Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fritz5 missed a win against Chesstiger in the wmcc in paris

Author: blass uri

Date: 00:13:12 08/01/98

Go up one level in this thread



On August 01, 1998 at 02:50:54, Thorsten Czub wrote:

>On July 31, 1998 at 22:51:18, blass uri wrote:
>
>>position:
>>2r2rk1/2q2pP1/p1b1pp2/1p6/4P3/2Np3P/PPP3Q1/1K4R1
>>white to move.
>>
>>Fritz5 arrived to a winning position because it has a better positional
>>understanding.
>
>Fritz has no "positional understanding".
>
>You believe it has, but it hasn't.
>
>>Fritz5 played 25.cxd3 and made a draw when 25.Qd2 is winning for white
>>
>>Fritz5 using 131072Kbytes hash tables needed almost 30 minutes to find
>>Qd2
>>Junior5(16 bit version for fritz) needed 10 minutes and 40 seconds
>>using 31000 Kbytes
>>
>>This is on pentium200MMX
>
>They used a p2/300 in Paris ! Also they had massive Hash-table-usage in Paris,
>as far as I knew.
>
>The draw-by-repetition-bug is a known problem/phenomena with the fritz-engines.
>The whole time in Paris Fritz had wired draw-lines and senseless repetition
>stuff.
>We made all our jokes watching the chessBase team sitting behind their
>"randomn-draw-generator -machine".
>
>>Genius3 needed 42 seconds to find Qd2 on a pentium100Mh with 15 Mbytes hash
>>tables.
>>
>>I think the strong point of Fritz5 is positional understanding
>
>Oh man, you really seem to have problems to differenciate tactics from
>positional stuff.
>You are the first guy I know who claims positional-understanding for fritz5.
>
>Whats next ?! Extraterrestrians on the moon ?! Nessi in Loch Ness ?


I can give you an example for this.
I did the nunn match between Junior5 and fritz5 won in 30 minutes per game
+30 seconds per move(the result after 16 games 10.5:5.5) because of better
positional understanding.

here is an example to position Fritz5 understand better than Junior

2r3k1/pq2rppp/b1p1p3/1n6/3PPP2/1N2Q1P1/6BP/RR4K1
black to move
Fritz(white) understood that white is better in this position(it knows not only
to count pawns) when Junior evaluated the position as a small advantage for
black.

Junior  positional understanding of its games is better than of the nunn match
so I cannot say practically in games fritz has better positional understanding
but Fritz has better general positional understanding.

I looked in the games fritz5 played as white and did not win in the
WMCCC and its evaluation in the end of the book opening was usually
small advantage for black.

>
>>The weak point of fritz5 is the opening book.
>>In tactics it is sometimes weaker and sometimes stronger than other programs
>
>It is a finder. it is good in finding key-moves, but sometimes oversees easy
>tactics because of massive pruning. HERE you are right, for the first time :-)
>
>>In the first round against Isichess Fritz5 with white came out of the opening
>>book with negative evaluation with white.
>
>And ?! The book is NOT tuned for Fritz5. It is a quantitative approach, isn't it
>?!
>
>>I believe fritz5's opening book has too many lines stupid humans played.
>
>Thats true. All books that were melted out of databases have this "problem".
>
>>It is better to delete lines Fritz5 does not like from the opening book of
>>fritz5.
>
>It needs a person to do so, that has to be paid for the job.


If they  delete lines when the evaluation of the computer is negative
for fritz5 the program will have  better results.

Uri

>
>
>>Uri



This page took 0.01 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.