Specific suggestions for improving the AI

[Selkirk]

Here are some of my specific suggestions for improving the AI...

COMBINED ARMS

A) The computer needs to be able to stack together multiple units and move the stack as a formation. The Age of Kings AI is good at creating combined arms formations. Formation support would also be an excellent addition to the UI. A formation in civ III would occupy only one square.

B) The computer should use catapults / cannons / artillery in offensive strikes. (This may require that the computer have an understanding of formations, as defined above to defend the bombardment units, although it seems to do OK with settlers.)

NAVAL INVASION

C) The computer should not allow a Galley to set sail unless it is full, or a troop transport to sail unless it has at least 5 units in it. (formation support might help the AI to do this -- sailing a formation, rather than individual troops)

D) The computer should use only its best units for naval invasion.

OBSOLETE UNITS

E) The computer should reserve a small budget (say 1% of its total income, or 2% for militaristic civs) to opportunistically upgrade units that happen to be in cities that have a barracks.

F) If the computer is at war and has a city that is threatened and has a barracks in that city and has the cash, they should pay to upgrade all the units in that city.

CULTURAL WARFARE

G) Founding a new city 1 square away from a influence size 1 city should result in a major degradation of the relationship between the civilizations. This crowding should be recognized as a "cultural attack." (The new city is inside the existing cities resource collection area.)

H) Founding a a new city that causes some squares to shared with an existing city of another civilization should result in a lesser degredation of the relationship between the civilizations. (Crowding is offensive.)

CITY GOVERNOR

I) When a cities growth is limited by lack of aqueduct or hospital and the governor is managing moods, the governor should re-apportion the population to maximize production or trade and produce near 0 food.

RESEARCH

J) The type of civilization (scientific, commercial, etc.) should factor more heavily into technology research choices and technology trading valuations:
* Does it lead to a preferred form of government for that civ?
* Does it allow an improvement or unit to be built that matches that civilizations attributes.
* Does it allow a wonder to be built that could trigger a golden age for that civilization?

Comments? Other specific suggestions?

[Code Monkey]

COMBINED ARMS

A) The computer needs to be able to stack together multiple units and move the stack as a formation. The Age of Kings AI is good at creating combined arms formations. Formation support would also be an excellent addition to the UI. A formation in civ III would occupy only one square.

B) The computer should use catapults / cannons / artillery in offensive strikes. (This may require that the computer have an understanding of formations, as defined above to defend the bombardment units, although it seems to do OK with settlers.)

The attack AI is much better than earlier civ games but there is a problem of the AI apparently only understanding offensive unit versus defensive unit. If they implement a stack idea for units then what it should do is draw from a list of suitable formations, e.g. 4 infantry, 2 cavalry, and 4 cannon OR 10 tanks OR 10 cavalry. I think the settler works because there's something in the AI which says always send a settler out with a defensive unit - that wouldn't transfer well except possibly to artillery.

NAVAL INVASION

D) The computer should use only its best units for naval invasion.

I would think this one would be open to exploits - the computer currently seems to send a representational mix of its troops - if the upgrade AI were tweaked more the problem of inferior units being sent would be lessened. Besides, sometimes it's a good way to get rid of older troops, send them out on a cannon fodder mission.

OBSOLETE UNITS

E) The computer should reserve a small budget (say 1% of its total income, or 2% for militaristic civs) to opportunistically upgrade units that happen to be in cities that have a barracks.

F) If the computer is at war and has a city that is threatened and has a barracks in that city and has the cash, they should pay to upgrade all the units in that city.

Agreed totally, the computer definitely needs to upgrade more aggressively. It costs the same upkeep for a warrior and a tank - which is better for a player? I'm not sure which is the culprit, the weighting for the CP to upgrade or the ability of the CP to create a decent economy which would allow them to upgrade, but this needs to be tweaked. There's always a gap where I know I can attack the AI because I just upgraded everything and I know it will be a long time before their forces catch up.

CITY GOVERNOR

I) When a cities growth is limited by lack of aqueduct or hospital and the governor is managing moods, the governor should re-apportion the population to maximize production or trade and produce near 0 food.

This is definitely a key problem with the current governor - no matter what it's set it on it will always seek to maximise food. I agree wholeheartedly with this one: the governor needs to understand how to push food up until the maximum population and then eliminate excess food in favour of production. This would both help the AI and cut down on my micromanagement.

RESEARCH

J) The type of civilization (scientific, commercial, etc.) should factor more heavily into technology research choices and technology trading valuations:
* Does it lead to a preferred form of government for that civ?
* Does it allow an improvement or unit to be built that matches that civilizations attributes.
* Does it allow a wonder to be built that could trigger a golden age for that civilization?

This is already implemented - it might benefit from some tweaking but I know I've read things from the designers as well as observed the prices various civs will pay me for a tech that show it's in place.

[Venger]

1) Suck less.

Seriously though, part of the AI's problem are the "improvements" to the game. God forbid the AI not be blessed with every resource - without resources the AI is absolutely CRUCIFIED.

The AI also doesn't seem to build as much culture as the human player (a rough, and maybe inaccurate, observation).

The AI is also absolutely HORRID at defending itself. I am constantly taking cities with battleships at anchor - WTF are they doing at anchor? Just sitting in the dock for me to take. I destroy bombers that have never attacked me, and usually a few units, often far worse than it should have (often riflemen). That's sorry. The Civ2 AI at least nutted up and had bombers and stacks of Mech. Inf. to blast through, and it WOULD use every available unit (sometimes when it shouldnt...)

Simply stated, not only can I outhink the AI, which is nothing new, but the AI rolls over like a 5 centavo whore on a Phillipino naval base. The AI isn't a challenge, the awkward and obtuse game system is (city reversion, corruption, goofy combat).

The AI ought to have modular decision making routines that are simple yet create a Civ that works - build, fight, etc.

Venger

[Moraelin]

As I've said in another thread, IMHO the way to do it would have been to allow people (and the AI) to still build units even without the resources, but at a cost penalty. Say, +50% shields cost for each resource that you are lacking. You can think of it as getting that stuff on the black market, or using other materials, or whatever.

But in gameplay terms it would mean that who has the resources still does have an avantage. And who doesn't, could still build modern units, only not as many. I.e., the AI could still build riflemen even if it isn't smart enough to grab a source of saltpeter, or those Rubber Tanks even if it doesn't have a source of Latex.

Also, it could have helped if the AI understood the concept of trade. Like maybe instead of attacking with obsolete units to get access to Saltpeter or Latex, try to import those resources first. Even if the player doesn't want to trade strategic resources (though some of us would take the money), maybe some other AI country is willing to make a profit out of its 5 extra saltpeter tiles.

As it is, the AI's idea of trade is simply trying to piss off the player, for the fun of it. Noone will accept a trade unless they're totally ripping off the other, extremely rarely propose a trade, and then will wantonly cancel it even though they were getting Gems and 100 gold per turn for their Furs. They also almost never connect to someone else's roads, except by accident. (E.g., if the border moves because of war or culture.) Which isn't good for trading, either.

So trade between AI countries is non-existent. Countries which could have prosperred and whooped my rear if they cooperated, will rather stay self-insufficient.

Maybe then there wouldn't have been a need to lobotomize combat, to make up for the AI's love of Impi (all the way into the middle ages.)