AI Pantheon Decision Making

[doublex55]

How do the AI choose their pantheon? I ask because I think they are making clearly sub-optimal choices that are really starting to piss me off. Stuff like taking Spirit of the Desert with 1 tile of desert without a resource. Or taking God of the Sun with 1 wheat in capital and none more in expansion range.

Do AI look at their expansion range or only capital? Do they look for resources on desert/tundra or only the terrain type?

And it is impossibe to get a pantheon before almost all the AI have one on immortal, so there is nothing i can do to secure these types of pantheon.

 

[Ziad]

Yea I agree it can be annoying when AI make what would constitute an irrational decision in the eyes of the player.

In this particular case, it is indeed apparent that the AIs choice is influenced by what it works in its current cities (not just the capital). Also, the AI tend to pick pantheon that are more aligned with their flavor. Ghandi for example picks Expanse if going Authority or Love if not.

I'm not sure how difficult it would be using the game's API, but I think it would serve the AI better if there was some sort of scanning algorithm that checks the nearby tiles of a few randomly selected cities to influence the pantheon choice.

Honestly it's more for the AI than the player. Seeing Isabella fail to found because she went God of the Sun cause of one wheat resource in the vicinity is a little sad.

 

[tu_79]

Yea I agree it can be annoying when AI make what would constitute an irrational decision in the eyes of the player.

In this particular case, it is indeed apparent that the AIs choice is influenced by what it works in its current cities (not just the capital). Also, the AI tend to pick pantheon that are more aligned with their flavor. Ghandi for example picks Expanse if going Authority or Love if not.

I'm not sure how difficult it would be using the game's API, but I think it would serve the AI better if there was some sort of scanning algorithm that checks the nearby tiles of a few randomly selected cities to influence the pantheon choice.

Honestly it's more for the AI than the player. Seeing Isabella fail to found because she went God of the Sun cause of one wheat resource in the vicinity is a little sad.

On the other hand, it would feel boring if Spain always picks Expanse (if that's the best pantheon for her). So it's ok if AI makes mistakes from time to time.

It would be nice to improve the pantheon algorithm if that's posible, though. It won't affect too much the game performance wise, since it's a decision that needs to be made only once, and it can increase significantly the difficulty, since picking the right pantheon has long lasting effects. (I think I need to improve my own pantheon selection algorithm )

Side note: I think the Sun pantheon is still bad, in a recent game I couldn't found a religion even with a 3 + 1 farms start (the best I've seen in a long time), king difficulty. It's ok to have some pantheons weak for founding if they scale well, or if they are made to give a strong starting game without founding, but I don't see this happening with farms.

 

[Ziad]

Most pantheons don't scale at all. God of the Sun does pretty well for a short-term boost by providing 7 yields for a city with just a Granary + 1 Wheat. It's actually one of the better ones that don't scale since it provides Food which can be used to gain other yields. It also does not require any technologies except Pottery (which you prioritize in general).

Of course if you only have 1 Wheat in the entire 20 tile radius, it's probably better to pick something else since you'll be starved of Faith, hence the problem I have with the current algorithm.

 

[Mad Madigan]

Isn't the likelihood of the AI to pick the optimal Pantheon also tied to difficulty? I could have sworn several patches ago a change was put into place where the variability of AI choices becomes more random the lower the difficulty, with Diety-level AI supposedly only being able to pick from the most optimal options at every decision point. So if not playing on Diety the AI is still "allowed" to randomly pick non-optimal Pantheons, Policies, build orders, etc. on occasion. I usually play on King or Emperor, and I've been noticing rather understandable Pantheon choices for the AI for the most part. The AI also seems to pick certain Pantheons based on their Unique Abilities. For example, Netherlands often seems to pick God of Festivals, India God of the Expanse or Goddess of Love, etc.

 

[tu_79]

Isn't the likelihood of the AI to pick the optimal Pantheon also tied to difficulty? I could have sworn several patches ago a change was put into place where the variability of AI choices becomes more random the lower the difficulty, with Diety-level AI supposedly only being able to pick from the most optimal options at every decision point. So if not playing on Diety the AI is still "allowed" to randomly pick non-optimal Pantheons, Policies, build orders, etc. on occasion. I usually play on King or Emperor, and I've been noticing rather understandable Pantheon choices for the AI for the most part. The AI also seems to pick certain Pantheons based on their Unique Abilities. For example, Netherlands often seems to pick God of Festivals, India God of the Expanse or Goddess of Love, etc.

