Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Non-reproducible blunder in Rybka automatic game
- - By ernest (****) Date 2007-12-31 16:34
On my Core 2 Duo @3GHz, I set up in the ChessBase/Shredder8 GUI a 100 game match between Rybka 2.3.2a 64-bit (single core) and Rybka 2.3.2a mp 64-bit (limited to 1 core, Max CPUs = 1), 4'+2", with pondering, 512MB hash. OS: Win XP Pro x64

I always watch the first few games, in order to check if everything is OK.

And the first game went like this:

(751) Rybka 2.3.2a 64-bit - Rybka 2.3.2a mp 64-bit [A07]
Ryb232a - Ryb232amp1p, pond, Blitz:4'+2 xxx-PC (1), 30.12.2007
[–0.06;–0.01]
W=15.2 ply; 238kN/s; 1 292 TBAsB=15.6 ply; 230kN/s; 620 TBAs 1.Nf3 d5 2.g3 Nf6 3.Bg2 c6 4.0–0 Bf5 5.b3 Nbd7 6.d4 e6 7.c4 h6 8.Nbd2 Be7 9.Bb2 0–0 10.Ne5 a5 11.a3 Nxe5 12.dxe5 Nd7 13.cxd5 exd5 14.e4 dxe4 15.Bxe4 Bh3 16.Re1 Nc5 17.Qc2 Qc7 18.Bh7+ Kh8 19.Nc4 Nxb3 20.Rad1 Both last book move 20...Rad8  -0.01/14  11  21.Bf5  -0.06/14  8  Bxf5  0.00/15  1  22.Qxf5  0.00/14  2  Rd5 (Nc5)  -0.08/15  10  23.Qf3  -0.07/15  15  Rxd1  -0.15/16  7  24.Rxd1  -0.16/16  7  Nc5  -0.18/16  12  25.Nd6 (Bc3)  -0.14/15  0  25...Kg8  -0.16/15  12  26.Qf5 (Nf5)  -0.23/16  51  26...b6  -0.27/15  7  27.Bc3  -0.29/16  14  Bxd6  -0.35/17  5  28.exd6  -0.32/16  6  Qd7  -0.29/17  8  29.Qc2 (Qe5)  -0.32/16  7  29...Qg4  -0.44/16  10  30.Rd4 (Bd4)  -0.42/15  10  30...Qg6  -0.44/15  3  31.Qb2  -0.46/15  2  Ne6  -0.61/15  2  32.Rd1  -0.69/15  7  Qg4  -0.70/15  3  33.Rc1  -0.71/14  5  b5  -0.87/14  3  34.Qd2 (Qc2)  -0.86/14  13  34...b4  -0.98/14  3  35.axb4  -1.00/15  7  axb4  -1.07/17  8  36.Bb2  -1.08/16  4  c5  -1.14/17  15  37.Qd3  -1.14/16  19  Rd8 (Qg6)  -1.15/17  26  38.Be5  -0.99/15  4  Ra8 (Qg6)  -0.94/16  21  39.Qd5  -0.90/13  5  Re8 (Rb8)  -1.06/16  15  40.f4 (Re1)  -0.89/13  5  40...Qf5  -1.04/15  5  41.Qc4  -1.04/14  5  Rd8  -1.00/14  4  42.Rd1  -0.88/14  6  Qg4 (g5)  -1.01/14  3  43.Rd5 (Re1)  -0.94/14  6  43...h5  -1.03/12  1  44.Kf2  -1.01/13  0  h4  -1.10/14  4  45.Rd3  -1.10/13  0  Qh3 (Qf5)  -1.03/13  1  46.Kg1  -1.10/12  1  Qf5  -1.03/13  1  47.Kf2  -1.17/14  10  f6  -1.14/15  17  48.Bb2  -1.22/15  3  Kf7  -1.19/14  2  49.Ke1 (Kg1)  -1.49/14  2  49...g6 (Rh8)  -1.25/12  1  50.Bc1 (Ba1)  -1.33/11  3  50...Kg7 (Qh3)  -1.59/11  4  51.Bb2  -1.72/13  4  Nd4  -1.68/14  1  52.Bxd4  -1.44/12  0  Rxd6  -1.68/14  1  53.Kd2 (Kf2)  -1.53/14  2  53...Qxd3+ (cxd4)  -6.14/14  1  54.Kxd3  -6.29/17  24  Rxd4+  -6.69/17  0  55.Qxd4  -6.30/17  18  cxd4  -6.47/18  8  56.Kxd4 (Kc4)  -6.70/16  6  56...hxg3  -6.46/14  1  57.hxg3  -6.51/15  12  Kh6  -6.87/16  0  58.g4  -13.81/15  10  f5 -7.16/15  15  0–1

Move 53.Kd2 is a blunder, but a non-reproducible one: only the correct 53.Kf2 can be reproduced.

