Author: David Rasmussen
Date: 16:18:04 02/08/03
Go up one level in this thread
On February 08, 2003 at 11:46:26, Andrew Williams wrote:
>First of all, if you want PM's log file for this game, just ask and I'll email
>it to you. It's too big to post here.
>
Thanks! But I don't think they're useful to me.
>The first interesting position is this one:
>
>[D]8/p4pk1/3p2p1/Pp2p3/1P1n2P1/6KP/4q3/2Q2R2 b - - fmvn 38; hmvc 3; id
>"PostModernist vs Chezzz, move 38"; sm Qc4
>
>PM definitely wouldn't exchange Queens here, as it thinks this is a nice attack.
> PM's score went from -3.1 to -1.4 here. (Why not Qd3+, which PM was expecting?)
>
Chezzz is liking Qd3+ for a long time, but then switches to Qc4:
time nodes depth score pv
00:00:00.00 6n 01/01/02 +1.705 1... e4 (6)
00:00:00.00 8n 01/01/02 +1.990 1... d5 (8)
00:00:00.00 12n 01/01/02 +2.005 1... f6 (12)
00:00:00.02 46n 01/01/02 ---------------------------------------------------
00:00:00.04 395n 02/04/05 +1.980 1... Qd3+ 2. Kh4 f6 (395)
00:00:00.05 684n 02/04/05 ---------------------------------------------------
00:00:00.05 2656n 03/06/08 +1.915 1... Qd3+ 2. Kg2 d5 3. h4 (2656)
00:00:00.07 4355n 03/06/08 +1.965 1... d5 2. Rf2 Qd3+ 3. Kh4 (4355)
00:00:00.07 4885n 03/06/08 +2.080 1... Qc4 2. Qe1 f6 (4885)
00:00:00.08 4937n 03/06/08 ---------------------------------------------------
00:00:00.08 7410n 04/06/08 +2.015 1... Qc4 2. Qb1 Ne2+ 3. Kh4 d5 (7410)
00:00:00.10 13kn 04/08/09 +2.185 1... Qd3+ 2. Kh2 Ne2 3. Qe1 d5 (13470)
00:00:00.13 15kn 04/08/12 ---------------------------------------------------
00:00:00.14 30kn 05/10/12 +2.155 1... Qd3+ 2. Kh2 d5 3. Qd1 Qxd1
4. Rxd1 (30371)
00:00:00.16 40kn 05/10/12 +2.235 1... Qc4 2. Qb1 Nc6 3. Rd1 d5 (40056)
00:00:00.22 54kn 05/11/14 ---------------------------------------------------
00:00:00.27 89kn 06/10/12 +2.510 1... Qc4 2. Qb1 Qc3+ 3. Kg2 Nc2
4. Rc1 Qd2+ 5. Kf3 Nxb4 (89621)
00:00:00.35 118kn 06/13/16 ---------------------------------------------------
00:00:00.54 231kn 07/13/15 +2.755 1... Qc4 2. Kg2 f5 3. Qxc4 bxc4
4. Rc1 d5 5. gxf5 (231833)
00:00:00.77 351kn 07/17/18 ---------------------------------------------------
00:00:01.32 683kn 08/16/18 +2.815 1... Qd3+ 2. Kg2 Nc2 3. Rf2 Nxb4
4. Qc7 Qd5+ 5. Kg3 Nd3
6. Qxa7 Nxf2 7. Qxf2 (683546)
00:00:01.63 849kn 08/16/18 ---------------------------------------------------
00:00:02.25 1222kn 09/19/20 +2.950 1... Qd3+ 2. Kg2 Nc2 3. Rf2 Nxb4
4. Qc7 Qd5+ 5. Kg3 Nc6
6. h4 e4 (1222754)
00:00:03.29 1823kn 09/19/20 ---------------------------------------------------
00:00:06.05 3552kn 10/21/24 +2.810 1... Qd3+ 2. Kg2 Nc2 3. Rf2 Nxb4
4. Qc7 Qd5+ 5. Kg3 Nc6
6. h4 b4 7. Rc2 (3552874)
00:00:08.16 4752kn 10/24/25 ---------------------------------------------------
00:00:12.27 7172kn 11/24/26 +3.050 1... Qd3+ 2. Kg2 Nc2 3. Kf2 Nxb4
