AI Suggestion for Settling - what am I missing?

[r_rolo1]

^^Not true

 

[pigswill]

I haven't checked this out on later upgrades but I've certainly seen AIs build cities in really naff locations and then 100 turns later there's oil or uranium in its BFC. The city is built long before the civ has the technology to reveal the resource.

I suspect that its not that AI civs spot hidden resources per se but rather that the AI civs follow the blue circle recommendation and its the blue circle algorithm that identifies spots with the most available resources (hidden or not).

Edit: On the other hand number of available resources is not the only factor in optimal city location. Factors like coastal location, riverside location, placement of other cities and optimising the total terrain (rather than resource tiles) can be just as important which human players realise but AI civs often don't (which generally works to our advantage).

 

[Wlauzon]

The AI sees resources that you haven't discovered yet

Yeah, I finally figured that out.

From reading some of the tutorials etc, it appears that while each individual CIV does not "know" where the future goodies are, the "blue circle" AI *does* know, and that is why some recommendations seem a bit weird.

I looked at about 10 maps with WB where the locations seemed pretty sucky, and most had some kind of yet unseen resource. But some were still sucky

 

[Bhruic]

From reading some of the tutorials etc, it appears that while each individual CIV does not "know" where the future goodies are, the "blue circle" AI *does* know, and that is why some recommendations seem a bit weird.

No, the "blue circle" AI (which is a useless distinction - there is only one AI algorithm for city placement) does not know where unrevealed resources are. The only time unrevealed tiles are taken into consideration is for the original settler placement (which is the starting city location for all AIs, if not always humans).

The main problem with the algorithm is the exception rules. For example, someone on the forum might say "you never want to build a city that's 1 tile away from an ocean". And that's true for the majority of cases. But another player might say "well, yes, but if you look at this example, it actually makes more sense to build 1 away because you get access to resource X and Y". Which also might be true. But it's a rule exception. You can't code an algorithm that would account for the thousands (or more) of rule exceptions there would be to city placements. So instead it does weighting. Each square is weighted as a potential city location, and the squares with higher weightings are chosen as settlement locations.

There are some issues with this - r_rolo1 quoted me mentioning one of them. Another would be that the AI doesn't "know" about various bonuses that might exist. It doesn't know that an water tile is going to improve if the city can build a Lighthouse. It doesn't know that a river tile is going to improve if the city can build a Levee. So the criteria that a human would use with that sort of fore-knowledge isn't really feasible for the AI.

Bh

 

[r_rolo1]

I just tested it and I can say with absolute certainty that BtS AI blue dots don't have in consideration hidden resources.....

I've made a little WBing...

And the only blue dot that I saw was this:

That didn't moved with me changing the required tech, or even with the elimination of resources that would be in BFC of that tile.

I still don't understand what rules BtS blue dots follow, but hidden resources are not one of criteria

WB file avaliable upon request

 

[Bhruic]

Actually, that example isn't likely to help, unless you were allowing a turn to pass after making changes - modifications via WB won't modify the city placement location immediately - it needs a turn to process the changes.

Bh

 

[r_rolo1]

I tested it a change per turn ( and retested it after seing your post , just to make sure... ) . 2 nd time there was no dot and no dot appeard with the tech changes....

 

[Wlauzon]

Actually, that example isn't likely to help, unless you were allowing a turn to pass after making changes - modifications via WB won't modify the city placement location immediately - it needs a turn to process the changes.

Bh

Hmm I was testing it by saving it and reloading, so maybe my results are not really valid, and just part of the randomness.

 

[Wlauzon]

No, the "blue circle" AI (which is a useless distinction - there is only one AI algorithm for city placement) does not know where unrevealed resources are. The only time unrevealed tiles are taken into consideration is for the original settler placement (which is the starting city location for all AIs, if not always humans).

There are some issues with this - r_rolo1 quoted me mentioning one of them. Another would be that the AI doesn't "know" about various bonuses that might exist. It doesn't know that an water tile is going to improve if the city can build a Lighthouse. It doesn't know that a river tile is going to improve if the city can build a Levee. So the criteria that a human would use with that sort of fore-knowledge isn't really feasible for the AI.

Bh

Perhaps I misunderstood how it worked. From what I read, when the map was generated, the AI figured out ALL the "best" places to start a city, and the blue circle recommendations only changed if that land was occupied. But it appears to be a bit more dynamic than that. Quite often I have noticed that the starting spot for a settler does NOT have a blue circle, but a spot 3-4 moves away will have one - yet when looking at WB, the original starting point seems to nearly always to be the best. Is it simply "presuming" a blue circle at the original start?

I guess I am still a bit confused about how the AI works, but your explanation that the AI is "stoopid" about any bonuses explains a lot.

 

[Bhruic]

From what I read, when the map was generated, the AI figured out ALL the "best" places to start a city, and the blue circle recommendations only changed if that land was occupied. But it appears to be a bit more dynamic than that. Quite often I have noticed that the starting spot for a settler does NOT have a blue circle, but a spot 3-4 moves away will have one - yet when looking at WB, the original starting point seems to nearly always to be the best. Is it simply "presuming" a blue circle at the original start?

Again, the AI takes the "hidden" resources into consideration when it is placing the original settlers. Once they are in play, it no longer does. Any "blue circle" that you see does not consider "hidden" resources. That's why you may see a blue circle somewhere other than your starting square.

Bh

 

[reknaw]

My understanding was that the AI places the blue circles in the places that would give the best empire, and not the best single city. That is to say that assuming you settled every blue circle on the map your cities would, overall, be better on average. This is completely not the way most humans play, as we try to get fewer, better quality cities.

 

[Incan Emperor]

My understanding was that the AI places the blue circles in the places that would give the best empire, and not the best single city. That is to say that assuming you settled every blue circle on the map your cities would, overall, be better on average. This is completely not the way most humans play, as we try to get fewer, better quality cities.

Not to metion the AI doesnt understand City Specialization.

 

[Wlauzon]

It also seems to have some odd ideas about resources. I have seen it "blue circle" 2 cows, which are common, vs 1 coal in the other direction, which on that map only had 2 sources total on a huge Pangea.

I think I got the basic idea of how it works to start, but quite often late game suggestions make no sense to me. For one thing it seems to like to put cities overlapping.