Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame Testposition (easy for some, hard for others)

Author: Thomas McBurney

Date: 19:59:35 06/21/04

Go up one level in this thread


>
>My opinion is that -5.5 is too good for black because you get it also for the
>following position when white is winning.
>
>
>You should give a smaller bonus for uunstoppable passed pawn in case that the
>unstoppable passed pawn is not more advanced than the most advanced enemy pawn.


I guess one of the goals of giving a large bonus is so your chess engine would
sacrifice pieces to achieve a won pawn end game.  Is it worth sacrificing a rook
if you know that you are going to queen first?  The question is how reliable is
your evaluation routine at evaluating these sorts of positions correctly?  In my
experience so far I have found it to be pretty reliable but I guess one can
always tune their evaluation to be better.  Of course testing will show if there
is a benefit or not.  The board position you have shown is good one because it
demonstrates a common flaw with people’s implementation of the pawn square rule
where the position has complexities that are not easy to detect in evaluation.
But then, on the other hand having the pawn square rule can show refutation
quicker than it would normally with out it.

Kanguruh 1.87

Static eval for Uri's position: +7.46

  2    744      29          18  h8g7 d5d6
  3    751      29          87  h8g7 f4f5 g6f5 g3f4
  4    721      29         212  h8g7 f4f5
  4    723      29         368  h8g8! f4f5
  5    685      29         682  h8g7? f4f5
  5    686      28        1004  h8g7
  6    685      27        3597  h8g7 f4f5 g7f6 g3f4 a5a4 f5g6
  7    686      23       10911  h8g7 f4f5 g7f7 g3f4 a5a4 f5g6 f7g6 h4h5 g6f6 h5h
6
  8    678       5       54756  h8g7 g3g4 a5a4 g4g5 a4a3 f4f5 g6f5 g5f5 a3a2
  9    673      25      138406  h8g7 f4f5 g6f5 g3f4 a5a4 f4f5 a4a3 d5d6 g7f7
 10    643      67      272131  h8g7 f4f5
 11    611     146      549828  h8g7? f4f5
 11   -325    1036     3104121  h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 a5a4 d6d7 f6e7 d7d
8 e7d8 h5h6 d8e7
 12   -295    1457     5029835  h8g7! d5d6
 12   -295    1648     5878122  h8g7 d5d6
 13   -265    2625    10803510  h8g7! d5d6
 13   -265    3227    13550805  h8g7 d5d6
 13   -264    4201    18891963  h8g8! d5d6
 13   -265    4622    20945027  h8g8 d5d6
 14   -294    5115    23559973  h8g8 d5d6
 15   -324    9117    45746847  h8g8 d5d6



Here is Lambchops analysis, as you can see it makes the same inital mistake as
Kanguruh.


  1  589     0       6 a5a4
  2  516     0      60 h8g8
  2  557     0     104 h8g7
  2  567     0     127 h8g7 d5d6 g7f6
  3  522     0     382 h8g7 d5d6 g7f6 d6d7 f6e7
  4  561     0     972 h8g7 d5d6 g7f6 d6d7 f6e7 g3f3
  5  560     0    2251 h8g7 d5d6 g7f6 d6d7 f6e7 g3f3 a5a4
  6  576     0    6268 h8g7 f4f5 g6f5 g3f4 a5a4 f4f5
  7  581     0   13695 h8g7 f4f5 g6f5 g3f4 a5a4 f4f5 a4a3 d5d6 g7f7
  8  491     0   38391 h8g7 f4f5 g6f5 d5d6 g7f6 h4h5 a5a4 d6d7 f6e7 h5h6 f5f4
g3f4
  9 -350     0  134011 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 a5a4 d6d7 f6e7 h5h6 f5f4
g3f4 e7d7 h6h7
 10 -530     0  231090 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 a5a4 h5h6 a4a3 h6h7 f6g7
d6d7 g7h7 d7d8Q a3a2
 11 -546   100  313043 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 a5a4 d6d7 f6e7 h5h6 e7d7
h6h7 b6b5 g3f4 b5b4 h7h8Q
 12 -594   100  430486 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 b6b5 h5h6 f6e6 h6h7 e6d6
h7h8Q d6d7 g3f3
1 13 -669   100  695769 h8g7
 14 -708   200 1154444 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 f6e6 h5h6 e6d6 h6h7 b6b5
h7h8Q d6c7 g3f4 b5b4 h8d4 c7b7
 15 -717   400 2037500 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 f6e6 h5h6 e6d6 h6h7 b6b5
h7h8Q d6c7 h8e5 c7c6 e5e6 c6c5 e6f5 c5b4 g3f3 a5a4
 16 -726  1100 5915377 h8g7 d5d6 g7f6 f4f5 g6f5 h4h5 f6e6 h5h6 e6d6 h6h7 b6b5
h7h8Q d6c5 h8c3 c5b6 c3d4



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.