Computer Chess Club Archives


Search

Terms

Messages

Subject: Fritz5 and Occam's Razor

Author: Larry Woodlock

Date: 23:35:59 02/22/98


     I've owned Fritz5 for some months now, and used Fritz 4.01 before
that.  Over the last months I have noticed a marked increase of chess
knowledge in Fritz5 vis a vis its predecessor.  An example:
     Just this past week, I started building up a King and Pawn endgame
database in Fritz5, transferring information from conventional books
into an "e-book" that I could use to study and refresh my recall of KP
endgames - something that "Fritzbase" is very good at.  I had ordered
the Hiarcs6 module to assist in the analysis, but it seems to be lost in
the mail.  So, "stuck" with Fritz, I would take out the CD (entering the
positions under "infinite analysis" leads to hangups during the
occasional CD search), and enter the positions with variations and
commentary.  I couldn't help but be struck by how well Fritz analysed
the positions - far better than I had expected any program to do.  I had
concluded over the past week that Fritz5 would be a truly formidible
opponent in KP endames, something I was not expecting to find.
     So in addition to what I thought of as a real increase in
positional sophistication in middlegames, Fritz5 is also skilled in the
endgame. Then came the new list.
     I had expected that Fritz5 would do well if it ever got up and
running at SSDF, but I could not quantify that expectation from my
subjective conclusion that Fritz had had a major knowledge upgrade.
Like some of the other correspondents, I am surprised to see that Fritz5
vaulted into first place, but I can't reject the SSDF data as freely as
others do.
     Who among us can specify just how much the extra chess knowledge
should be worth at 40/2 against a variety of opponents?  Occam's Razor,
William of Occam's injunction to select the most economical hypothesis
capable of explaining a given dataset, suggests that we not presume
latent flaws in SSDF to explain away this result, but rather that we
accept a theory that Fritz5 has seen a dramatic increase in real playing
strength at longer time controls.
     This is certainly a big leap, and more data may moderate the
finding, but for now we should consider congratulating the Fritz team
for shaking up our little world.



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.