Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: They are all human...

Author: Marc-Olivier Moisan-Plante

Date: 00:15:17 10/29/05

Go up one level in this thread


On October 29, 2005 at 01:29:32, Uri Blass wrote:

>On October 28, 2005 at 22:56:20, Marc-Olivier Moisan-Plante wrote:
>
>>
>>Who would doubt that once in a while programs like to jettison a pawn on move 4
>>for obscure compensation?
>>
>>[D] rn1qkb1r/ppp1pppp/3p1n2/5b2/2PP4/2N5/PP2PPPP/R1BQKBNR w KQkq - 0 4
>
>e4 is not a sacrifice.
>White can get the pawn back in few moves.
>
>examples:
>
>1.d4 Nf6 2.c4 d6 3.Nc3 Bf5 4.e4 Bxe4 5.Nxe4 Nxe4 6.Qf3 d5 7.Qb3 with a fork.
>
>Note that 6.Qf3 is typical for humans who go for material gain.
>I guess that it is the reason that I find only 6.Qf3 in Fritz8.ctg and in
>Junior9.ctg and not computer typical move that do not win material like 6.Qb3 or
>6.Bd3
>
>Uri

As you point out, this is *not* a sacrifice. Just a plain bad move which gives
away the advantage for white (at least!). So just a bad move. This is different
from the examples presented before where "Fritz 9" absolutely (and wrongly)
sacrifies a pawn. You get a point here.

However in both Fruit 2.2 and Fritz 8 PVs', "4.e4" is a genuine pawn
sacrifice... (also Ruffian 1.0.5 "agrees") so they both intend to sac the
pawn... for dubious compensation (in my opinion)... hence my post... They look
like an over optimistic human here...strange pseudo-sacrifice... Do they look
like human? If yes, it is hard to say that "Fritz 9" is more human... or just a
weak human?!

I think that saying "Fritz 9" is more human is fluke (though this is an
uninformed opinion as I don't own it - I guess it is a great chess program in
general) and I would guess that Hiarcs 9 (9...9.6...10) is (will be) more human
in general.

What I wanted to point out is that "sceptical" program like Fruit 2.2 do
sometimes evaluate PV's in which there are "strange moves" (which GM would play
4.e4 seriously?), so that anti-Fritz data mining is unfair.

PS: Chessbase did positive data mining for their product here:
http://www.fritz9.de/experten/fritz9engine.htm . I know you already know that.

PPS: I already suggested on this forum (without success!) that fans of program
xyz should gather moves found *uniquely* by their favorite program, so the
competition would turn around finding "good moves" instead of publishing
"mistakes". That sounds more interesting to me... (general comment, not
something related to your post).

PPSS: With your suggestions (6.Qb3 or 6.Bd3), I still think that white stands
worse (are you an ICCF GM?, so you can confirm that?! - almost a free pawn,
isn't it?).

Cheers,

Marc-Olivier





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.