Xref: utzoo rec.games.chess:5222 comp.sys.ibm.pc.misc:2139
Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!mailrus!iuvax!news!cartan!ndmath!nstar!inland!pals
From: pals@inland.com
Newsgroups: rec.games.chess,comp.sys.ibm.pc.misc
Subject: Re: Chess program recommendation wanted (for IBM)
Message-ID: <103.2703c1f5@inland.com>
Date: 28 Sep 90 21:34:45 GMT
References: <1399@pedsga.UUCP> <6302@hub.ucsb.edu> <85.26ff813b@inland.com> <1990Sep27.121239.5258@cs.rochester.edu>
Organization: Inland Steel Research Labs; East Chicago, IN
Lines: 67


     I am posting this followup as added information/counterpoint
     to the followups posted by Ray Frank, Thos Sumner, and Michael
     Kaufman on this issue.  Take a deep breath, and we're off!

  1. WHO DOES THE RATING?  Mr. Frank brought up the issue of "official" 
     ratings of chess computers/software.  The person I have come to 
     trust on this issue is IM Larry Kaufman, who writes for Computer 
     Chess Reports, published by ICD Corp.  He has done a great deal of 
     work in rating chess machines, and I couldn't possibly begin to go 
     into all he has discovered here.  Kaufman understands the need for 
     but distrusts ratings of computers based on playing humans, and has 
     an in-depth discussion of the USCF's official Computer Rating Agency 
     (C.R.A.) in the Fall/Winter 1988-1989 Computer Chess Reports.  A 
     couple of flaws he points out:  C.R.A. Ratings of computers are 
     positively correlated with the number of tournament rounds played in 
     a day (people tire, computers don't).  Also, if manufacturers don't 
     like the C.R.A. rating they get, they can simply retest.  This 
     allows them to keep retesting until perhaps they get lucky and get a 
     rating they like.  Kaufman plays machines against each other. He 
     also checks his results against those of leading European testers.  
     He discusses his methods at length in the above-mentioned issue of 
     C.C.R. (you can try to get a copy from ICD at 1-800-645-4710). 

  2. CHESSMASTER 2100.  My biggest gripe with CM 2100 is that I already 
     had CM 2000 and (quite justifiably) expected an improvement in 
     playing strength.  I didn't get one.  To quote Kaufman, CM 2100 is 
     "an improvement of only 25 points or so over CM 2000".  His rating 
     of CM 2100 on a 10MHz XT is 1825 (USCF rating points), and 2010 on a 
     25 MHz 286.  Kaufman says "to reach the 2100 level would probably 
     require at least a 33 MHz 386 machine with cache".  At least when I 
     bought it, Software Toolworks had no such label on the box.  Kaufman 
     also writes "I would like to warn readers about misleading ads by 
     Software Toolworks for CM 2100.  It is claimed to have 'substantially 
     the same program that earned a 2325 rating in the World Open.' "  
     But "when Software Toolworks attempted to translate the Fidelity 
     68000 program [the one that earned the 2325 rating] to IBM PC code, 
     they found that it played even weaker than CM 2000."  So, Software 
     Toolworks chose only to "upgrade" the CM 2000 into the 2100.  My 
     final shot from Kaufman - "It may be a slight improvement over the 
     CM 2000, but the claimed 2100 (never mind the 2325!) is only 
     achievable on a fast 386 machine." 
     
  3. SARGON IV.  I'll be as brief as possible.  Kaufman calls Sargon IV 
     "a disgrace."  One more sentence - "Sargon IV seems weaker than its
     predecessor Sargon III, and should never have been released."

  4. MODERN MACHINES.  In response to Mr. Sumner's comments - of course, 
     I agree that CM 2100 would look increasingly good on hotter 
     machines.  Shoot, if we ran it on a CRAY, it might be a master.  But 
     its showing is poor compared to the software below.  I would also 
     point out that even though fast PCs are getting cheaper, there are 
     "billions" of us out here who bought XTs and early (slower) ATs a 
     few years back, and can't afford to trash them just yet. 

  5. RexChess.  Kaufman estimates this at 2177 USCF running on an 25 MHz
     286 or a 20 MHz 386.  This is 167 points above CM 2100.

  6. AI Chess.  Kaufman estimates this at over 2300 (!) on a 20 MHz 386, 
     a whopping 290 points above CM 2100.  He rates it a "strong Expert" 
     (lets call this 2100+) on a 10 MHz XT (8088).  But he says "the 
     graphics are still awful and the price still excessive."

     Whew!  This was fun!  And I still think the name "Chessmaster 2100"
     is misleading.

     Randy Pals