4. Qe3 Qxe3+ 5. Kxe3 d5
6. Rc1 Na6 7. h4 f6 (7172486)
00:00:17.38 10Mn 11/25/28 ---------------------------------------------------
00:00:29.83 17Mn 12/25/28 +3.000 1... Qd3+ 2. Kg2 Nc2 3. Rf2 Nxb4
4. Qc7 Qd5+ 5. Kh2 Nc6
6. h4 e4 7. Kg3 Qe5+
8. Kg2 Qd5 (17028716)
00:00:43.16 24Mn 12/27/28 ---------------------------------------------------
00:00:57.72 33Mn 13/27/30 +3.025 1... Qd3+ 2. Kg2 Nc2 3. Rf2 Nxb4
4. Qc7 Qd5+ 5. Kg3 Nc6
6. h4 e4 7. a6 b4
8. Kf4 Qd4 (33130578)
00:02:42.97 96Mn 13/29/32 +3.320 1... Qc4 2. Qb1 Nc2 3. Kh4 Nxb4
4. Rc1 Qe2 5. Kg3 Qe3+
6. Kg2 Nd3 7. Rf1 Qe2+
8. Kg1 b4 (96048298)
00:02:48.33 99Mn 13/29/32 ---------------------------------------------------
00:04:18.60 154Mn 14/30/34 +3.370 1... Qc4 2. Qb1 Nc2 3. Kh4 Nxb4
4. Rc1 Qe2 5. Rf1 d5
6. Qd1 Qc4 7. Qf3 f5
8. Kg5 (154778146)
00:04:52.32 174Mn 14/30/34 ---------------------------------------------------
I don't know why.
It may be an error, I don't know. But in this game, I feel that Chezzz's lack of
understanding of passers is wrong. Maybe it evaluates passers too high. It
certainly does not know anything about outside or potential/hidden passers. I
really want to implement some more knowledge about passers, since it seems to be
a weak point of Chezzz, but I can't figure out what knowledge is missing, or how
people in general take care of such situations (potential and outside passers
for instance).
>Unfortunately in that score of -1.4, PM is underestimating the Chezzz's passers
>and the score drops again to -2.4 here:
>
>[D]8/p5k1/8/Pn1ppp2/1Pp5/5K1P/8/6R1 b - - fmvn 46; hmvc 1; id "PostModernist vs
>Chezzz, move 46"; sm Kf6
>
>I don't think Kf6 is right in this position, although it looks like it ought to
>be right. What happens is that the h3 pawn starts to get dangerous, and it's
>hard for the King and Knight to keep control with threats on both sides of the
>board. With Kh7, although it looks counter-intuitive, I think that the four
>passers in the centre can overrun the King and Rook.
>
You're probably right. But I have a feeling that some move not too long before
this would not have been Chosen by Chezzz if it knew more about the passers that
white gets.
>My program gives the h3 pawn an "outside" bonus for being on the edge of the
>board, but it's *tiny*. If the King was on the c-file (maybe the d-file, I can't
>remember), the bonus would start to get bigger, because both Black's pieces are
>far away from the pawn. In this game, the situation is confused by the presence
>of the Rook on the g-file when the King is on the f-file, I think. Because of
>the Rook, the King might as well be on the c-file, as he can't ever interfere
>with the Pawn.
>
>Note that PM didn't have a positive score in the ending until after Ne6,
>although the mistake is much earlier, I believe.
