Resource icon

AI+ v13.1

Hi!

I have been using your mod for a time now and in this version and the 9th and 10th as for as i remember has this "problem" with other civilization. They are not really expanding. I don't know, I use larger worlds 2.31 mod (maybe this couse the issue) but the AI give a crap about expanding :D Turn 250 and Norway has 2 cities. Maybe China has 5. This is my actual game with v11. My prev games turn 4-500 7-8 cities tops and those are belonged maybe 2-3 civilization. The rest of them has 4-5 tops.

Difficulty: Emperor
Map: Continets
 
So I've done a boatload of testing. And well. I can only come to the conclusion that settling on watery maps below emperor is simply not going to work this patch. It doesn't work on vanilla, nor does it work with extreme values on tweakable numbers. Increasing the valuation of spots, even to ridiculous numbers doesn't work, nor does boosting the actual yields of water. Many civs simply refuse to settle after some point, occasionally without ever settling at all. Only civs lucky enough to find themselves on larger landmasses end up settling, regardless of what I do. They just don't start any settle operations, which means they'll never even consider building settlers.
 
Even on emperor the issue appears to some degree, it's just not 75% of all civs on the map. And yeah I think it's mostly the two cities. Not only do they have two starting spots to build further settlers from, but extra cities also decrease the required threshold to settle new spots (probably to make them fill the terrible spots up lategame). Add the extra production bonuses so that they'll run out of alternatives for settlers to build, and add their extra barbarian killers and escorts (barbs are the most common settle problem besides this bug) and it all adds up to much higher chances of settling. And even if they won't build additional settlers, well, at least they still have 2 cities, which can still be sort of acceptable on maps that barely have any space anyway.
 
It's a strange issue for sure. On watery maps below emperor, even if there's space, the AI often refuses to build settlers. My hypothesis had always been that they consider most spots to be terrible spots because of the lower coast yields and end up not meeting the minimum threshold required to build a settler (no settler 'operations' and 'contracts' are ever started when this happens). I've tried exploring that by increasing their spot valuation and by decreasing the threshold on watery maps but with no success, so I'm not sure if that is the reason. I'll try to have another look at it when I have some time. Perhaps there's some alternative way of fixing it that I haven't found yet

Meanwhile, I'd recommend either playing on emperor+ or on land heavy maps like pangea/continents.

I know from other modders that there is a "fertility" rating in the game, that is also used to decide viable starting locations for civs. Assuming this is also what makes AI settle, could a factor for coastal be added/buffed so that coastal cities are seen as viable more easily by AI?

Also, just thinking about my last game, which I played without this mod but on the Summer patch, I went on an early conquering spree on my continent, and one of the cities I conquered (Emperor level, so settled by the 2nd starting settler) was actually sitting on the edge of a peninsula where there were maybe 10 land tiles in range of the city, while there was a much better spot (river, almost exclusively good tiles, etc) to the other side from the starting position of the AI, so I'm kind of wondering why it chose to put the city there, if the other city had far less coast?
 
I play another v11 game, this time pangea, AI civs have 2 cities, still don't settle coast...
(the 2 cities isn't too bad, as it is tiny map with 6 civs and lots of CSs, only that AI still leaves good coastal land unsettled)
 
I know from other modders that there is a "fertility" rating in the game, that is also used to decide viable starting locations for civs. Assuming this is also what makes AI settle, could a factor for coastal be added/buffed so that coastal cities are seen as viable more easily by AI?

Also, just thinking about my last game, which I played without this mod but on the Summer patch, I went on an early conquering spree on my continent, and one of the cities I conquered (Emperor level, so settled by the 2nd starting settler) was actually sitting on the edge of a peninsula where there were maybe 10 land tiles in range of the city, while there was a much better spot (river, almost exclusively good tiles, etc) to the other side from the starting position of the AI, so I'm kind of wondering why it chose to put the city there, if the other city had far less coast?

The starting spot fertility value is calculated in lua code and not related to the settling evaluation system. I've done some tests that include buffing coastal yields, and while it makes them more likely to settle on coast if they decide to build a settler, it doesn't seem to affect whether they'll consider building a settler. The main bug here is that they don't start 'settle operations', which means they won't build settlers or use already available ones.

