Improper AI settling

[Matara]

There's a problem that sometimes AI are settling improperly and ahistorically. E.g. Egypt spamming useless desert cities everywhere. Or France conquering barbarian Anjou as capital (1N of Bordeaux).
Sometimes moving AI settler or change settler value could work but sometimes it couldn't.
So the question is:
1. What factor will AI settler consider？
2. Is it always possible to change such improper settlements？

 

[MrFeline]

I unfortunately cannot say I know how to change this behavior. However, I will say it is somewhat accurate in the case of Egypt - spamming desert cities along the Nile is a fair assessment of their civilization. Where this becomes a problem, though, is when those cities last to the present and cripple the productivity of the Mamluks and modern Egypt.

I wonder if a mechanic could be put in place that would consolidate these cities / raze the smaller ones when Egypt collapses or flips. This could allow for a densely packed ancient and classical Egypt with a more sustainable medieval to modern one.

 

[zinedine]

This is an issue with the AI itself and there are multiple ways to solve it:
- modify the ai code (idk if possible)
- specialize the ai code per civ (idk if possible, v. expensive)
- pigeon-hole the ai behavior through parameter-pressure (limit food, city spawns).
I went with number 3 because i don't have time to check out the DLL. I've had some success in limiting the city-spam in Egypt in my micromod (historical colonization micromod).
This also comes with the additional benefit of better modeling the history of egypt (nubian expansion, militarism, etc)

Attached is an image with a typical ~1200BC egypt with those changes.
by eliminating desert food sources (oasis, a few flood plains, marked with red) and by script-adding some key very early cities:
Nubians/Buhen/Khartoum (independent, 2500BC, historically accurate)
Alexandria/Ra-Kedet (egypt or independent, 2500BC, historically inaccurate but necessary)
Tanis (1300BC, historically accurate)
i managed to pigeon-hole the Egyptian AI into being very militaristic (historically accurate), deprioritizing colonization and prioritizing war. It captured Buhen and it is also pressuring Jerusalem.

Note that once that parameter-pressure is alleviated, the AI will look to expand (Memphis building a settler), but these settlers have a limited space for botching up the good and historically relevant population centers (Delta - Thebes - Buhen). Statistically the next city is either Akhenatan (red sea port) or Abu-Simbel/Mut/Selima (on the sahara-side of the Nile), both have limited growth potential and can be rather easily razed.

 

[DanLT3]

This looks promising. Though its not only Egypt AI that have these problems. I think it needs to be applied to all civs.