Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: CM 5500 vs. Fritz 5 & Rebel 9

Author: Komputer Korner

Date: 06:38:21 08/06/98

Go up one level in this thread


On August 05, 1998 at 12:14:25, Shaun Graham wrote:

>On August 05, 1998 at 07:41:47, Freddie wrote:
>
>>Shaun,
>>
>>I was surprised by a previous posting where you said that Chessmaster 5500 would
>>beat Fritz 5 or Rebel 9 on a Pentium II 400 computer.  Aren't the latter two
>>programs supposed to be superior in both knowledge and search speed?  I would
>>appreciate it if you or someone else could explain how and why this would be
>>possible.
>
>
>You can go to this page
>http://www.geocities.com/TimesSquare/Ring/8448/compchess2.html#t08 were
>chessmaster has played almost a 100 40/2 games with excellent results surpassing
>all the top programs it recently beat rebel 9 7 - 2 in a 40/2 match on two
>P233s.  Chessmaster for some reason benefits considerably more from faster
>processors than other programs, i believe this to be the main reason for it's
>much increased strength over the ssdf rating.  Most people like to use CM5555
>settings to enhance chessmasters performance, but the new settings in my
>oppinion only increase the strength slightly.  I've played many games with
>Chessmaster vs Rebel, Genius and fritz 5 myself, when on comparable fast
>hardware (p233s) chessmaster most often has a slight edge in winning
>percentages.  You can also find some tournaments on the gambitsoft page
>demonstrating the increased strength chessmaster has running on faster hardware.
>The winning resulys are even more spectacular because chessmaster has a
>non-optimized book.   Chessmasters combinative ability surpasses all other
>programs i believe and it finds mate faster than any other program as well.


I have always said that highly selective (razoring) programs will gain more from
faster machines than non razoring programs because the minimum razored search
depths increase enough to catch most blunders caused by the razoring process.
Vincent Diepeveen coined the term TACTICAL  BARRIER to cover this. While I
believe that there is no tactical barrier, as the search becomes deper the
blunders caused by razoring become fewer and fewer. I believe that this gain
will be a steady one the deeper you go into the search.  So engines like CM5500
(the King) have a good future. Bob Hyatt doesn't trust razoring but does do a
form of forward pruning called null move. Note that razoring and forward pruning
are different. Forward Pruning is cutting off complete non critical candidates
and not looking at them again, whereas razoring is looking at these candidates
but at a lower search depth than the others. Of course as you go deeper into the
search, the minimum ply depth of the razored candidates grows to the point of
catching almost all the blunders caused by the razoring itself. CM5500 is proof
that on very fast machines and long time controls, the razoring idea works.
Today almost all SELECTIVE SEARCH PROGRAMS ARE BUILT ON THE RAZORING IDEA. Hyatt
believes that the gain in depth on the critical candidates is completely
outweighed by the blunders caused by the razoring process and that increased
speeds will not improve the situation. Programs like CM5500 are proving him
wrong.
--
Komputer Korner



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.