The challenge of creating a competent AI

[berto]

I think we all agree that creating a competent AI to play hundreds or even many dozens of "turns" is extremely difficult to impossible.

I have a solution to this problem: Just bundle with the game dozens of shorter, AI-capable scenarios, both historical and hypothetical, scenarios lasting just a half dozen, or dozen, or a couple dozen "turns".

Although ideally I'd love to face an AI capable of fighting the entire war in an intelligent fashion--and the designers should try their best to achieve it in a reasonable time frame (say, within the next 3-6 months?)--I have to admit that I will simply never have the time to play a campaign game beyond a few dozen "turns" anyway. Wargaming is a hobby, not my life. Besides, I have many other games that will compete with WITP for my attention (Uncommon Valor, SPWAW, Europa Universalis, TOAW and other Talonsoft games, etc.), making it even less likely I will ever play a WITP campaign game anywhere near to completion. So, for me, a very good campaign-game AI is moot.

So, I suggest:

--For the campaign game, have a really good PBEM option, and a half-way decent AI.
--For the (hopefully) dozens of short-term scenarios, try for a darned good AI (and a good PBEM option).

BTW, I will never play PBEM. I would drive my opponent crazy with impatience waiting for my next move. I'll go weeks at a time when I simply don't have the time to wargame. No, AI is critical for me. If the game has no AI, I won't buy it. (In 40 years of wargaming, easily 98% or more of my games, both computer and tabletop, have been solitaire.) If the game's AI is weak, I won't play it for very long.

[Von Rom]

BTW, I will never play PBEM. I would drive my opponent crazy with impatience waiting for my next move. I'll go weeks at a time when I simply don't have the time to wargame. No, AI is critical for me. If the game has no AI, I won't buy it. (In 40 years of wargaming, easily 98% or more of my games, both computer and tabletop, have been solitaire.) If the game's AI is weak, I won't play it for very long.

Agreed.

Is there anyway to have BOTH scripted tactics/strategies and a program to allow the AI to "learn" as it plays?

The new game "Superpower" is using this type of AI. It's not out yet, so it's hard to say how successful this approach might be . . .

Also, it might be helpful to get the experienced players' input of PacWar as to the strategies they use for both the US and Japan for each year of the war, and program in some of their devious strategies for the AI

[Ron Saueracker]

Any further thoughts or decisions made regarding AI in WITP? I've read in UV forum that UV AI will be the model and UV will act as basic test bed.