New Beta Version (3-20b)

[BiteInTheMark]

@BiteInTheMark I want to point out that #2 is very hard to correct. If they're not aggressive enough they'll have this happen to them all the time too. I would have settled a city similarly to pen you in. I think it's more an evaluation of your willingness and capacity to fight them that could maybe be improved.

I do not think that the AI includes blocking the player in their calculations. Ilteroi can correct me there, if it is the case.
Especially peaceful civilizations, like Brazil, should not necessarily consider such tactics. Brasilia is already closer to my capital than to his (and the settlers of him even closer) and he is moving towards a civilization that has advantages in and through warfare.
Of course the position was a good spot (in 3 distance a natural wonder, horses and luxury good [he didnt get the luxury cause shoshone settled there, which was also silly, cause they started in top right corner and settled in the bottom left corner of the swedish continent above me (ring map)]), but the distance to their own capital, the possibility of logistical development and the reaction of aggressive neighbors should be much more involved.

"If they're not aggressive enough they'll have this happen to them all the time too."
If you change the behavior in that way they will not settle that far away, the other AI will also not do this and they dont get blocked by extreme aggresiv settling.

 

[Gazebo]

I do not think that the AI includes blocking the player in their calculations. Ilteroi can correct me there, if it is the case.
Especially peaceful civilizations, like Brazil, should not necessarily consider such tactics. Brasilia is already closer to my capital than to his (and the settlers of him even closer) and he is moving towards a civilization that has advantages in and through warfare.
Of course the position was a good spot (in 3 distance a natural wonder, horses and luxury good [he didnt get the luxury cause shoshone settled there, which was also silly, cause they started in top right corner and settled in the bottom left corner of the swedish continent above me (ring map)]), but the distance to their own capital, the possibility of logistical development and the reaction of aggressive neighbors should be much more involved.

"If they're not aggressive enough they'll have this happen to them all the time too."
If you change the behavior in that way they will not settle that far away, the other AI will also not do this and they dont get blocked by extreme aggresiv settling.

The AI absolutely does include strategic settling in its calculations. As an aside, this is the problem with discussing high-level AI stuff in settings like this: lots of assumptions are made about the AI's behavior and those are used to generate arguments that, while informed, aren't reflective of what the AI does or can do.

In this case, I think it's more fair to ask for modest changes, specifically that the AI should simply prefer closer locations when possible.

G

 

[BiteInTheMark]

The AI absolutely does include strategic settling in its calculations. As an aside, this is the problem with discussing high-level AI stuff in settings like this: lots of assumptions are made about the AI's behavior and those are used to generate arguments that, while informed, aren't reflective of what the AI does or can do.

In this case, I think it's more fair to ask for modest changes, specifically that the AI should simply prefer closer locations when possible.

G

Interesting to know, I didnt thought the AI is that complex already. You may be annoyed by discussions like thise, cause you have the overview of all those aspects, but we player only see the interface and the results and think about the possible sources. The only source of information about the background of AI behavior are the developers, like you. So please do not be angry, should we once again make naughty guesses.

 

[Gazebo]

Interesting to know, I didnt thought the AI is that complex already. You may be annoyed by discussions like thise, cause you have the overview of all those aspects, but we player only see the interface and the results and think about the possible sources. The only source of information about the background of AI behavior are the developers, like you. So please do not be angry, should we once again make naughty guesses.

Definitely not angry or annoyed - as I said, these are high-level discussions and your points are informed, but - as with the tactical AI - it's often not clear what the AI is doing compared to what it knows to do in the code.

G

 

[ilteroi]

ok, many interesting comments here.

trying to condense the points:

0. non-expansion
1. settling aggressively in indefensible locations
2. useless cities in bad locations

comments/questions in response:

0. seen that a couple of times. they don't build settlers. root cause might be as simple as AI not researching pottery in time, which can happen through bad luck, because of random choices. gazebo topic, maybe solved already.

1. many aspects to consider here:
* what is aggressive? some AI players are supposed to be aggressive ... i assume it would be a good move if they put enough units there to be able to defend the city? so in fact it's not really about the settling but about the defense, right? what it boils down to is that the AI needs to be able to judge the likelihood of an attack. and that's very difficult to code. who is a pushover and who isn't? who has room to expand and who is desperate?
* anyway, there's already a check for military power, if the neighbor is stronger they are supposed to keep their distance. there was an interesting proposal "don't settle near developed cities of powerful neighbors" ... sounds good but how to define developed? also this would lead to areas of no-mans land between players. the more powerful player would be able to grab it without a fight. is that an improvement?

2. again, many aspects:
* what is a bad location? a city has 37 plots in its three rings. with specialists and building bonuses etc it's very rare that all plots are being worked. so some overlap between cities is acceptable. but of course that's a matter of personal preference and the AI leaders have differences here too. in my experience, even marginal cities will grow (especially with a food trade route) and become productive eventually, due to the bonuses from buildings. the terrain is surprisingly unimportant.
* so how many "good plots" does a city need? what is a "good plot" anyway?
* a human will sometimes turn a good site into a bad site by rampant plot buying. i might discount the value of plots adjacent to enemy territory, but the real lesson for me is to add a cooldown between plot acquisitions.

