Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hiarcs7.32-fritz6Light 44-36(blitz 5' match no opening book)

Author: blass uri

Date: 06:10:56 08/01/00

Go up one level in this thread


On August 01, 2000 at 08:44:15, Mogens Larsen wrote:

>On August 01, 2000 at 08:14:44, blass uri wrote:
>
>>I try to find which engine is better at blitz without opening book.
>>
>>I guess that I am not going to get significant result in 800 games but I am not
>>sure.
>>
>>Uri
>
>You won't get significant result in general due to your unfounded speculations
>about hashsize. Are the hashsizes you use for each program recommended by the
>authors given the machine and timecontrol you use. If not, the exercise is
>rather pointless IMHO.

I know that for Hiarcs7.32 the suggestion is hash size as big as possible.

I can give hiarcs bigger hash tables then 128 Mbytes but I think that part of
the users cannot use even 128 Mbytes so I decided to do a compromise.

I do not know the best hash size for Fritz6 light but I know that Fritz can lose
on time  with big hash tables so I guess that  my 8 mbytes hash size is better
than the 16 Mbytes of the previous test.

I do not say that 8 mbytes are optimal but they are close to be optimal for
Fritz6light and I think that the programmer also does not know the optimal hash
size(with 8 mbytes it can lose on time in some rare cases but with lower hash
size it may have other problems).

It is possible to criticize my experiment because of the no opening book but I
decided to test in this way after seeing Fritz6 lsoing against tal with the
opening 1.e4 a6 because I think that knowing strange opening better can be
pracitcal in games when the opponents try to get you out of book.

I do not think that Fritz6light is at disadvantage because of these strange
openings but we need another experiment from known opening to check it.

Uri



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.