AI gurus (i.e. Mark), is this possible?

[alms66]

I've been pondering this idea for some time now, and it seems to me possible, and also quite reasonable for our AI, it would basically be Civ2's AI on steroids...

Is it possible to make the AI perfect?

that is the heart of this post...

And I think it is possible (this could possibly mean no generic algorithms and no heuristic AI, i.e. less time spent on AI).

F.E. If the AI is in square 35,37 and it's opponent is in square 35,38; give it all the info it could possibly need, exact military str., exact supply lines, etc., then based on the difficulty factor (chieftain, warlord, prince, etc,[Civ2 explaination]) a player put in, give it a "dummy" variable to add or subtract from all it's math..., so for instance, if 350 military str. points were in 35,38, and the player chose deity (the hardest level), the AI would get no bonus or negative to what is there (still 350 military str.), it would be exactly what is there, whereas if he chose chieftain level (easiest) the AI might receive a +50%(just a guess) to it's calculations, making it 475 military str., and therefore easier for the player(i.e. the AI, wouldn't attack that strength).

Is this possible (I'm sure it is ), Is it feasible as an alternative? and is it good?...I'm not sure, I'm not an AI Guru...

Mark, can you enlighten us?

[Mark_Everson]

TK:

If I have your suggestion right, we're already doing it in even demo 4. I know its disgraceful that the civ2 AI didn't even go that far, but it is really trivial, as you say. It doesn't quite work as well in clash as in civ2 because we have simultaneous movement (and will have fog-of-war), and so you never know what exactly is going to be in that other square... The reason you need all the high-powered stuff for the AI is to deal with different levels of uncertainty, especially when looking many moves ahead. That's where the cutting-edge stuff comes in.

[Twinge]

Most of this strongly depends on how much the AI can cheat. As previously mentioned, how much it can cheat will likely be mostly controlled by the user, but what it sees and "how good it is" depends on how much it cheats. If the user wants it particularily easy, the computer will "cheat" in a negative manner... I assume that Mark is already aiming for a "Perfekt" AI, which will be nearly as good as a good player without cheating, so I do think it will be implemented.