Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!uunet!in1.uu.net!ncrgw2.ncr.com!ncrhub6!daynews!falcon!news
From: Dick Menninger <Dick.Menninger@DaytonOH.ATTGIS.COM>
Subject: Re: [Q] Is there a tank game like this?
X-Nntp-Posting-Host: 149.25.99.44
Message-ID: <DCnFuM.Kq8@falcon.daytonoh.attgis.com>
Sender: news@falcon.daytonoh.attgis.com (News administrative Login)
Reply-To: Dick.Menninger@DaytonOH.ATTGIS.COM (MENNID)
Organization: AT&T Global Information Solutions
X-Newsreader: DiscussIT 2.5.0.10 for MS Windows [AT&T Software Products Division]
References: <3vh2pn$7q4@nnrp4.primenet.com>
Date: Tue, 1 Aug 1995 20:35:10 GMT
Lines: 44


> ==========Walt Mazur, 7/30/95==========
> 
> steve@empire.com.au (Steve Taylor) wrote:
> 
> >Come to think of it, Scientific American once had a Computer Recreations
> >collumn on a Rogue player called Rogomatic, which was good enough to win
> >the game. I've never seen it, but friends have. They say it has a very
> >dull style of play.
> 
> I got ahold of the source code and looked at it long ago.  Perhaps
> it's still on the net.  As near as I could tell, it was all heuristic.
> It played a good game largely by doing things far too tedious for a
> human player to tolerate, like move over the minimum number of squares
> to minimize chance of triggering a trap.
> 
What is provided as primitives has a huge effect on how easy
various things are to do.  The game did not provide primitives to
the human to make these things easy.  They were, however, things
that an automated opponent could do readily.  AI can give us such
insights into how to make some things easier for humans.  What are good
primitives and how easily they can be manipulated will vary between
machines and humans.  A game AI generally has some hardcoded knowledge
as primitives.  Often, the quality of the AI is strongly affected by
the primitives available to it.  If it has to work out what should be rote
knowledge, then it will not get to a higher level of play.  This is also what
happens with humans, for the most part.  They start by wallowing through
stuff that needs to be rote primitives, probably even the building blocks
of what shoul be even larger rote primitives.  As a result, they play
poorly.  Later, they have more, and higher level, primitives with which
they work, and they have a better formed strategic framework.  Then
they play much better.  Often, they are uneven in their primitive framework
and strategic framework.  As a result, they vary from higher level work
to grunging in the muck, depending on context.  (Humans mirroring
computer game AIs? :-) )

So the question is, how do you distinguish heuristics from a framework
of primitives that has developed to a high enough level of sophistication
to effectively solve the given game?  Apply that answer to Rogomatic.


Good Day
Dick

