Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty 17.10 not that strong

Author: Robert Hyatt

Date: 14:22:52 04/22/00

Go up one level in this thread


On April 22, 2000 at 14:20:57, A.L.Mourik wrote:

>Hello dear CCC friends,
>
>Although earlier reports from e.g. Jouni Uski, suggest an enormous increase in
>strenght for Crafty 17.10
>Nuun 2 match result Fritz6 against Crafty  17.10 ends in a very clear
>29,5-10,5!! victory for Fritz6.
>Played on PII 400 8mb for HT, Timecontrol 5 min + 3 sec per move.


This keeps coming up, so I may as well add my two cents worth one more
time..

You did _not_ reproduce the results of others.  But then you didn't run
the same test either.  You used one computer, presumably with ponder=off.
That is _not_ a reasonable test.  Lots of reasons, but to make it simple:

Let's play a match with no opening book.  That favors the program with the
lousy book.

Let's play a match with search extensions and tricks disabled.  That favors
the program with the best eval.

Let's play a match with evals disabled. That favors the program with the
best search.

Lets play a match searching to fixed depth.  That favors the program that does
the most work per node.

Let's play a match searching to a fixed time limit. That favors the program
that has a horrible automatic time allocation algorithm.

In short, when you turn off _anything_ you most likely affect one program more
than the other.  You turned off pondering.  I have reported several times that
such hurts crafty, definitely.  I allocate time based on the _assumption_ that
I will save some time later in the game with correct pondering predictions.  I
don't when I can't ponder.

The results are interesting, but they have _nothing_ to do with the way I would
actually play a game against fritz in a real tournament.  I would _not_ play
without pondering enabled...




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.