Tile selection on border expansion question

[Wit]

This question has been raised about a year ago without a satisfactory answer. Perhaps second time is the charm.

The question is: when borders expand, how is the specific tile that is added to the city chosen? I have encountered repeated instances when the city screen suggested that borders will expand to a certain tile only to have it changed later on to a different tile. This suggest a random ingredient in the tile selection algorithm. If true, why is that?

Does anyone have insight? Or can point me in the right direction?

 

[Trav'ling Canuck]

Ah yes. That old chestnut. One of those "why are you telling me false information?" issues.

Last time, by memory, we got as far as wondering why the UI even shows you a tile if it's not the one you're getting, followed by some speculation about whether placing a Builder on a tile could influence the game to pick that tile. Sort of a 21st century reading of entrails type thing.

I became so enured to ignoring what the UI tells me, I don't recall if I've actually looked at the supposed next tile expansion in 2018. Was this maybe fixed in R&F?

 

[FlyinPhil]

"Was this maybe fixed in R&F?"

Nope. Definitely had it happen on at least a couple of occasions in a game I finished last night. Went to see which tile would be chosen next, saw it was one I needed, so I didn't bother buying it. Ended up buying it later when I realized it was not the one chosen.

 

[Tech Osen]

Also rather amusing how Feraxis was all about the player having to make choices yet this is some cloudy automatic mechanic.

 

[Victoria]

I cannot be bothered looking for the thread @Trav'ling Canuck is mentioning but to summarize what we found out as bast as I can remember.

1. There is some weighting factors in the xml on this but taking them alone does not seem to be adequate and certainly when it comes time for expansion there appears to be a random decision made if the weightings of 2 tiles are similar.

2. The first rule is the city now will expand all of ring 2 tiles before moving on to ring 3. This used to be different but some people complained about odd shaped borders and as a result our city will ignore the iron hills in ring 3 and instead will choose a mountain in ring 2.

3. There is a huge bias toward lux/resources as there should be. It will always take these before land without them. So 1 lux/ resource in ring 2 means that will be the next tile. There did not appear to be any bias for lux or resource, but my memory is fading.

4. Sea or land I found a bit odd, if I remember there was a slight bias for land but once again there seemed to be RND involved.

5. There has been anecdotal swearing that placing a worker on a tile helped the bias, I have tried and I see no change in rates, it is always easy to overweigh the likelihood when you see it happen once.

 

[Wit]

@Victoria Thank you for the recap. I did some additional searching and found the thread I believe you were referring too. It was a good read, but alas we are pretty much still in the dark, it seems.

 

[Zdarg]

2. The first rule is the city now will expand all of ring 2 tiles before moving on to ring 3.

Playing just now: Guangzhou (AI's) has exactly one ring 2 tile and one ring 3 tile, which is cotton.
Perhaps rules are different for them?

 

[WillowBrook]

Playing just now: Guangzhou (AI's) has exactly one ring 2 tile and one ring 3 tile, which is cotton.
Perhaps rules are different for them?

Or they bought the tile. I'm pretty sure the AI cities follow the same rules as human player cities.

However, I have noticed that the CIVITAS city states mod changes the way city states acquire tiles, such that third-ring resources are chosen before boring second-ring tiles. But the fact that you can mod this behavior doesn't necessarily tell us anything about the default behavior. (I haven't dug into their code to see how they do this.)

 

[FlyinPhil]

4. Sea or land I found a bit odd, if I remember there was a slight bias for land but once again there seemed to be RND involved.

I think there must be some weighting to the amount or type of resources, because I've found it generally favors land, and even favors coast over ocean. There are times when I want a desert tile to place a district or wonder, but it picks a coast tile instead. I'm guessing that's because it was programmed to prefer tiles with resources.

 

[Trav'ling Canuck]

Regardless of the calculation used to pick the tile, the bigger mystery is why does it show you one tile as the one you will get, then give you another one?

The three most likely answers are:
(a) the UI display and the actual selection use different code for tile selection; possibly they were synced at some point and then one was tweaked and the other not
(b) there's a random factor involved, and one RNG number is used for the UI display, then another when the tile selection actually happens
(c) there's a non-random weighting factor that changes from the time a tile is picked for the UI display and the time the actual tile is picked

I still say that if the UI is going to show you a tile that you don't actually get, it would be better for it not to show you anything at all.

 

[WillowBrook]

Regardless of the calculation used to pick the tile, the bigger mystery is why does it show you one tile as the one you will get, then give you another one?

The three most likely answers are:
(a) the UI display and the actual selection use different code for tile selection; possibly they were synced at some point and then one was tweaked and the other not
(b) there's a random factor involved, and one RNG number is used for the UI display, then another when the tile selection actually happens
(c) there's a non-random weighting factor that changes from the time a tile is picked for the UI display and the time the actual tile is picked

I still say that if the UI is going to show you a tile that you don't actually get, it would be better for it not to show you anything at all.

Yes, this is the most annoying aspect of the whole thing. I'm guessing that there some sort of ranking, and if more than one tile come out on top, it's random among those. But the RNG is rolled each turn, so you can get any of the possibilities as the predicted next one.

IIRC, the pink prediction indicator was not in the game on release, but was added in an early patch due to fan demand. My guess is that they didn't think through this issue initially, and then the patched solution was sub-par.

 

[Wit]

I think Loyalty should factor into the tile selection decision. I think this would be a valid approach -- tiles between two civs should be harder to get automatically than a comparable one in your backyard. This would provide incentives to use cash to claim those precious tiles at your borders to a neighboring civ (bc the algorithm is more likely to choose tiles on the other side that experiences less loyalty pressure from other civs). I think this has intriguing potential.