Let me share my experience of playing v.6 of the mod:
Map: Fractal
Size: Large
Difficulty: Emperor
Player: Cleopatra
Mods: AI + V.6, Improved harbour, Smaller boosts (33%), Slower techs (based on Gedemon mod, greater costs for techs, civics and GP depending on the era), other small changes (50% faith costs increase for religious units, 10% discount on unit costs, 10% increase of building/districts costs.
Speed: Epic
Turns played: 250 (end of medieval era)
I monitored AI behaviour in individual components and got the following:
1. City location
Before playing this game, I played as France on Pangea map in vanilla and noticed significant improvement of AI settling after fall patch. I could barely find any city which would be built not on river/lake or coast.
With the mod enabled I noticed that AI in general settling cities ok, however sometimes it settles 2 or 3 tiles away from seaside or 1 tile away from river, e.g. please see Sparta, Valencia, Solikamsk or English cities screenshots. There might be some logic in it, like settling closer to luxuries or natural wonders but I think same result could be achieved if settling on a seaside tiles.
Another thing is that after good initial settling (4-5 tiles away from each city), AI again started to put cities 3 tiles away from each other. It's ok if there would be lack of available territory but there were still better spots
Also, I have a feeling that AI settles too fast and give too much attention to settling new cities. It's still medieval era and AI almost filled all suitable territory. Luckily I was separated by mountains and sea from other AIs but in my previous games AI was constantly trying to encircle me with their cities.
I am not sure how vanilla AI would settle on island maps but overall its settling improved a lot after fall patch and I would keep it untouched until we find some flaws in it.
2. Settlers behaviour
In general AI tries to accompany a settler with military units however I am not sure whether he uses escort function. In my game I saw a Roman settler rushing to the settling point and manage to settle first in this spot. Then for next 10-15 turns this settler stopped at my border doing nothing. He was unescorted although he had two warriors in neighbour tiles, so in theory I could declare quick war and steal it. It's good that finally AI found another spot for this settler, but it worries me that it took him so long.
On a positive side: I haven't seen hordes of settlers which we had before. AI usually has one, max two settlers who are trying to find good spot.
3. Diplomatic AI
I can't say much here because I had one big war where after killing 4-5 units in attempts to come to my city AI offered a peace with 2-3 gpt. Very fair peace tbh, in vanilla (after patch) after taking one of AI cities he offered me ridiculous 102 GPT. How he got this money I don't know. Overall I saw wars between AIs which lasted relatively long, prob 30-40 turns.
Got two stupid denouncations after sending envoy to city-state but that's probably ok.
4. Military AI
This is where we got a problem. AIs have really sizeable forces but they are not really using them even when going for aggressive wars. I've seen the following picture: Sumeria besieged Roman city (no walls), they had 4-5 war carts + allied city-state units. Romans has some forces as well but they kept them a bit away from this city (prob defending their capital against Greece). Sumeria managed to damage city a bit and then all units pulled away from the city for 2-3 tiles (!) instead of trying to finish it. The same situation happened several times until they finally had a peace. As mentioned before, AI taking another AI's city is the event of millennium.
If possible, I would like to see the following logic:
a. AI accumulates 10-13 units with 3-4 siege units.
b. AI makes huge push without concerning about his unit with priorities destroying player's ranged units, encircling the city and taking it.
c. If AI left with 3-4 units from his initial force and city still has walls undamaged or there are many player units around, then push away for 4 tiles to your closest city. If player's city has no walls or there aren't many units around, keep attacking.
I like the idea of AI trying to save his units but it's really difficult to implement. AI starts to push even its lightly damaged units, messing battle orders and forcing other units to stop and give way instead of doing proper attack.
I would leave this option only for AI promoted units which are heavily damaged.
5. Districts, improvements and wonders
This is where I would like to go one by one for each AI I met:
a. Sumeria
Improvements: Insane number of ziggurats, almost in every single cell (hell, Ner-Zul would be pleased seeing this!).
By the way, does anybody know how to put a mod which would prohibit all unique improvement to be adjacent to each other (apart from Great Wall and Roman fort)? I am pretty sure the reason why Sumeria is ahead in Science in most of the games is purely because of huge number of ziggurats.
Also, Sumerian cities have 4-5 population (because of lack of normal improvements) but they improved 10-12 tiles... why?!
Districts: The only district I see is I science district (although they get 3.1 GSP per turn, so maybe they have a second one). Nothing else, despite it is turn 250 now (end of medieval era).
Wonders: They are trying to build now Aztec wonder on a lake, but it will probably take them ages
b. Russia
Improvements: lots of farms, again they don't need so many for now, other resources covered normally
Districts: Lots of lavras + 1 cultural district, 1 production and 1 military district in the capital. Also they are constructing now 1 commercial hub and they got 2 or 3 harbours.
Wonders: Managed to get Petra.
Overall Russia was probably the most developed and close to vanilla district building. Mb there were too many lavras and lack of Science/Commercial/Entertainment districts
c. Spain
Improvements: again, tons of unnecessary improvements at that point in the game
Districts: No districts at all, even no faith district. AI hadn't been at wars, the only thing he was doing was building army, improvements and settlers.
Wonders: nothing
d. Rome
Improvements: fair number of improvements, not too much, I would even say Roman AI should have more but probably they didn't have a chance to build them because of many wars with Sumeria and Greece.
Districts: a military district in Rome and bath district in another city
Wonders: Tried to build Chichen-Itza but was too slow, somebody else built it.
e. Greece
Improvements: As usual, lots of mines and other stuff and they got La Venta allied, so lots of stone heads as well (they should be non-adjacent
).
Districts: Prob one or two cultural districts and 2-3 faith districts
Wonders: Nothing I would see
f. Others and city-states
I see America and Scythia got 2 great merchant points per turn so they are doing something at least. I see only one city-state who is trying to build faith district, all other CS don't have a district (despite normally they try to build it asap)
Here are all screensots:
Overall impression: Maybe I am too early in the eras but comparing to vanilla game I have an impression that AI is very reluctant to build districts and develop their cities. Instead he tries to build a lot of settlers and workers to make insane number of improvements. I haven't seen any entertainment district in the game (despite building two myself). I have a feeling that in vanilla AI is much better at that component and really concentrates on district building, so if possible I would return pseudoyields to vanilla in both peace/war situation and see what would happen. Also, preference towards unique districts (as well as unique improvements) is a bit too high in my opinion.
I hope all those points above will serve as a good feedback in the attempts to improve AI behaviour.
Siesta, keep up your work, AI+ is the only hope for us (beside Firaxis) to get the challenging AI opponent.