Improving AI priorities

[mystikx21]

Libraries are useful for offense by giving new techs and new techs can mean better units or better production, more gold, etc, so I could see the GL being useful on those grounds for "offense". The other way to look at that is that it can be used to deny better techs to an enemy they are trying to rush.

Pyramids is useful more for expansion though.

Walls et al don't add very much food or production and not all AIs will use the same policies. Seems better to put a light priority on the policy than on the walls.

Agree whole liberty and tradition tree needs reworked priorities.

Terracotta would get the old Stonehenge priorities I should think for culture and expansion.
Problem with Petra being over prioritized is that the AI might not put it in a very competitive place (I agree it's potentially very useful). Putting a high AI priority could at least mean that if you want to build it somewhere awesome, you'd better hurry.
Happiness buildings are good for expansion AI because the AI doesn't do as well at calculating future or potential happiness. Giving it a priority on additional happiness would seem like it would help there.

I assumed the liberty finisher wasn't completed yet myself.

 

[Ahriman]

Libraries are useful for offense by giving new techs and new techs can mean better units or better production, more gold, etc, so I could see the GL being useful on those grounds for "offense".

I don't think this is a helpful way of setting flavors. I think flavors should be assigned to buildings, wonders and policies based on their direct effects, not on indirect effects.

Then set leader flavors appropriately; some might favor early brute rushes (Aztecs, Huns) while others might favor a combination of military and science (eg Germany) and so have preferences for both science and offense.

Similarly, granaries and aqueducts should give Growth, not Science.

Pyramids should be expansion.

Walls don't need production or food flavors; just defense flavor, and then put production or food flavors on the appropriate social policies. If you aren't using Tradition or Liberty then walls get neither of those bonuses.

 

[Thalassicus]

@mitsho
Thanks for that helpful list!

@Ahriman
In general I try to have only science on science buildings, growth on growth buildings, like you say. I limit weird stuff to specific cases where I found the AI is not doing what it should. There are some basic problems with the game core's priority system.

For example, the AI doesn't plan ahead well. The AI is clueless that the Piety tree has culture policies, unless the Piety policy itself has a culture priority. The expansion/science numbers on the Pyramids and Great Library were not enough to encourage militaristic AIs to get these, so I added "offense" numbers. It's not always ideal. The Great Library has tons of flavors because almost every human player tries to get it, no matter how much I tried to nerf it, so I just returned it to its free-tech bonus and encouraged every AI player to act like a human.

 

[mitsho]

Right, forgot the libraries = better tech = useful for offense (but really, everything ) link. It does make sense and these indirect links are probably very important to create a competitive AI. I guess, much work should go into finetuning these lists.. (which assumes that no big changes are made to the buildings themselves ).

Walls et al don't add very much food or production and not all AIs will use the same policies. Seems better to put a light priority on the policy than on the walls.

I only suggest a small flavor on them, these buildings don't hurt the AI (maintenance) and it's a shame for them to miss out on the bonusses. Also conquerors (Honor) are the ones that don't profit, normally not the ones to build walls (themselves). If it has a low priority, it will get chosen before things the AI doesn't really need, like a shrine (if they're not religious) or a amphitheatre (just examples). It's an indirect link of course, like the library one mentionned above.

Of course, if we want the AI cities to fall more easily, it'd probably best to not encourage them to build walls

Problem with Petra being over prioritized is that the AI might not put it in a very competitive place (I agree it's potentially very useful). Putting a high AI priority could at least mean that if you want to build it somewhere awesome, you'd better hurry.

The situation seems rather similar to the Great Library above... If the AI randomly builds it in a very good place, good for them. If they don't, well then they at least denied it to the player

@Thal, I just got to Late Ancient Era in my game though, otherwise I would've done more...

 

[Ahriman]

In general I try to have only science on science buildings, growth on growth buildings, like you say. I limit weird stuff to specific cases where I found the AI is not doing what it should. There are some basic problems with the game core's priority system.

For example, the AI doesn't plan ahead well. The AI is clueless that the Piety tree has culture policies, unless the Piety policy itself has a culture priority.

Ok, this sounds sensible.

The expansion/science numbers on the Pyramids and Great Library were not enough to encourage militaristic AIs to get these, so I added "offense" numbers.

I'm not convinced that Pyramids and the Great library are amazing picks for offensive players; wouldn't they be better off building a bigger army than losing a bunch of time to Wonder-building? With a bigger army, they could conquer their neighbor who built a Wonder.

 

[mystikx21]

I thought the priorities are available on buildings and wonders throughout by looking over the tech tree. If you don't want to wade through the xml itself that is. Balancing the early ones does give some stages of work at least.

Those bonuses on walls, etc, really only add up to much growth or production if the AI really puts a premium on defence. Which is to say that it has much better ways to grow or produce that it should be doing...but I will say it needs to put a high-ish premium on defence. For... other reasons.

@Ahriman/Thal. It would seem like it would be useful to have some science/offence links and that it could still be possible to have say, the Germans or Greeks or Romans still be more science focused than the Aztecs or Huns. That doesn't mean that everything listed as "offence" deserves it either. I'm not sure the Pyramids does and could be talked out of the GL over regular libraries.

