AI has many problems and needs to be improved

[fomalhaut]

my dissatisfaction with ver 2.5 AI

1. They don't attack an enemy city by air units.
(AI needs to recognize an effectiveness of fighters or cruise missiles.)
2. They don't attack an enemy ships by triremes or caravels.
(At least ships carrying expensive cargo or fragile units who are on the coast should be attacked.)
3. In the first place, they are less aggressive than ver 2.3 AI because they are very afraid of making their city empty.
(We could easily fish ver 2.3 AI. So they were idiot of course. But ver 2.5 AI are timid and also idiot.)
4. They think nothing but producing city defenders and ignore some useful buildings. (library, harbor,market,temple and so on)
5. Because of too many city defenders, their cities soon suffer from lack of production point.
(They should disband unnecessary units.)
6. Ignoring Freeciv help ("need an escort"), they independently operate artillery (catapult, cannon and artillery).
7. They don't understand the importance of terrain defense bonus. So their units land at flat terrain instead of hills or mountains.

When will AI satisfy us? How long should I wait?
Do you plan to improve them?

[cazfi]

4. They think nothing but producing city defenders

Interesting. What ruleset and settings do you use? My experience until some recent changes (and quite much even after them) is that most AI cities go totally undefended as they concentrate in completely different things.

[Ceres]

Chiming in here because I've experienced the same thing - in my experience, as soon as an AI starts lagging behind its neighbors technologically, it starts mass-producing defensive units and literally nothing else. This is with close to default settings, standard ruleset and hard AI - I've seen the same thing happen in civ2civ3 as well though (except with periodic mass starvation because of the food upkeep for additional units).

[fomalhaut]

Unfortunately I don't know the best way to verify the difference in AI's behavior.
But whenever I play ver 2.5 classic ruleset, I always notice that AI continues to anchor warships at their city.

In addition, I want AI to avoid changing research goal.
Evidently this is a waste of time for a development of AI country.

[cazfi]

I always notice that AI continues to anchor warships at their city.

Thanks for giving such specifics. That might be something gone unnoticed when I have observed AI players. This is unlikely to be AI considering them as "defenders" (ships make bad defenders), but there's some reason it's not using them properly. Do you have an savegame where one could check such a situation?

[cazfi]

In addition, I want AI to avoid changing research goal.
Evidently this is a waste of time for a development of AI country.

Actually i don't know why the AI code bothers with the research goal at all. It should set the current research every time anyway. The (longterm) Research Goal is something I consider mainly a micro-management reduction helper for human players (assuming game with low researchcost so that it doesn't so much matter in which order you research as you will get the tech in a couple of turns even if something else gets researched first)

[cazfi]

Made a ticket about the warships case so it won't get forgotten just to an old forum post: bug #24099

[fomalhaut]

Made a ticket about the warships case so it won't get forgotten just to an old forum post: bug #24099

thx.

I have observed AI-only game again and noticed some problems.

Overview and progress of the game:
6AI (hard) ,tilesperplayer=110 (42*56 tiles), generator =RANDOM, startpos =VARIABLE
Only Westphalian and Pictish shared an island. Everyone else had an own island.

T35: Belgic were producing archers instead of settlers. Pictish were producing settlers in spite of being war with Westphalian.
T50: Although Guanche had the biggest island, they had only 3 cities.
T80: Westphalian...3 cities and 16 archers /Belgic... 6 cities, 10 warriors and 16 archers
They had an excessive number of troops and suffered from lack of production point.
T160: 4 AI players had "gunpowder". But only 1 city among 90 cities in the world had Library. Harbor or Marketplace didn't exist.
T270: Every AI continued to anchor warships(Ironclad, Destroyer or Cruiser) at their city.
T300: There were 91 cities in the world. (90 coastal cities) Harbor...1, Library...12, Marketplace...0, Port facility...66
T337: Belgic built Isaac Newton's College but they had only 1 University in their 16 cities.
T370: Belgic and Khwarezmian were building SAM Battery or SDI Defense. Why was it necessary?
Harbor...1, Library...19, Marketplace...0, Factory...0
T400: Stalemate. AI didn't produce any missiles and continued to anchor Battleship.

Subsequently I was worried that this server settings were unsuitable because the map was smaller than default one.
But even after I set map size as 4 (4000 tiles), similar problems were found.
*Some AI don't try to produce Settlers. And they continue to produce archers even when their enemies are in the different island.
*AI change the technology that they are researching very frequently. Therefore their research doesn't complete and they are wasting the time.
*So many warships are anchored and converted into museum ship.
*Furthermore they don't use Howitzers against adjacent enemy country. There were foolish AI, who had 60 Howitzers and never tried to use them.
*They love to produce AEGIS Cruiser instead of Cruise Missiles. Every turn many AEGIS Cruisers try to suicide attack and sink.
*AI are indifferent about city improvements. After Apollo Program was built, they hardly have Factory in their cities.
*AI don't attack effectively and don't improve their cities. Game is easy to fall into stalemate.

The most incomprehensible question is why AI don't try to attack enemy cities with air units.
Is it an intentional design?

[fomalhaut]

I understand how to upload a savefile.
Please check Kenyan Howitzers and warships.

[djconklin]

>as soon as an AI starts lagging behind its neighbors technologically,

I'd like to see that! I'm playing 2 dozen at novice level and they are two steps ahead of me AND building the wonders for that tech!