Author: Vincent Diepeveen
Date: 05:32:01 09/09/01
Go up one level in this thread
On September 09, 2001 at 04:50:14, CLiebert wrote:
>On September 08, 2001 at 15:55:00, Vincent Diepeveen wrote:
>
>>On September 08, 2001 at 12:01:16, ALI MIRAFZALI wrote:
>>
>>Kramnik will win chanceless and not because he has that copy
>>already, of course it will act diff at 8 processors, but
>>most important thing is that this is a good sales trick.
>>
>>Of course it's very unlikely the fritz version facing kramnik will
>>be much different from deepfritz.
>
>
>
>It will be different, because
>
>a.) it is (for the first time) optimized against humans!
>(I hope that cb will publish both versions:
>Fritz-Human and Fritz-Comp like before)
After 20 years of fulltime work it would be commercially stupid
to say it's the same version, but obviously it will be not much
different, otherwise Frans Morsch would already have put in that
feature 10 years ago.
The program is so well tuned and optimized that any change to it
is hardly possible without years of fulltime work and testing.
>b.) it will be stronger again (as far as I know may be 50 points on si.-pc)
>c.) it uses 8 proc. in this match
That doesn't matter against humans, you know it and i know it and Bob
knows it.
Against humans you don't want to search 5 ply of course, but i remember
fritz3 already getting 10/11 ply and sincethen it has become a more
actively playing program.
Tuning fritz 'more' for humans would result in letting it give away
direct pawns which against a technical player like Kramnik, also one
of the stronger tactical players on earth, is not very smart.
Fritz now already wants to give away a pawn for 2 half open files towards
the king.
This is very well tuned code i bet, so obviously even increasing that
would be a complete disaster.
And because of little knowledge in the evaluation searching even more
deeper than the 15-17 ply it already got is completely useless.
>nevertheless I guess that kramnik will win, but not as easy as some people
>think...
With his eyes closed. Where Kasparov's strength is for 700 points based
upon opening and how he during the game invents how to play further after
that, for Kramnik his strength is not exactly based upon opening.
Nevertheless i think that playing Berlin defense with black would be
not so smart for Kramnik, though who am i to say kramnik would lose it,
being the strongest chessplayer in the universe right now...
He better can play Najdorf, trick Fritz in the opening and win easily.
Knowing kramnik he goes for a draw with black though.
Who am i to say that you lose if you play berlin with black against white
a computer? Because it's kramnik. Even with 2 pawns less in the endgame
he can still draw it against most programs.
That past years Fritz endgame slowly has improved doesn't change that.
The real weakness of fritz is obviously middlegame as it lacks
loads of chessknowledge. For example it doesn't know anything from
good or bad bishops anymore (i had the impression fritz6 knew more
from that than deepfritz does), i would be amazed if Kramnik fails to
find that weakness.
Nevertheless Russians have weird logics, it can change the number of
draws most likely, but not the result of the match.
Berlin defense is of course completely stupid to play against a program
objectively spoken:
a) you have a cool majority king side and even when a program doesn't
realize that, it still wants to advance, of course fritz is no exception
here
b) king in quadrant of a passer when it exists after exchanging is
knowledge any program has
c) all pieces are in the center
So on paper only superb play from Kramnik can win the berlin defense.
My feeling says he will, because of some lacks in pawn structure of fritz
which i know which might get on the board on the queen side :)
But if your bet is that the games will be hard fought, then for any position
which is derived from Berlin defense you may be right, but then we talk
about kramnik having black. Kramnik having white is an entire different
thing. He'll blindfolded win that because in the kind of positions he plays
you need all positional basic knowledge, and fritz is not even close to
that. In fact i'm always amazed why the thing scores so well at world
champs and other tournaments. It always is near the top.
Perhaps its strong point is the same from Kasparov, book!
In a match vs kramnik that doesn't count. Kramnik prefers to get out of
book at move 1 of course, against *any* opponent.
This is what my expectation is based upon. Of course playing
with a preprocessor would be way worse than what fritz is, but
all what is going to happen is a very boring match where fritz loses
chanceless.
Nevertheless Kramnik with black might beat programs at their strongest points:
tactics and advancing with a pawn to passer (obviously with loads of passers
programs have wealth problems which cause confusion and bad play, which
hardly will be the case here).
Perhaps it's better if Kramnik wins with a real huge score, but my expectation
is that Kramnik is only out to win the match and not with a 100% score.
Most likely approach from Kramnik is playing a safe middlegame, exchanging
to a minimal disadvantage or advantage and then going for a draw and
win if the program makes a major mistake.
The active play which in blitz is so effective for fritz against
humans like me (2300 FM) will be completely useless in this match!
>Christian
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.