Author: Dieter Buerssner
Date: 19:22:55 09/06/03
Go up one level in this thread
On September 06, 2003 at 21:58:14, Kim Roper Jensen wrote:
>Position from the game Larsen - Kavalek Montreal Tournament of Stars
>
>[D]1k6/8/B3K1P1/5P2/1pRr2Nn/2b5/8/8 b - -
>
>Only move here for black is Nxf5 and get both white pawns and it should be a
>draw - according to the notes by Whiteley (its taken from the tournament book,
>by Tal, Chepizhny & Roshal)
>
>Shredder 6 Paderborn Classical finds Nxf5 fast, but no drawscore after a long
>time ... this on 2Ghz AMD with 256MB hashtable but very limited
>endgametablebases
Yace needs much longer than Shredder, to find Nxf5, but it shows a somewhat
better score:
13182076 30.090 -3.07 10. 1...Rd1 2.Ne5 Re1 3.Rxc3 bxc3 4.g7 Rg1 5.Ng6
Re1+ 6.Kd7 Rd1+ 7.Ke8 Re1+ 8.Kf7 c2 9.g8=Q+ Ka7
10.Qc8 {-851}
18509925 38.821 -3.25 11t 1...Rd1 2.Ne5 Re1 3.Rxc3 bxc3 4.g7 Rg1 5.Ng6
Re1+ 6.Kf7 Nxf5 7.g8=Q+ Ka7 8.Qd8 Kxa6 {-421}
25301247 56.569 -3.24 11t+ 1...Nxf5 2.Kxf5 Rd5+ 3.Ke6 Rg5 4.Kf7 Rf5+ 5.Kg8
Rg5 6.Rc8+ Ka7 7.Rxc3 bxc3 8.Bc8 {-190}
27300699 1:00.2 -2.67 11++ 1...Nxf5 2.Kxf5 Rd5+ 3.Ke6 Rg5 4.Kf7 Rf5+ 5.Kg8
Rg5 6.Rc8+ Ka7 7.Rxc3 bxc3 8.Bc8 {-190}
28541573 1:02.8 -1.67 11++ 1...Nxf5 2.Kxf5 Rd5+ 3.Ke6 Rg5 4.Kf7 Rf5+ 5.Kg8
Rg5 6.Rc8+ Ka7 7.Rxc3 bxc3 8.Bc8
29155992 1:03.8 -1.66 11t 1...Nxf5 2.Kxf5 Rd5+ 3.Ke6 Rg5 4.Kf7H Rf5+H
5.Kg8H Rg5H 6.Kh7H Rh5+H 7.Nh6H Bd2H 8.Rc8+H
Ka7H 9.Rc4H Kxa6H 10.Rxb4H {-70}
30881556 1:07.0 -1.66 11. 1...Nxf5 2.Kxf5 Rd5+ 3.Ke6 Rg5 4.Kf7 Rf5+ 5.Kg8
Rg5 6.Kh7 Rh5+ 7.Nh6 Bd2 8.Rc8+ Ka7 9.Rc4 Kxa6
10.Rxb4 {-70}
42662815 1:33.0 -1.26 12++ 1...Nxf5 2.Kxf5 Rd5+ 3.Ke6 Rg5 4.Kf7 Rf5+ 5.Kg8
Rg5 6.Kh7 Rh5+ 7.Nh6 Bd2 8.Rd4 Rxh6+ 9.Kg7 {0}
48237358 1:42.9 -0.91 12t 1...Nxf5 2.Kxf5 Rd5+ 3.Kf4 Rd6 4.g7 Bxg7
5.Rxb4+ Ka7 6.Rb7+ Kxa6 7.Rxg7 Kb6 8.Ne5 Kc5
9.Kf5 Kd4 10.Re7 {-340}
53235180 1:53.3 -0.91 12. 1...Nxf5 2.Kxf5 Rd5+ 3.Kf4 Rd6 4.g7 Bxg7
5.Rxb4+ Ka7 6.Rb7+ Kxa6 7.Rxg7 Kb6 8.Ne5 Kc5
9.Kf5 Kd4 10.Re7 {-340}
The -0.9 can be interpreted as: Actually I am quite confident, that this is a
drawish position, even if I a less material.
You could ask: Why then not give a draw score, when quite confident? Because it
will yied in problems! Crafty, for example, will evaluate KR + minor piece vs KR
as 0.01 (it is much more confident of the draw). But by this, it can (and does)
easily lose drawn games. Now, all defender moves look exactly the same, and it
is easy to do some bad moves sooner or later. Then a draw will go to a loss.
Those close to 0.0 scores cannot differntiate between - uff my K is already at
the edge, and the opponent has mating possibilities, and between - no problem.
Of course, in the actual scenario I mentioned (only 5 pieces) TBs will help. But
when you add a pawn to the inferior side, the will not help much anymore. For my
engine, it seemed best, to give such very probably drawish positions somewhat
higher scores, giving the engine the possibility to differntiate between really
critical positions, and other ones, that are not that tough.
Regards,
Dieter
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.