 

[Moi Magnus]

ok, many interesting comments here.

trying to condense the points:

0. non-expansion
1. settling aggressively in indefensible locations
2. useless cities in bad locations

comments/questions in response:

0. seen that a couple of times. they don't build settlers. root cause might be as simple as AI not researching pottery in time, which can happen through bad luck, because of random choices. gazebo topic, maybe solved already.

1. many aspects to consider here:
* what is aggressive? some AI players are supposed to be aggressive ... i assume it would be a good move if they put enough units there to be able to defend the city? so in fact it's not really about the settling but about the defense, right? what it boils down to is that the AI needs to be able to judge the likelihood of an attack. and that's very difficult to code. who is a pushover and who isn't? who has room to expand and who is desperate?
* anyway, there's already a check for military power, if the neighbor is stronger they are supposed to keep their distance. there was an interesting proposal "don't settle near developed cities of powerful neighbors" ... sounds good but how to define developed? also this would lead to areas of no-mans land between players. the more powerful player would be able to grab it without a fight. is that an improvement?

I suppose you could use the economic power of the city (used for warmonger penalty when capture), but I don't know if it is doable, or anything which is the approximation of the age of the city.
The undelying idea is that forward settle a wide empire and forward settle a tall empire is not the same thing.

In truth, the main problem is not that they settle those cities, the problem is that they don't expect the human to declare war to raze them, so are not prepared to defend them. Maybe they underestimate the importance of the diplomatic penalty for forward settling ?
Another idea: is it possible to add a "connection check" ? which make sure that the AI is able to secure a reasonnably short path from its army to the city. The worst case for defense is when the city is isolated from the main empire.

For no mans lands, If the city is settled at ~6 tiles of distance from the other civ, there is not any remaining land, and not that much contested borders compared to settling at minimal distance. Moreover it give more space for defensive fight. Also, I do not advocate strict rules, forward settling a natural wonder is worth it.

2. again, many aspects:
* what is a bad location? a city has 37 plots in its three rings. with specialists and building bonuses etc it's very rare that all plots are being worked. so some overlap between cities is acceptable. but of course that's a matter of personal preference and the AI leaders have differences here too. in my experience, even marginal cities will grow (especially with a food trade route) and become productive eventually, due to the bonuses from buildings. the terrain is surprisingly unimportant.
* so how many "good plots" does a city need? what is a "good plot" anyway?
* a human will sometimes turn a good site into a bad site by rampant plot buying. i might discount the value of plots adjacent to enemy territory, but the real lesson for me is to add a cooldown between plot acquisitions.

How many good plots : well, usually, the reasonning is the converse. I want a city in a particular area, for luxuries, for defensive purpose, for land grabing, ...
But the AI is usually fine with it. The only problems is when there is no longer any "good enough" spot, and the AI still has settler to use. Then they either wanders, either settle crappy forward settled cities, while they should probably either wait for the upgrade of the settler, either settle a city in a more conservative way nearby their territory, taking advantage of the tiles no longer worked by develloped cities working specialist instead.

For plot buying, I don't think puting a cooldown will really solve the problem. It would balance it better, but would be very frustrating for the player. If any cooldown, I would forbid to buy tiles the turn you settle a city, in order to make sure the forward settled civ is able to buy the tile "it should have".
Personnaly, when settling, I consider that tiles at 3 tiles or less of a city belong to the city (or only 2 tiles for a "newborn" city), so I don't count them (except if I have the gold to purchase them immediatly, but I feel like cheating when doing so).

 

[Stalker0]

even marginal cities will grow (especially with a food trade route) and become productive eventually, due to the bonuses from buildings. the terrain is surprisingly unimportant.

Considering how much debate went into nerfing the hammer bonus from a forest that you settled on, I can't imagine this is true. Now yes, maybe when you are plopping down Pioneers later in the game this is true, but early game those yields are very important. Everything is a snowball at that point in the game, more food = more people = more yields. More hammers = more buildings = more yields. Culture = more policies = more yields.

So yes eventually any city can become productive, there is a huge difference between productive at turn 50 vs turn 150.

 

[tu_79]

what is aggressive?

there's already a check for military power, if the neighbor is stronger they are supposed to keep their distance

Aggressive is a civilization that has an army more developed than its economy. I guess there can be a check on military strength to hammer per turn ratio, or to population, or something like that. A human player may be considered (30%?) more aggressive than what his military strength shows.
By 'keeping the distance', do you mean that AI still is able to settle near a warmonger, but keeping a fair distance of 5-6 tiles to the closest city? Or is it simply that the AI looks at other directions?
If the AI wanted to, say, forward settle to launch an attack on a neighbour, could it ignore the cautions and escort the settler with a few military units instead?

in my experience, even marginal cities will grow (especially with a food trade route) and become productive eventually

so how many "good plots" does a city need? what is a "good plot" anyway?

My take on this. My spots are usually revealed resources or natural wonders that can be placed within the first two rings of a city, call this spotted resources.

For settler: 2+ resources
For pioneer: 1+ resources
For colonist: No requirements.