>
Chezzz's score dropped also after Ne6, but "only" from 3.x to 2.x. Thinking
about move 50, Chezzz suddenly realizes that something is wrong:
time nodes depth score pv
00:00:00.01 37n 01/00/02 +3.880 50... axb6 51. axb6 (37)
00:00:00.03 70n 01/00/04 ---------------------------------------------------
00:00:00.03 96n 02/02/03 +3.880 50... axb6 51. axb6 (96)
00:00:00.03 142n 02/02/04 ---------------------------------------------------
00:00:00.05 232n 03/04/05 +4.213 50... axb6!
00:00:00.05 314n 03/04/05 +4.230 50... axb6 51. axb6 d4 (314)
00:00:00.05 615n 03/04/06 ---------------------------------------------------
00:00:00.06 1194n 04/05/06 +4.030 50... axb6 51. axb6 d4 52. Ke2 (1194)
00:00:00.06 1822n 04/06/07 ---------------------------------------------------
00:00:00.08 4165n 05/06/08 +4.055 50... axb6 51. axb6 Nd8 52. Ke3 Nb7 (4165)
00:00:00.09 7111n 05/08/10 ---------------------------------------------------
00:00:00.09 15kn 06/09/10 +3.880 50... axb6 51. axb6 Nc5 52. h5 Nb7 (15626)
00:00:00.12 22kn 06/09/11 ---------------------------------------------------
00:00:00.14 39kn 07/10/12 +3.955 50... axb6 51. axb6 Nc5 52. h5 d4
53. Rg6+ Ke7 54. Ke2 (39406)
00:00:00.19 56kn 07/11/12 ---------------------------------------------------
00:00:00.26 112kn 08/13/14 +4.030 50... axb6 51. axb6 Nc5 52. Ke2 d4
53. Rg3 c2 54. b7 (112805)
00:00:00.33 143kn 08/13/14 ---------------------------------------------------
00:00:00.84 498kn 09/13/14 +3.610 50... axb6 51. a6 Nc7 52. a7 Kf7
53. Rg3 d4 54. Rg5 Na8
55. Rxf5+ (498842)
00:00:01.00 586kn 09/15/16 ---------------------------------------------------
00:00:02.12 1384kn 10/15/18 +2.835 50... axb6 51. a6 Nc7 52. h5 c2
53. a7 d4 54. Rg6+ Ke7
55. Rc6 Na8 56. Rxc2 (1384938)
00:00:02.61 1720kn 10/20/21 ---------------------------------------------------
00:00:04.51 3130kn 11/18/20 +2.855 50... axb6 51. axb6 Nd8 52. h5 d4
53. h6 e4 54. h7 Nf7
55. b7 e3+ 56. Ke2 (3130977)
00:00:05.47 3786kn 11/22/25 ---------------------------------------------------
00:00:12.26 8598kn 12/21/22 +2.100 50... axb6 51. axb6 Nc5 52. h5 c2
53. Ke2 d4 54. Kd2 d3
55. b7 Nxb7 56. Kxd3 (8598811)
00:00:14.80 10Mn 12/23/24 ---------------------------------------------------
00:00:28.19 19Mn 13/23/25 +0.595 50... axb6 51. axb6 Nc5 52. h5 c2
53. h6 c1=Q 54. Rxc1 Ne4+
55. Ke2 Nd6 56. h7 Kg7
57. Rc6 (19726686)
00:00:37.31 26Mn 13/25/26 ---------------------------------------------------
00:01:28.08 61Mn 14/26/28 -0.170 50... axb6 51. a6 Kf7 52. h5 c2
53. a7 Nc7 54. Rc1 Na8
55. Rxc2 Kf6 56. Rc8 Kg5
57. Rxa8 (61070836)
00:01:40.20 69Mn 14/28/30 ---------------------------------------------------
It takes 28 seconds to realize that something is wrong. And that's after having
reached the position. That seems to indicate that Chezzz could use some form of
knowledge that would have evaluated the position at move 50 much lower. Then
this position would have been discarded by search. I just don't know what that
missing knowledge is. Something with passers...
/David
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.