Not exactly this, if I understood Gazebo correctly. It sorts pantheons for order of 'optimallity', then picks the best on deity, chooses randomly between best and second best option for lower difficulty, and chooses among some other good options the lower the difficulty. So it should never choose a plainly bad option.

 

[Gazebo]

AI randomness is tied to difficulty yes. Lots of complaints about the AI choosing the same pantheons over and over, this we slightly increased randomness at lower difficulties. Nothing to see here.

G

 

[doublex55]

I'm fine with the randomness for the non-terrain dependent ones, but this is occuring at Immortal difficulty that if there is even a tiny bit of terrain they take the pantheon. Same thing happens with Machu Pichu in 1 mountain cities.

 

[Gokudo01]

On the other hand, it would feel boring if Spain always picks Expanse (if that's the best pantheon for her). So it's ok if AI makes mistakes from time to time.

It would be nice to improve the pantheon algorithm if that's posible, though. It won't affect too much the game performance wise, since it's a decision that needs to be made only once, and it can increase significantly the difficulty, since picking the right pantheon has long lasting effects. (I think I need to improve my own pantheon selection algorithm )

Side note: I think the Sun pantheon is still bad, in a recent game I couldn't found a religion even with a 3 + 1 farms start (the best I've seen in a long time), king difficulty. It's ok to have some pantheons weak for founding if they scale well, or if they are made to give a strong starting game without founding, but I don't see this happening with farms.

no the best pantheon for spain is god of love + authority : turn 1 you settle -> turn 2 you get the first faith and GA yields and food from growing border -> turn 3 you get another faith and golden age pop up which represent 24 faith and 24 golden age with zero investment : no worker needed.

 

[Ziad]

Agreed. Goddess of Love has a lot of synergy with Isabella due to her UA. Isabella with Goddess of Love and Cooperation is a monster.

Of course you have to pray that the AI doesn't pick either before you, seeing as they are both top-tier.

 

[Gokudo01]

Agreed. Goddess of Love has a lot of synergy with Isabella due to her UA.

Isabella with Goddess of Love and Cooperation is a monster. Of course you have to pray that the AI doesn't pick either before you, seeing as they are both top-tier.

yes but it still prefer to get two faith buildings because I need to dump all those extra faith

 

[Ziad]

yes but it still prefer to get two faith buildings because I need to dump all those extra faith

But faith buildings provide even more faith (compounding the problem) and you can't really build more than the number of your cities. You also can't build them in puppeted cities which is what you'll probably have if you're playing Isabella.

I would still prefer going with at least one passive ability (Cooperation, Inspiration, Diligence, etc...). Order is great for a building that you would only need in your non-puppet cities, although you could make any work depending on your strategy.

If you need a faith dump, I would suggest Zealotry (also very much appropriate given her theme). I also think Gazebo is experimenting with letting her buy ships with faith. Might be interesting.

 

[Gokudo01]

But faith buildings provide even more faith (compounding the problem) and you can't really build more than the number of your cities. You also can't build them in puppeted cities which is what you'll probably have if you're playing Isabella.

I would still prefer going with at least one passive ability (Cooperation, Inspiration, Diligence, etc...). Order is great for a building that you would only need in your non-puppet cities, although you could make any work depending on your strategy.

If you need a faith dump, I would suggest Zealotry (also very much appropriate given her theme). I also think Gazebo is experimenting with letting her buy ships with faith. Might be interesting.

I don't puppet much as Spain, I need mission rolling asap and after the initial population lost, you should get lots of yields from the Mission's mechanism. If the city is too bad but has got some strategic or luxury ressources, I puppet but I raze most of the time.

Ah and I get zealotery AND Jesuit Education/Faith of the Masses. it frees me LOTS OF PRODUCTION and allows me to develop cities at the lightning speed.

 

[Ziad]

Ah you probably play on Standard-sized maps. I always forget that my settings are not the default lol

 

[Gokudo01]

Ah you probably play on Standard-sized maps. I always forget that my settings are not the default lol

Yes