Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!howland.reston.ans.net!torn!watserv2.uwaterloo.ca!undergrad.math.uwaterloo.ca!eadengle
From: eadengle@watcgl.uwaterloo.ca (Ed "Cynwrig" Dengler)
Subject: Re: Scrabble Algorithms
Message-ID: <D78LK9.EGE@undergrad.math.uwaterloo.ca>
Sender: news@undergrad.math.uwaterloo.ca (news spool owner)
Nntp-Posting-Host: watcgl.uwaterloo.ca
Organization: University of Waterloo
References: <3mj43i$300@ucthpx.uct.ac.za> <3ml747$qbi@news.xs4all.nl> <3musin$o0m@inferno.mpx.com.au>
Date: Tue, 18 Apr 1995 15:26:33 GMT
Lines: 64

Greetings!

In article <3musin$o0m@inferno.mpx.com.au>,
Scott Lemon <slemon@jolt.mpx.com.au> wrote:
>carroll (carroll@xs4all.nl) wrote:
>: mwodrich@cs.uct.ac.za (M Wodrich) writes:
>
>: >Hello
>
>: >I am interested in computer algorithms for playing scrabble (human vs.
>: >computer, or computer vs. computer). At present I have only found one
>: >article, published in the ACM in 1988 or 89, called "The fastest scrabble
>: >program in the world". It is no longer the fastest (apparently there is
>: >a game called Crosswise that is faster), so I am interested in other
>: >algorithms/ modifications of the one I have seen....
>
>: Generating a list of all legal moves using the algorithm you list is 
>: very quick.  Since look-ahead doesn't make much sense in Scrabble (until 
>: the tile pool is nearly empty), the trick to writing a good Scrabble 
>: program is to have a good evaluation function.  All Scrabble programs 
>: I've seen make their moves very quickly, within a couple of seconds.  
>: Some of these moves aren't so good, but they are all quick.
>
>I don't really see much point in a computer based Scrabble program, all 
>bar the stupidest dictionary searching algorithm should beat 99.99% of the 
>general population.
>
>For instance, theres a shareware scrabble game for the Mac that I have 
>seen that answers in probably less than a second, and it comes up with 
>the most obscure words you can imagine. The only way to beat it is to a) 
>get the prgram to suggest a word for you or b) be one of the smartest 
>0.01% of the population
>
>
>Scott Lemon
>slemon@jolt.mpx.com.au

I had a quick chat with some friends who are serious Scrabble addicts.  Their
opinion of current state-of-the-art Scrabble programs is that while the
computer may possibly some up with some stronger scoring words on a particular
move, the overall placement strategy of the computer program is still lousy
(most programs optimize the score for the current word, rather than the score
for the game).  Thus in their opinions, the computer programs do not present
a great challenge to human play as yet.

Playing well for a game requires being able to plan, and gamble, ahead by
several words to generate good winning layouts.  The amount of lookahead
explodes even faster than Go due to the uncertainty as to which tiles
you will get.  So the current "trick" that needs to be worked on is in
fact lookahead.

Ed

P.S. As an interesting observation, lookahead and evaluation seem to be
very complementary types of information about a position, but which seem
to be handled separately on most of todays programs.  Guided lookahead should
be able to simplify an evaluation function (since "in the future" results
would not have to be explicitely coded) while an appropriate evaluation
function should be able to provide the guidance needed.  However, most Go
and Chess programs work on the two as separate entities, which is where
the problems of "combinatorial explosion" comes about (as an example of how
to reduce the explosion, expert human players seem to consider only the
"correct" moves, and by doing so, prune out much useless search).

