Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!swrinde!pipex!sunsite.doc.ic.ac.uk!warwick!sunserver1.aston.ac.uk!pcx_10.aston.ac.uk!user
From: P.Coxhead@aston.ac.uk (Peter Coxhead)
Subject: Re: Opponents to move...but how ?
Message-ID: <P.Coxhead-0411941000210001@pcx_10.aston.ac.uk>
Sender: usenet@aston.ac.uk
Organization: CSAM, Aston University
References: <AigJQQq00iV247Og5w@andrew.cmu.edu>
Date: Fri, 4 Nov 1994 09:58:39 GMT
Lines: 30

In article <AigJQQq00iV247Og5w@andrew.cmu.edu>, Joe Eddy Demers
<Coyote+@CMU.EDU> wrote:

> I've found that raw, unadulterated agression works pretty well.
> Basically, if you have enough strength to attack, do so.
> If not, build up your forces to attack next time.
> Simple rules like always try to attack smaller forces with large,
> and assignment of priorities to goals are also steps in the right direction.
> If that's too simplistic, alpha-beta search works well,
> and generalizes to nondeterminism nicely.
> 
> Ignore the following if you understand alpha-beta search:
> (alpha-beta search: find the best possible outcome for your side,
> assuming both sides make perfect choices, by enumerating your possible
> moves, and for each one of these, your opponents possible moves, etc.
> for a certain number of moves. Then you evaluate the positions, and give
> numerical values to each:
> positive if it's good for you, negative if good for opponent, 0 if even.
> Assuming the player makes the best move, take the largest number to be
> the value of the position at the next to last move. (largest = smallest
> (negative) if it's your opponent's turn) repeat this for the last move,
> the one prior to that, the one prior to that, etc. until you get to the
> top, and you just choose the move with the biggest value)
> (oh, nondeterministic: having some elements of randomness)
> 
> Joe

This isn't actually alpha-beta (which is a pruning method) but just
mini-maxing.  See e.g. Chapter 6 of Winston's 1992 "Artificial
Intelligence" (Addison-Wesley).