Also, if a resource is on the border to another civ territory, consider its value halved (the risk of not being able to get it). And if the chosen locations is either tundra or flat desert, raise the resource requirements by 1. Optionally, prioritize fresh water and hills.

 

[Gazebo]

I really think you are all overthinking this. The AI planning in place is already sufficient, it just needs to be more sensitive to threat and proximity than it is now.

G

 

[tu_79]

I really think you are all overthinking this. The AI planning in place is already sufficient, it just needs to be more sensitive to threat and proximity than it is now.

G

I agree. We were giving suggestions on how to be more sensitive to threat and proximity.

 

[Moi Magnus]

I really think you are all overthinking this. The AI planning in place is already sufficient, it just needs to be more sensitive to threat and proximity than it is now.

G

And reducing the valuation of tiles that can be bought by the player. That's probably the main current problem : the AI settle to have ressource the player can easily deny to him.
But yes, the AI planing is reasonnably good.

 

[Gokudo01]

I think casus belli doesn't work as the tooltip may suggest :
"
The world recognizes the need for justifiable war to resolve diplomatic disputes. While this resolution is active, unit maintenance costs are lowered by 15% Gold. Warmonger penalties for capturing cities and declaring war are greatly reduced, and Warmonger scores decay much more rapidly than normal."

ok so the warmonger score that you get from war and capturing cities is indeed greatly decreased but the decay is a joke.
I mean in my last game, Indonesia which was friendly, was losing 1 warmonger scores per turn over 180 ... with casus belli active.

 

[Moi Magnus]

I think casus belli doesn't work as the tooltip may suggest :
"
The world recognizes the need for justifiable war to resolve diplomatic disputes. While this resolution is active, unit maintenance costs are lowered by 15% Gold. Warmonger penalties for capturing cities and declaring war are greatly reduced, and Warmonger scores decay much more rapidly than normal."

ok so the warmonger score that you get from war and capturing cities is indeed greatly decreased but the decay is a joke.
I mean in my last game, Indonesia which was friendly, was losing 1 warmonger scores per turn over 180 ... with casus belli active.

Which game pace ? Maybe its just the scaling with gamepace that does not work properly.

 

[Gokudo01]

Which game pace ? Maybe its just the scaling with gamepace that does not work properly.

standard standard communitas

 

[civplayer33]

ok so the warmonger score that you get from war and capturing cities is indeed greatly decreased but the decay is a joke.

Yeah I've noticed that as well; I don't play with transparent diplomacy but the rate of decay with Casus Belli seems to not or just barely be increased vs. no Casus Belli active.
Either the description should be changed or the rate should actually be increased significantly to reflect the description.

 

[Gokudo01]

Yeah I've noticed that as well; I don't play with transparent diplomacy but the rate of decay with Casus Belli seems to not or just barely be increased vs. no Casus Belli active.
Either the description should be changed or the rate should actually be increased significantly to reflect the description.

I was monitoring it turn after turn with transparent diplomacy

 

[Rekk]

I'd like to show off a questionable decision England made in a 18-1-14 Epic King game.

Turn 301, there is a large space of resourceless jungle stuck between city states, me and India:

Fully 70 turns later, England decides to colonize this area:

You can see that she occupies the northeast corner of the Pangaean continent. In order to settle Liverpool, she had to walk across regions controlled by Poland/Russia, India, the Aztecs, and Rome (Me), to get to an area that had been devoid of resources for at least an era. What sort of decision making process could she have gone through to make this decision?

 

[Stalker0]

So on Emperor Small Terra I just got my butt whipped 4 times in a row in early warfare. So Iteroi you must be doing something right

 

[anip94]

I'd like to show off a questionable decision England made in a 18-1-14 Epic King game.

Turn 301, there is a large space of resourceless jungle stuck between city states, me and India:
View attachment 491744

Fully 70 turns later, England decides to colonize this area:
View attachment 491745
You can see that she occupies the northeast corner of the Pangaean continent. In order to settle Liverpool, she had to walk across regions controlled by Poland/Russia, India, the Aztecs, and Rome (Me), to get to an area that had been devoid of resources for at least an era. What sort of decision making process could she have gone through to make this decision?

AI just likes to spam city now thats it. any freelands they just spawns the city there. i've witness it many times.

 

[Moi Magnus]

I'd like to show off a questionable decision England made in a 18-1-14 Epic King game.

Turn 301, there is a large space of resourceless jungle stuck between city states, me and India:
View attachment 491744

Fully 70 turns later, England decides to colonize this area:
View attachment 491745
You can see that she occupies the northeast corner of the Pangaean continent. In order to settle Liverpool, she had to walk across regions controlled by Poland/Russia, India, the Aztecs, and Rome (Me), to get to an area that had been devoid of resources for at least an era. What sort of decision making process could she have gone through to make this decision?

If you have a save (and use reasonnably few mods), I'm sure Ilteroi would be glad to have this on Github to correct AI's logic.

Also, does that zone has late game ressources unrevealed ? If the AI is settling for those ressources, that's not a bug in the AI settling logic, but that's a bug since the AI is not supposed to know the position of unrevealed ressources.