Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel-eecis!news.mathworks.com!uunet!inXS.uu.net!news.new-york.net!actcom!news
From: bruck@actcom.co.il (Uri Bruck)
Subject: Re: On making more AI-friendly wargames
X-Nntp-Posting-Host: p1.haifa2.actcom.co.il
Message-ID: <Du4FCF.I63@actcom.co.il>
Sender: news@actcom.co.il (News)
Reply-To: bruck@actcom.co.il
Organization: ACTCOM - Internet Services in Israel
X-Newsreader: Forte Free Agent 1.0.82
References: <4q5l8q$h0a@fountain.mindlink.net> <4qp840$g9t@newsbf02.news.aol.com> <4r02it$kph@fountain.mindlink.net>
Date: Sat, 6 Jul 1996 12:10:14 GMT
Lines: 44

Alan_Douglas@mindlink.bc.ca (Alan Douglas) wrote:

>>As for checkers -- there is a world of difference between it and a
>wargame like Third Reich.  In checkers you only move one piece per
>turn, giving you at most about 20 possible moves to consider.  For an
>AI to factor in its opponents reply, it must consider about 20*20=400
>possible positions.  Another move ahead would take it up to 8000.
The branching factor for checkers is much lower, because the number of
legal moves for any given turn is much lower than 20.


>We would be at a disadvantage compared to conventional wargames which
>are designed specifically to be easy for us to understand.
I think they are designed to be playable.


>As I mentioned in my post "AI-friendly war games - PART II", I could
>very well be proceeding from a false assumption about the true nature
>of a battlefield.  If front lines are merely a convenient abstraction
>and are generally unoccupied by forces except for those concentrated
>in isolated patches, then the standard discrete unit approach would be
>more appropriate.  My continuum representation assumes forces to be
>deployed along the entire length of the front, as I assumed to be the
>case in large scale conflicts such as the WWII invasions of France and
>the USSR.  If I am wrong about this, then please correct me.
I don't see how this is different than influence mapping.


>With standard wargames, a skilled and experienced player can look at
>the large scale disposition of forces and determine what the best
>strategy is and wheres the best place to attack.  Computer AIs are
Good. That's what makes her a skilled and experienced player. If
players could not learn these things with experience of playing the
game, then such games would not very different than games of chance.


>Now what gives computers an edge is brute force calculation.  But if
>youre forcing an AI to play a wargame with 10 to the 40th possible
>moves each turn, then its advantage is nullified.  All Im proposing
What makes you think wargames work this way?

Uri