 

[mitsho]

Medieval Era

Hagia Sophia loses Great People, Religion up to 16 ( or more?)
Chichen Itza loses Great People (what's the connection between Golden Ages and Specialists)
Macchu Picchu gets Food/Growth and Culture (10).
Angkor Wat gains Religion (the Pagoda)
Alhambra loses Religion, gains Offense (16)
Notre Dame/Lalibela needs total overhaul
Harbor gains Production (base hammers)
Garden gains small Growth (if the stays)
Should Pikeman have the same Defense value as Crossbows? (Archers as Spears, etc.)
Galleons gain Ranged (If the AI uses it)
Commerce policies have too much naval (Mercantilism, Guilds and Protectionsim, if no links)
Guilds higher Gold (16?), no Offense, no Expansion
Protectionism (cheaper purchase costs) could gain something else?
Merchant Navy loses Gold and Expansion.
Patent Law loses offense, gains naval changes into naval_tile improvement
Patronage is a tricky thing, is gold correct as a AI priority for better gold gifts to AI? Should they get low amounts for the other yields cs can provide (units, food, culture, religion)?

Renaissance Era
Taj Mahal gets the other priorities for Golden Ages as well (food, production, etc. )
Kremlin gains culture
Himeij gains low offense (if you fight offensive wars, after having conquered the cities, you will fight in your own territory)
Seaport gains naval_tile_improvement
workshop loses grwoth
Constabulary gains whatever effect we give the building
Lancer loses Defense?
Cannon loses Defense (Ranged units are better at that, not?)
Knowledge loses Naval (to not make the AI pick this one before Commerce), Religion (using faith for scientists is not a religion acts),
All Knowledge loses Culture,
Scientific Revolution and Sovereignty loses Happiness
Humanism has lower Spaceship (and maybe lower science)
Should all the specialist slot providing buildings gain a small number on Great People? (I think so)

 

[mystikx21]

Not sure gardens are much use for growth vs harbors being good for production.

Alhambra has a culture bonus too (and a free castle).

 

[Ahriman]

I wouldn't add flavors for every little +1 whatever. If you do that you just end up diluting the whole point of the flavor system.

Gardens are about great people; give them great person flavor.

 

[mystikx21]

I don't think specialist slots ought to do much with great people either. Do we build something because it has slots? Seems more like we build it because it increases production or science.

 

[Ahriman]

I don't think specialist slots are have much on great people either. Do we build something because it has slots? Seems more like we build it because it increases production or science.

Well, in my view we *should* build things because they have slots... the fact that we don't is in my view because there are a few too many slots available, and because specialists are too easily substituted for working tiles, because the great people aren't enough about gpps rather than yields.

 

[mystikx21]

If the two things are equal, I'd build something with a slot. Since they're in order rather than comparing a mint and a market or university and school, it makes enough sense that the AI would get some GP elements out of whatever it is that it is preferring (science or production). That way the AI gets slots for what it seems to prefer as is.

 

[Ahriman]

If the two things are equal, I'd build something with a slot. Since they're in order rather than comparing a mint and a market or university and school, it makes enough sense that the AI would get some GP elements out of whatever it is that it is preferring (science or production). That way the AI gets slots for what it seems to prefer as is.

I agree with you that it's fine to not have great person flavor on things just because they have a specialist slot.

 

[mystikx21]

Other notes:
Cannon and Lancer should have some defence value. They are useful in those roles, say for counter-attacks. They're not as useful as other units. So they can have some defence priority, but not nearly as much as other units of their era.

 

[mitsho]

Well, if I understand the AI priority system correctly, giving something a low value for food (say the garden) will mean that it will be the last thing built if the city/AI wants to build food in there. They are absolute values and they will build the highest one available, (if they decide to boost that priority which is based on leader values somehow?)

IF it works this way, it does make sense to give low values on buildings that help that thing a little. If I want to have the highest production possible in this city, I will build a stable, even if I only have one horse tile, of course after having built the forge, blast furnace, windmill, workshop, harbor, etc. ... Maybe not before the bank, but still, you gotta work with what you have

@Alhambra, I feel it does feel like a wonder to be changed, seems a bit unfocused right now, not?
@Building with Slots, yes, I agree, they shouldn't feel like an afterthought, and I guess, I'm not sure myself as I'm not sure about how the priority system itself is working.

 

[mystikx21]

Alhambra is all over the place, but the religion priority that seemed to be on it didn't make any sense with the places it was going.

I'm not sure the AI needs to think in terms of maximizing growth over a +1 yield. I tend to think of that as a bonus rather than a necessity. This is less true of the bonus production on stables, which can be anywhere from the base +1 to 4-5 bonus tiles. I don't think the AI does very well at figuring that out, so we could think in terms of an average yield and assign it a weight based on that. For the garden the average yield is just +1 though. Given the cost of upkeep, it's really only useful to build if there are GP rates to consider.

 

[Ahriman]

I think the point is that the AI always wants multiple things; if it wants gold and food and it is a leader who has a high food flavor rating, there are times where it might get confused and build the garden instead of the market.

 

[Zaldron]

Ok, this sounds sensible.


I'm not convinced that Pyramids and the Great library are amazing picks for offensive players; wouldn't they be better off building a bigger army than losing a bunch of time to Wonder-building? With a bigger army, they could conquer their neighbor who built a Wonder.

I actually agree with this. I just had one of my best games ever by completely ignoring all wonders and only building units, settlers, and buildings.

 

[Zarggg]

Re: Gardens, I was always under the impression that they were the decorative/cultural variety, rather than the sustainable foodstuffs kind.

 

[mystikx21]

Could think of them as mild irrigation or agricultural knowledge from botanical practices and then later as communal vegetable gardens in urban districts. Or some such. The bonus is small, and thus not crucial to the building.