and Production. Militaristic civs could then also build often offensive units, whereas the "builder" civs could build often scientific improvements instead (Germany could do both... ignoring "culture" since their culture will come from the science improvements anyways).

I think this would help the AI a lot. Good for mods/scenarios.

-Arrian

[DeepO]

Originally posted by alexman
Thank you!! Making a killer AI was the point of this thread before DeepO started bashing the idea!

You do know that I was only trying to give constructive criticism, and had no bashing in mind, don't you?

And even if Soren might correct me again, I still feel that in scenarios, the biggest killer AIs would be those that have good starting positions, good tastes as to what to build, and 'weak' AI neighbours... these would be more killer than those in the same position but with strong AI neighbours as they more or less balance each other out. In that respect, the biggest killer AIs will come from the original .bic, although you have more chance of getting (lesser) killer AIs with your changes.

No, I won't give in that easily

DeepO

[alexman]

Originally posted by Arrian
- do courthouses fall under "production" ?

My guess is no. The game probably considers "production" improvements the ones that have a non-zero value in the production box in the editor.

- do markets also fall under "happiness" ?

Again, I think no. Just like for production, there is a "happy face" box in the editor. Marketplaces have zero there.

- does "expl" mean scouts and explorers? If so, that should be a "never build" for all civs.

Not sure if it's only for those units, but if it is, I think it should be set to "often" for expansionist civs so they can take advantage of their scouts. More scouts = more goody huts = more power! Explorers are another story, but I have never seen the AI build one - not even the Iroquois that have this flag checked.

Originally posted by DeepO
You do know that I was only trying to give constructive criticism, and had no bashing in mind, don't you?

Of course I do! That's why I added the smiley. I appreciate your feedback!

[DeepO]

Oh, BTW, when you want the AIs to build marketplaces, how about modding them so they give, on top of the usual advantages, 1 shield, 1 culture, 1 beaker? This would mean they are absolutely the highest ranking improvement an AI can build, without over doing their effects... and this regardless of the above table

DeepO

[lockstep]

Originally posted by alexman
I think no. Just like for production, there is a "happy face" box in the editor. Marketplaces have zero there.

OTOH, marketplaces have the 'increases luxury trade' flag.

[lockstep]

Originally posted by DeepO
Oh, BTW, when you want the AIs to build marketplaces, how about modding them so they give, on top of the usual advantages, 1 shield, 1 culture, 1 beaker?

Only no. 2 (1 culture) is currently possible with the editor.

[Soren Johnson]

Originally posted by alexman

My guess is no. The game probably considers "production" improvements the ones that have a non-zero value in the production box in the editor.


Again, I think no. Just like for production, there is a "happy face" box in the editor. Marketplaces have zero there.

actually, the answer is yes in both cases...

[DeepO]

And 5% production, 5% science extra? it would more or less be the same...

DeepO

[lockstep]

Hmm ... okay, courthouses and police stations reduce waste, therefore increase production.

Another guess: harbors would be 'growth' as well as 'trade' improvements.

EDIT: If I split harbors into docks (which would yield +1 food per water tile) and harbors (which would provide trade links and veteran sea units), will the AI cope with these changes?

[alexman]

Originally posted by Soren Johnson Firaxis


actually, the answer is yes in both cases...

Wow! That would mean marketplaces belong to three groups (trade, wealth, and happiness). Why don't they get built more often? A mystery to me...

Thanks for the answer, btw. I don't suppose it would be possible to get a complete list of what falls in each category, would it? It kind of defeats the purpose of having these things editable if we don't know what they do!

[dunk]

Originally posted by alexman

Thanks for the answer, btw. I don't suppose it would be possible to get a complete list of what falls in each category, would it?

This would be great for modders. Not only the where the current buildings lie, but why the buildings are in those categories. This way, we'll have an idea as to what happens when we create new buildings / wonders. Ideally, we should have a manual for the editor with this kind of information. I mean, look at all the conjecture we've had in this thread as to what these flags do. Even if it's just a big test file available through download only.

Firaxis, please, a user's guide or manual for the editor?