Your peninsula settling observation could've been (but I can't be sure without seeing the map/game obviously) an instance where they recaptured a settler from barbarians. They appear to use a different system if that happens, frequently settling in place regardless of the quality of the spot. Alternative explanations may be that there's hidden oil in the water somewhere -which they can detect before getting the tech- or that the better spots contains desert, which is valued lower than coast tiles or that the civ is England and this spot is on a different continent than the better spot.
 
The starting spot fertility value is calculated in lua code and not related to the settling evaluation system. I've done some tests that include buffing coastal yields, and while it makes them more likely to settle on coast if they decide to build a settler, it doesn't seem to affect whether they'll consider building a settler. The main bug here is that they don't start 'settle operations', which means they won't build settlers or use already available ones.

Your peninsula settling observation could've been (but I can't be sure without seeing the map/game obviously) an instance where they recaptured a settler from barbarians. They appear to use a different system if that happens, frequently settling in place regardless of the quality of the spot. Alternative explanations may be that there's hidden oil in the water somewhere -which they can detect before getting the tech- or that the better spots contains desert, which is valued lower than coast tiles or that the civ is England and this spot is on a different continent than the better spot.

Siesta --- have been playing a game with AI+ 10.5 (not 11), and have had no issues at all with civs having multiple cities (in fact, some are just spamming cities which has made the game fun). In the late middle ages, on immortal, and no problem with settlement.

On a separate machine, I ran AI+ (v11) with AI Autoplay to see the settling behavior. Continents map, by 1800s, no civilization had more than 4 cities, most only had 2. I'm going to run a few more tests as well (upping the difficultly level maybe to immortal or diety, vs. prince or king) -- and will report back. I will also run similar scenarios on AI+ v 10.5 to see if there is a difference there.
 
Siesta --- have been playing a game with AI+ 10.5 (not 11), and have had no issues at all with civs having multiple cities (in fact, some are just spamming cities which has made the game fun). In the late middle ages, on immortal, and no problem with settlement.

On a separate machine, I ran AI+ (v11) with AI Autoplay to see the settling behavior. Continents map, by 1800s, no civilization had more than 4 cities, most only had 2. I'm going to run a few more tests as well (upping the difficultly level maybe to immortal or diety, vs. prince or king) -- and will report back. I will also run similar scenarios on AI+ v 10.5 to see if there is a difference there.

Thanks for doing some tests! It's really helpful. Testing always takes so long..

The AIs ought to mostly settle at intended rates (pretty high rates) on emperor+ and anything lower should also be fine on land heavy maps such as pangea and the earth map. If you find any settling problems there or any other major issues it'd be great if you could share the map and game seeds.

I'll keep trying to look into prince/king settling on watery maps, perhaps there's still something I'm missing.
 
Found it!

Turns out there were two problems with identical results. One only present in vanilla (poor spots not meeting a settle threshold), and one only present in AI+. A priority of 4 instead of a priority of 3 on an operation made for naval maps made them not settle. I hadn't looked much at the details of that operation because of the very similar problem in vanilla.

The results are quite a bit better on watery maps, much more in line with appropriate settle rates. They'll even do some island hopping. It's by no means perfect still, especially when there's barbs around, but it's a step. The one exception is extremely tiny island starts where they can still get stuck at 1 city.

Uploading the new version now
 
Siesta Guru updated AI+ with a new update entry:

AI+ v11.1



Read the rest of this update entry...

Ran two quick AI-Autoplays with AI+ v11.1. Standard sized maps, standard settings, no other mods other than Autoplay and AI+.

Was ONLY focusing on Settler behavior -- nothing else relating to combat, AI, etc.

First one was Islands on Immortal level. Settler behavior appeared to be normal, and some civs founded cities off of their home island. Ran autoplay until the middle ages, by which time most inhabitable areas were taken. My "host" civilization, Nubia, was overrun by an invading AI.

Second one was Islands on Warrior level. Settler behavior was not nearly as normal. Took longer for cities to be founded, and each civilization ultimately got to 3-4 cities. Only one, Russia, founded a city well away from their home island. Ironically, they were the most delayed in getting moving on building cities.

I can say that settler behavior is better -- I do think that at levels below Emperor, the AI is just simply too dumb to be competitive enough, and hence, slow on Settling. At least that is my theory and I am sticking to it :)

I'm going to now experiment with Quo's mod, which is the one I really enjoy playing with -- will do similar autoplays to see if there is anything wompy with the combination of the two of them.
 
