Oh lord, it's hard to be humble... but you can help!

[jed]

One more bonus for the Iron Works. It lowers corruption.

I'll try to upload my mod in case anybody wants to down-load it directly without entering the changes themselves. I used version 1.16f.

[barefootbadass]

The ai basically uses similar tactics to what the ai of civ 2 used, although some of the ai's are better. Those that have mobile special units seem to do really well(before industrial) because they focus on those(esp. chinese). During the same period, those that rely on infantry suck. The germans seem to send nothing but foot soldier type units in these times, save an occasional knight or calvary.

Most abysmal is ai use of artillery. Mostly packed defensively in cities, this is easily exploited to capture loads of free catapults and cannons early in the game. I city can often have 3 or 4 pikes and 4 catapults. Take the city and the catapults(beats building them en masse) then give it back, and come back later when it has been restocked(hitting other cities with the help of the catapults/cannons in the meantime). Anyway, it might do better if stacks were in the game that the ai could use as well, then it could be programmed to make certain types of stacks for various attacking and defensive purposes.

As far as science goes, the ai should recognize that optional techs like music theory are limited in value because they are dead ends and zero in value once the wonder has been built. This shouldn't be too hard, the techs already have dynamic value in the trading screen based on how many civs have it, adding a new condition to the civ's evaluation should be very simple(assuming object oriented design). Seems there should also be a way for the ai to evaluate the value of skipping an optional tech to get on to another age. This is probably much more difficult, but not impossible, some kind of recursive call to all future techs in the tech tree should work. The hard part would be evaluating how this tech now compares to waiting to get to another tech later in the tree. That would be something that might need extensive playtesting to see if the ai reaches good decisions based on that model.

[rentgen]

It is a very interesting issue that is brought up in this thread, but i would like to note a few things.

First, concerning the AIs teaming up on the #1 player. Here we meet an interesting dilemma - if we look at Civ3 as a some abstract turn-based strategic game with a clear purpose - it should definetely be the way. Once someone has got even a slightest edge everyone else should forget their own quarrels and turn their attention to the leader to bring him down. On the other hand, I think Civ3 tries to model "real world" and if we think about it it is really not so typical for the real world civilizations to act this way. Obviously, be it so all the world would turn on the USA by now, as the current "leader". Why this is not happening - because the purposes are slightly different.
Personally I, and i think many other players, find it particularly attractive that Civ3 does try to model real world. Even though we would like more challenge as well in the game. However, i would wish to see a game with much more realistic civilization modelling aspects. Like possibility to have small but flourishing countries like Andorra that lives on tourism mostly or Kuweit that (at least before the war) was able to make all its' citizens happy by simply exporting oil.

Now, to implement all this a program would have to be several orders of magnitude more sophisticated than the Civ3, so as much as i wish to play something like that i realize that i won't be able to for quite a long time. Taken into account the speed Civ3 plays, a program like that would require a supercomputer to run.

Ok, so the speed is the second issue i guess. From my experience in IT, the main reason for slow game running is AI. Not graphics, sound, holding the "shift" key or whatever else that was discussed here in this forum. After all, it is a business and if they are to make any money - it is always a tradeoff between complexity, playability, time-to-market and many other things. Money here are made by selling as many copies as possible, and they sure could create a much more challenging AI, or make things so that a tank would never be killed by a hoplite - but at the expense of the game size, running speed and time-to-market.

I can assure you, being an IT project manager for quite sometime, - developers and designers always have tons of great ideas on how to make things "cooler", "nicer", "more challening" etc. But if they were allowed to do all that, they would spend some ten billion $, some ten years, and even *if* they do create something playable (workable) at the end - it *will* require a supercomputer to run and would cost a couple of millions$ a copy. I am exaggerating, but that gives you an idea

Ok, now for some constructive things
What would be interesting is to make up a list of AI improvements that *we think* would not affect performance or game size significantly, but improve the gameplay and give more challenge for those who want that. Good examples that i saw on this thread are not researching deadend techs without the need and not building obsolete units. It is so obvious and so easy i wonder why it wasn't done With the units btw, can it be that they build swordsman instead of tanks just because they are lacking a resource?
Anyway, do put your suggestions if you feel like it and then we can compile the list and give it with comments to Firaxis.. And no, i am not affiliated with them in case you were wondering

[barefootbadass]

Originally posted by rentgen
With the units btw, can it be that they build swordsman instead of tanks just because they are lacking a resource?

Swordsmen and longbowmen still get built even with better units around, primarily because they remain an option forever. Of course units that are upgradable get built if the civ lacks the resources for the better one.