Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Example: inaccuracy against accuracy: Gambit-Tiger vs. Crafty17.13

Author: Uri Blass

Date: 13:32:57 11/09/00

Go up one level in this thread


On November 09, 2000 at 12:49:07, Bruce Moreland wrote:

>On November 09, 2000 at 11:01:59, Thorsten Czub wrote:
>
>>Here another really nice and lucky game of Gambit-Tiger versus Crafty17.13.
>>
>>if you want we can post the different scores too. its funny.
>>most often gambit-tiger said +3.0 and crafty was 0.3x down !
>>
>>you can see very good in this example how gambit-tiger tries and tries to
>>attack,
>>how it betrays crafty.
>
>In this game, Tiger was white and had an attacking setup.  Crafty got pretty
>passive and Tiger was very aggressive.  Crafty didn't do anything particularly
>stupid to get into this situation, I can imagine any program having problems in
>that position.  You can't blame someone for playing black and getting
>king-attacked in a Sicilian.  But Tiger could not break through.  Crafty
>defended well and counterattacked.  The balance shifted to neutral around the
>time Crafty played ... d4.
>
>This isn't about inaccuracy or accuracy or anything else.  It's about being
>aggressive and kicking heads.  Tiger played very aggressively and turned all of
>its units into attacking units.  I think Crafty's evaluation was low.  But the
>fact remains that Tiger's evaluation was not borne out, the attack couldn't
>overwhelm a point in Crafty's position.
>
>For those expecting to see a sacrificial attack here, it didn't happen.  (By the
>way, in another post, I said Tiger had not made any long-term sacrifices, and I
>was talking about in my games, not in the game where it played Rc6, which is an
>obvious long-term sacrifice.)
>
>>then suddenly the king attack gets lost, but the endgame takes over.
>>this was very risky, and only due to craftys good defense... and some
>>luck, it was possible.
>
>It was BN vs R and two more pawns.  Tiger made a very obvious exchange sacrifce.
> I don't know why Crafty didn't see it.  Mine thinks the game is about even
>prior to the sacrifice, and it takes it 2.7 seconds to get a score of in excess
>of +1 on that sacrifice, and it's a good +1, I think it knows what is going on.
>I think that anything could find that.

Crafty17.13 at the end of iteration 15 still likes Ne8 but expect the right move
Rxc6 with a small advantage for white.

I did not use tablebases but I believe that they cannot change the score because
I see no tablebase positions in the relevant lines.

The score at depth 16 is more optimistic for white and is 0.54 pawns for white
and there is a good chance that crafty is going to change its mind at depth 16
but it is too late for 60/60 game even if crafty use hardware of 2 gh.

I think that the problem here may be a different evaluation of crafty and
another problem is that crafty does not consider Rxc6 for enough time(it is a
known problem of chess programs that they do not consider for a long time moves
that simplify the position).

Uri



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.