Alright, thanks for sharing! I'll continue to work on it a bit more. Though I agree, that below emperor it's essentially hopeless in any real competitive sense without proper code reworks. It's still nice to have them somewhat functional for the players who prefer more relaxing games. Is anyone also playing below prince btw? I very rarely do any testing on that.
 
v11.1 - smoother mod King level, tiny continents, 6 civs, 2 Css, no barbs, quick speed (330)

turn 70, I met 3 AIs -

I have just put down my 4th city (and I forced it quick)
Gilgamesh and Monty has 3 cities, Gorgo has 2

AI is also great at having science and culture, army too

I think AI settling is much better... :)
 
Alright, thanks for sharing! I'll continue to work on it a bit more. Though I agree, that below emperor it's essentially hopeless in any real competitive sense without proper code reworks. It's still nice to have them somewhat functional for the players who prefer more relaxing games. Is anyone also playing below prince btw? I very rarely do any testing on that.

Siesta -- I used warlord just to be on the extremes. Running one with Quo now on Immortal to look at behavior. I ran one last night with Quo on Warlord, and by turn 250ish (late middle ages), everybody had 3-5 cities.

One of the next things I want to look into is districts and if the AI is building enough districts. On a watery map, you would expect a lot of harbors.

g
 
Glad to hear soma :) Hope they'll be a challenge!

Gorgo attacked me with a varied army of 2 hoplites, 2 horses, archers, warriors and a ram... She went for my 2nd city in the middle and took it quick (!)

I was kind of behind in tech and army: warriors and slingers... It took me much effort to get back that city and was lucky to get peace for 20 golds...

EDIT:

it is turn 124 now, AD 1260:

I have 5 cities,
Gorgo has 5,
Trajan has 5,
Gilga has 4,
Tomyris has 3
Monty has 2

tech:
I am now next to last in tech with 16 techs and 20/turn
civs have 20-21 and ca. 30/turn - almost all civs are in the Medieval era
well, Gorgo now makes 70sci/turn

culture is kind of the same as science... here Gorgo is in Renaissance...

I play Quo's Combined mod...
 
Last edited:
Siesta -- I used warlord just to be on the extremes. Running one with Quo now on Immortal to look at behavior. I ran one last night with Quo on Warlord, and by turn 250ish (late middle ages), everybody had 3-5 cities.

One of the next things I want to look into is districts and if the AI is building enough districts. On a watery map, you would expect a lot of harbors.

g

Update -- Autoplay with Quo's mod -- settler behavior is mostly good on islands map. However, two nice islands are not populated by turn 250 (1750s ish) on immortal.

Decent mix of districts, tho not seeing many production districts by the AI.

If you want to run the same map -- you start as Kongo, it is the big island Kongo is a part of that goes mostly uninhabited, as well as the island above.

1329294617 -- game
1329299631 -- map

I'm going to run same game/map but without Quo's mod to see if I get any appreciable difference.
 
Hey Siesta, it's great you are back!
I did some testing of the latest version with slight amendment (slightly increased desire to built Science and cultural districts, as well as naval and air units):
My observations:
1. Settling - very good, usually by renaissance/modern era the land is filled with cities, also AI doesn't just try to place new city 3 hexes away from the old, he can build 4-5 hexes if the closest land plot has desert or so.
2. District building - very good, I slightly increased cultural and science district desire to built (as I play with mod which reduces culture and science from pop by 50%). I've seen AI building all districts and especially I like the fact that I see production and market district built in many cities.
3. Land warfare - good, apart from many stupid moves AI does. I also play with Rocketboots and AI siege help so it help AI moves quicker to cities and conquer them. Cities even fall in latest eras and also AI tries to use support units (mainly rams) which helps him a lot.

4. Naval warfare/amphibious landing - oh god, poor AI doesn't know what to do with ships. It either sends them for scouting or just keep them at bay in the harbour. I tried to add ships to a normal city attack groups (by increasing max number) but so far I haven't seen them in action. Siesta, is it possible to create special operations "Naval assault" or something like this? I am really missing good naval fighting from civ5.

5. Diplomacy/war declaration - AI stopped steamrolling city states at the beginning of the game. Now it is happening slowly throughout the game. Also AI now is less aggressive, especially in the latest eras, I would say it would be good to increase AI desire to wage wars in modern eras

Overall, version 11.1 is probably the best one so far, on a land maps it gives you lots of fun fighting against AI hordes who is now also build districts and try to be ahead in science/culture.
 
Back
Top Bottom