8/6k1/3r1pp1/2p2q2/1pQB1P1p/3R2P1/7P/4K3 w - - 0 53


Kd2???     Kf2!

Any idea how such things can occur: hash problem, cosmic ray :-) ?
Parent - By BB (****) Date 2007-12-31 16:47
Non-reproducible blunder

The technical phrase is Heisen-bug. :)
Parent - - By Vasik Rajlich (Silver) Date 2008-01-03 14:35
This looks normal, it takes some time for Rybka to get a score drop here. Apparently, the K+P endgame needs to be searched quite deeply.

Vas

New game
8/6k1/3r1pp1/2p2q2/1pQB1P1p/3R2P1/3K3P/8 b - - 0 1


Analysis by Rybka 2.3.2g7 mp 32-bit bench :

1...Rxd4
  -+  (-1.79)   Depth: 2   00:00:00
1...Rxd4
  -+  (-1.73)   Depth: 3   00:00:00
1...Rxd4
  -+  (-1.67)   Depth: 4   00:00:00
1...Rxd4 2.Rxd4 cxd4 3.Qc7+ Kh6
  -+  (-1.54)   Depth: 5   00:00:00
1...cxd4 2.Qxb4
  -+  (-1.61)   Depth: 5   00:00:00
1...cxd4 2.Qxb4 Qd5
  -+  (-1.64)   Depth: 6   00:00:00  10kN
1...cxd4 2.Qxb4 Qd5 3.f5
  -+  (-1.63)   Depth: 7   00:00:00  22kN
1...cxd4 2.Qxb4 Qd5 3.Kd1 Qe4
  -+  (-1.69)   Depth: 8   00:00:00  41kN
1...cxd4 2.Qxb4 Qd5 3.Rb3 Qg2+ 4.Kd1 Rd7
  -+  (-1.85)   Depth: 9   00:00:01  88kN
1...cxd4 2.Qxb4 Qd5 3.Rb3 Qg2+ 4.Kd1 Rd7 5.Qd2
  -+  (-1.84)   Depth: 10   00:00:01  104kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Qc3+ 7.Ke2 f5 8.Qf8+
  -+  (-1.83)   Depth: 11   00:00:04  292kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Kg7 7.Qa7+ Kh6 8.gxh4
  -+  (-2.01)   Depth: 12   00:00:06  396kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Kg7 7.Qa7+ Kf8 8.Qd4
  -+  (-1.98)   Depth: 13   00:00:07  525kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Qc3+ 7.Ke2 Qc4+ 8.Kf3
  -+  (-1.85)   Depth: 14   00:00:13  920kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Qc3+ 7.Ke2 Qc4+ 8.Kf3
  -+  (-1.85)   Depth: 15   00:00:18  1335kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Qc3+ 7.Ke2 Qc4+ 8.Kf3
  -+  (-1.85)   Depth: 16   00:00:28  2003kN
1...cxd4 2.Rxd4 Rxd4+ 3.Qxd4 b3 4.Qa7+ Kh6 5.Qb8 Qc2+ 6.Ke3 Qc3+ 7.Ke2 Qc4+ 8.Kf3
  -+  (-1.85)   Depth: 17   00:00:45  3254kN

(Rajlich,  03.01.2008)
Parent - - By ernest (****) Date 2008-01-03 18:42
Yes, but the point was:
how could Kd2 be played during the actual game (immediate evaluation -6),
rather than Kf2 (immediate evaluation -1.6)?
Parent - By Vasik Rajlich (Silver) Date 2008-01-04 10:23
The immediate eval after Kd2 is in the -1.X range, as you can see from the analysis I posted.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Non-reproducible blunder in Rybka automatic game

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill