Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rebel10's anti-GM revised...

Author: Amir Ban

Date: 03:23:45 06/30/98

Go up one level in this thread


On June 29, 1998 at 18:07:31, Bruce Moreland wrote:


>So, in this case, both programs scored a 1/2 point more by not going nuts on e6,
>although to be fair, I didn't test to see if they'd actually play fxe6 after
>Rxe6, and this is a surprisingly hard move to find.
>

This is the reason Rxe6 is good after all. fxe6 is a very difficult move, not
only for a computer opponent, but for the GM. It's not reasonable to expect the
GM to see that he comes out of the complications on top, let alone alive. You
have to swallow moves like Kxd7 and look real deep in positions where every
piece is under attack. I think a reasonable player would spend here 80% of his
time to find a reasonable way to decline the sac, and would accept it only if he
doesn't find any.

The name of the game here is "bluffing", not a bad strategy.

I am guessing that Ed's anti-GM plan involves a bonus for sacrifices, which will
make the program play borderline or slightly inferior sacs. Unless the tactical
sequence is very simple or forced, this will easily pay back the slight
"objective" eval cost either by forcing the master into a position where he
understands less than the computer, or, even easier, by making the master
decline the sac.

The old Fritz strategy against humans, of willingly taking on a central isolated
pawn, was a similar idea: Take on a positional liability that you have to be a
good tactician to exploit.


>
>How many lines of code is this feature, Ed?  Does it only come into play when
>the program is down material?  Does it get more correct on ECM?
>

Who cares about lines of code ? These kind of things may be just adding a term
or raising a flag in the right place. I would be disappointed and suspicious to
hear that Ed wrote 2000 lines for this.

Being correct on ECM counts for nothing. Ed is not going to throw a game because
he reached an ECM-like position and failed to solve it in 30 seconds. He's not
going to reach any such position if things go as Anand plans (or maybe he will,
but from the losing side).

Amir




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.