Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!necntc!necis!encore!pierson
From: pierson@encore.UUCP (Dan Pierson)
Newsgroups: comp.emacs
Subject: Re: Gomoku: a strategic board game for GNU Emacs.
Message-ID: <1974@encore.UUCP>
Date: Wed, 23-Sep-87 13:46:10 EDT
Article-I.D.: encore.1974
Posted: Wed Sep 23 13:46:10 1987
Date-Received: Sat, 26-Sep-87 04:48:29 EDT
References: <1547@culdev1.UUCP> <13301@bu-cs.BU.EDU>
Reply-To: pierson@encore.UUCP (Dan Pierson)
Organization: Encore Computer Corp, Marlboro, MA
Lines: 20


The complete search tree for Go obviously has to be very large, if just
because of the board size (19x19).  On the other hand, my (limited)
experience is that Go players tend to look much further ahead than
chess players because the scope of look-ahead is more bounded.  Chess
pieces *move*, many of them large distances; therefore a single move
can have direct, immediate consequences in a distant part of the board.
Go pieces are placed or (rarely) removed, but never moved; therefore most
tactical decisions are localized, leading to a deeper but simpler tree.
Srategic decisions are another matter altogether; the consequences can be
subtle and take a long time to mature.  About fifteen years ago I heard
that one of the leading Go programs was using potential theory instead of
search trees for its strategic moves...

Can we please move this discussion to a more appropriate newsgroup?
-- 

In real life: Dan Pierson, Encore Computer Corporation, Research
UseNet: {talcott,linus,necis,decvax,ihnp4}!encore!pierson
ArpaNet: pierson@multimax.arpa