Cargo helicopters...

[Hans Lemurson]

I think Martin just fixed it so the AI could build more than one a turn.

That fool, he has doomed us all! DOOOOOOOOMED!!!

Now, on a more rational note: Has the AI's city-congestion always been a problem?

[Maquiladora]

That fool, he has doomed us all! DOOOOOOOOMED!!!

Now, on a more rational note: Has the AI's city-congestion always been a problem?

There are ways of improving it, but you might have to change the game.

IIRC in the Call to Conquest mod the AI spaces it's cities quite wide, so they start overlapping later, on the 3rd or 4th rings. But the maps in CtC have far less bad terrain and ocean, and plenty of farm-friendly land too to keep them growing. There are other little things that help the AI settle wider too. Like free militia defence units when a city is built, making them almost invincible to barbarians early on when they're most at risk. Although the human gets these units too.

I see no reason why we shouldn't incorporate more game play changes into the AE mod though, if they help the AI. The wonders are already more balanced (although the AI still doesn't build wonders early enough), so perhaps it's a good idea to help the AI in other areas too.

[Martin Gühmann]

That fool, he has doomed us all! DOOOOOOOOMED!!!

Now, on a more rational note: Has the AI's city-congestion always been a problem?

So you mean AI expansion is now faster, because the AI isn't limited to just settling one city per turn? So you are overrun, aren't you?

And by the way what problem, so far I just have seen a few cities that hadn't no full first ring.

-Martin

[Hans Lemurson]

So you mean AI expansion is now faster, because the AI isn't limited to just settling one city per turn? So you are overrun, aren't you?

And by the way what problem, so far I just have seen a few cities that hadn't no full first ring.

-Martin

How would being limited to settling only one city per turn have any significant effect on expansion rate? "1 city per turn" should just delay the AI by 1 turn in settling; instead of 2 on turn 117, it settles one on 117 and the other on 118, shouldn't it?

Now if this was "The AI only sends out one settler at a time on a city-building mission" then I could see that having a noticeable effect.

[Maquiladora]

How would being limited to settling only one city per turn have any significant effect on expansion rate? "1 city per turn" should just delay the AI by 1 turn in settling; instead of 2 on turn 117, it settles one on 117 and the other on 118, shouldn't it?

Now if this was "The AI only sends out one settler at a time on a city-building mission" then I could see that having a noticeable effect.

There's no doubt it speeds up the AI at least a little bit, and if it's early (< t50) in the game it can snowball quite a lot. So it's a good fix either way.

I think Martin is exaggerating/joking a little bit anyway

Just thinking of ways to help the AI in the meantime. Using slic we can check the map size, number of players, and current map settings (the land/ocean slider), to get an idea of what settle distance we want our AI's to have in each game. Then we can create some new strategies each with different settle distances, and make the AI use them using slic.

The current minimum settle distance for almost all strategies is 4. That is, "eg. new cities must be founded at least 2 cells from nearest cities collection border". The only difference is in STRATEGY_SETTLE_LARGE which is a minsettledistance of 6. There are only two AI personalities (Ghandi and Caesar types) and the barbarians that use this strategy. All the other personalities use STRATEGY_SETTLE_COMPACT, which is obviously minsettledistance of 4.

In the AE mod I changed ISLAND_NATION and SETTLE_COMPACT to 3, so you might see this using that mod.

These distances aren't always accurate, so a minsettledistance of 4 may see spaces of 3 or 5, or even closer if two cities are settled simultaneously. Really bad land will make the AI look further and further away, and really good land (rivers on plains/grassland) seems to make the AI settle quickly once it finds such land. But generally you do have some control of the distance.

Over larger distances this type of control seems to work much better - a mistake of 1 tile over a distance of 8, means less than a mistake of 1 tile over a distance of 4.

Having just tested this some more I still see some unavoidable problems though. See the attached picture. In this game I set all strategies to minsettledistance of 8. You see some cities at 7, some at 8. All was going fine until Bedford and Winchester were settled (not from ruins or settled simultaneously). Why did the AI do this? The settler for Bedford came from London, so it's not like the settler was wandering for a long time, it just went straight there and settled. Did it decide the travel distance of another distant city wasn't worth more than the immediate effect of another river city?

[Martin Gühmann]

Having just tested this some more I still see some unavoidable problems though. See the attached picture. In this game I set all strategies to minsettledistance of 8. You see some cities at 7, some at 8. All was going fine until Bedford and Winchester were settled (not from ruins or settled simultaneously). Why did the AI do this? The settler for Bedford came from London, so it's not like the settler was wandering for a long time, it just went straight there and settled. Did it decide the travel distance of another distant city wasn't worth more than the immediate effect of another river city?

To figure out what happend you could replay this, but having on debugai and debugcell, so that you see the settle map and that you see the AI goals. And maybe you should also check what that AI uses of strategies.

And check whether you have missed any strategy.

-Martin

[Maquiladora]

To figure out what happend you could replay this, but having on debugai and debugcell, so that you see the settle map and that you see the AI goals. And maybe you should also check what that AI uses of strategies.

And check whether you have missed any strategy.

I double checked all the strategies minsettledistance and they're all either 7 or 8, for this test.

I tried again on another map (since I didn't have that save anymore), here are the results with debugcells. The big red numbers are just image numbers, not turns.

1: London is settled where the settler starts, the other two go off towards the red numbers West and North of London.

2: The first settler from London decides it wants to settle right near London. You can see the settle score is 740, that's well above the minimum settle score of 600 (this is the same for all strategies), but it's obviously ignoring our minimum settle distance of 7/8.

3: The second settler from London, again a good settle score, but totally ignoring our minimum distance. The settler East of Birmingham is going off toward the brown 1130 in the North-West.

4: Two new settlers from Birmingham and London now head off South. Strangely they have very close settle goals. The one without the yellow reticule in the screen shot is going to the green number 615. Notice they will arrive, and settle at exactly the same time...

5: They both settle at the same time and POOF, one of the settlers disappears. Didn't get on a transport, just gone...

So yeah, I seem to have found a different bug. Perhaps don't give the AI settle goals that are that close to each other. And why does the AI ignore minimum settle distance after the initial phase, this needs to be enforced stronger.

[EPW]

And they should settle on the darned river.

[Maquiladora]

Or the goods at least.

[EPW]

The placement of London isn't that bad, but Manchester, Canterbury, and Birmingham make me cringe.

[Maquiladora]

Birmingham is in a good spot for a min settle distance of 7/8, right next to the dye with lots of beaches. Cantebury should've been on the elephant and Winchester on the wheat, but I don't think goods make any difference to the settle score (yet).

What's strange is that Manchester and the poppy location above London only became settle targets after the other 7-8 distance targets were settled.

[EPW]

What file calculates all those city placement scores? Has anyone worked on a better formula?

[Maquiladora]

I'm looking for it now. I don't get some of the scores, it's certainly not based on just the immediate 9 tiles.

[Maquiladora]

Well it appears the radius for the score calculation is 2.

I'm wondering what would happen if we modified the score for tiles that are already being worked by another city. If we modify those tile's score to 0 they'd be a much less desirable, practically like settling near glacier. This should make it much less likely the AI settles within 1 or 2 tiles of another city.

[EPW]