Computer Chess Club Archives


Search

Terms

Messages

Subject: Fritz 5 SSDF - settings

Author: Moritz Berger

Date: 08:42:35 02/23/98


Here is more information about the configuration which got Fritz to the
top of the SSDF list. Doesn't sound like a killer book (quite the
contrary, an book created from a huge GM database, almost without any
fine-tuning for Fritz), the only significant thing is the hash table
size (44 MB, which is not really that much nowadays ...) and the choice
of opponents - playing many weak opponents (like Fritz 4 did because
they couldn't use the autoplayer with it) is too risky on "THE LIST",
because even the occasional draw that is forced out of book will cost
too many rating points.
Since the SSDF now has an autoplayer, Fritz will be matched mostly
against Auto232 compatible opponents and therefore not be endangered to
lose too many rating points against non-Auto232 opponents (Sparc 20,
Chessmaster 5000, ...).

(quote from www.chessbase.com/news.htm)

Fritz5 on SSDF list
Fritz5 has been included into the latest SSDF list with the highest
rating a chess program ever achieved.

1 Fritz 5.0 Pentium MMX 200 MHz     2589 54 -50 198 67% 2463
2 Nimzo '98 Pentium MMX 200 MHz     2534 43 -40 297 63% 2436
3 Hiarcs 6.0 Pentium MMX 200 MHz    2533 36 -34 414 63% 2441

Fritz5 setup for SSDF games: 44MB hash tables (on 64MB machines),
openings book writeable on hard disk, CD with endgame database in  CD
drive, chatter sound and music off, book learning parameters = default.

Fritz5 unfortunately scored around 40 points weaker in the SSDF result
than in our own published tests (Nunn-matches). Possible reasons:

We used 100 MB hash tables. The 44MB possible on a 64MB machine get
filled in less than a minute on a P200MMX.

The Nunn-match excludes opening book preparation effects. In contrast to
the books of other programs neither the Fritz5 standard book nor the
PowerBook CD is in any way engine-dedicated: There isn't a single move
in the book which hasn't been played before between two human players in
a published game. Statistical probability of Fritz' opening choice is
determined by popularity and score of opening variations in human chess.



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.