Deep Blue blunder

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

kgburcham
Posts: 2016
Joined: Sun Feb 17, 2008 4:19 pm

Deep Blue blunder

Post by kgburcham »

In this position Deep Blue played 44.Kf1. This move draws.
44.Kh1 wins. Too deep for Deep Blue.
Try your 8 way or four way overnight with large hash running Rybka 3.
Please post results.

[d] R7/1r3kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/6K1 w - - 0 44
Top
royb
Posts: 572
Joined: Thu Mar 09, 2006 12:53 am

Re: Deep Blue blunder

Post by royb »

It does not take Rybka 3 long to spot that Kh1 is better than Kf1:

[+2.05] d=20 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Kg2 Be5 53.Rxb5 Bxc3 54.Bxe4 Be5 55.Rc5 (0:01.36)
[+2.08] d=19 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Kg2 Be5 53.Rxb5 Bxc3 54.Rb7 Be5 55.Rf7 (0:01.07)
[+2.08] d=18 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Kg2 Be5 53.Rxb5 Bxc3 54.Rb7 Be5 55.Rf7 (0:01.02)
[+1.94] d=17 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 (0:00.40)
[+1.94] d=16 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Bxe4 Be5 53.Rxb5 Bxc3 54.Kg2 Be5 (0:00.29)
[+1.94] d=15 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Bxe4 Be5 53.Rxb5 Bxc3 54.Kg2 Be5 (0:00.24)
[+2.03] d=15 44.Kh1 (0:00.22)
[+1.83] d=14 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 (0:00.20)
[+2.32] d=13 44.Kf1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Rc7 48.Ra5 (0:00.03)
[+2.32] d=12 44.Kf1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Rc7 48.Ra5 (0:00.02)
[+2.17] d=11 44.Kf1 Rb8 45.Ra6 Qxc6 46.dxc6 Ke8 47.Ra7 Rc8 (0:00.01)
[+2.10] d=11 44.Kf1 (0:00.01)
[+1.90] d=10 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7 49.Rxb5 Ra8 (0:00.00)
[+1.90] d=9 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7 49.Rxb5 Ra8 (0:00.00)
[+1.99] d=9 44.Kh2 (0:00.00)
[+1.79] d=9 44.Kh2 (0:00.00)
[+1.59] d=8 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rc8 48.Kg3 (0:00.00)
[+1.79] d=8 44.Kh2 (0:00.00)
[+1.59] d=7 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 (0:00.00)
[+1.40] d=6 44.Kh2 Rb8 45.Ra6 (0:00.00)

This was run with 32 MB hash on a Dual Core 2.5 GHz system (under Linux).
Top
Uri Blass
Posts: 11057
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Deep Blue blunder

Post by Uri Blass »

royb wrote:It does not take Rybka 3 long to spot that Kh1 is better than Kf1:

[+2.05] d=20 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Kg2 Be5 53.Rxb5 Bxc3 54.Bxe4 Be5 55.Rc5 (0:01.36)
[+2.08] d=19 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Kg2 Be5 53.Rxb5 Bxc3 54.Rb7 Be5 55.Rf7 (0:01.07)
[+2.08] d=18 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Kg2 Be5 53.Rxb5 Bxc3 54.Rb7 Be5 55.Rf7 (0:01.02)
[+1.94] d=17 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 (0:00.40)
[+1.94] d=16 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Bxe4 Be5 53.Rxb5 Bxc3 54.Kg2 Be5 (0:00.29)
[+1.94] d=15 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Kf8 48.Rd7 Bc7 49.g4 Bb6 50.Rb7 Bc7 51.Bd5 e4 52.Bxe4 Be5 53.Rxb5 Bxc3 54.Kg2 Be5 (0:00.24)
[+2.03] d=15 44.Kh1 (0:00.22)
[+1.83] d=14 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 (0:00.20)
[+2.32] d=13 44.Kf1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Rc7 48.Ra5 (0:00.03)
[+2.32] d=12 44.Kf1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 Rc7 48.Ra5 (0:00.02)
[+2.17] d=11 44.Kf1 Rb8 45.Ra6 Qxc6 46.dxc6 Ke8 47.Ra7 Rc8 (0:00.01)
[+2.10] d=11 44.Kf1 (0:00.01)
[+1.90] d=10 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7 49.Rxb5 Ra8 (0:00.00)
[+1.90] d=9 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7 49.Rxb5 Ra8 (0:00.00)
[+1.99] d=9 44.Kh2 (0:00.00)
[+1.79] d=9 44.Kh2 (0:00.00)
[+1.59] d=8 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rc8 48.Kg3 (0:00.00)
[+1.79] d=8 44.Kh2 (0:00.00)
[+1.59] d=7 44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7 (0:00.00)
[+1.40] d=6 44.Kh2 Rb8 45.Ra6 (0:00.00)

This was run with 32 MB hash on a Dual Core 2.5 GHz system (under Linux).
It is one of the things that I hate about rybka.
The last time rybka showed a score for Kf1 it was +2.32(depth 13)
Rybka never showed more than +2.32 for Kh1

It is clear that Rybka has a failed low at depth 14 for Kf1
What was the last score of rybka for Kf1 at depth 14 and why does she hide the score?

Maybe rybka is not interested in exact score but she could at least say something like
[<=+1.82] d=14 44.Kf1 0:00.14

Uri
Top
ernst
Posts: 354
Joined: Thu Mar 09, 2006 6:00 pm

Re: Deep Blue blunder

Post by ernst »

Shredder doesn't need much time either on my Q6600 to prefer Kh1.

Analysis by Deep Shredder XP:

44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Kh7 48.c7 Re8
+/- (1.20) Depth: 6/14 00:00:00 38kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Kh7 48.Rd7 Bf8
+/- (1.37) Depth: 7/19 00:00:00 60kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Kh7 48.Rd7 Bf8 49.c7 Re8
+/- (1.34) Depth: 8/16 00:00:00 87kN
44.Kh2 Rb8
+- (1.59) Depth: 9/19 00:00:00 177kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rd8 48.g4 Kg8 49.Rb7 Kh7 50.Kg3 Rb8 51.c7
+- (1.57) Depth: 10/24 00:00:00 308kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rd8 48.g4 Kg8 49.Rb7 Rb8 50.Rd7 Bf8 51.c7 Re8
+- (1.68) Depth: 11/21 00:00:00 447kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rd8 48.g4 Kg8 49.Rb7 Kf8 50.Kg3 Ra8 51.Rxb5 Ra3 52.Rb7 Rxc3+ 53.Kh4
+- (1.75) Depth: 12/27 00:00:00 728kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rd8 48.g4 Rc8 49.Rb7 Rc7 50.Rxb5 Ra7 51.Kg1 Ra1+ 52.Kf2 Ra3 53.Rb7 Rxc3
+- (1.74) Depth: 13/27 00:00:00 1030kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rd8 48.g4 Rc8 49.Rb7 Rc7 50.Rb8+ Ke7 51.Rxb5 Rc8
+- (1.81) Depth: 14/31 00:00:00 1684kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rd8 48.g4 Rc8 49.Rb7 Rc7 50.Rxb5 Ra7
+- (1.81) Depth: 15/33 00:00:00 2438kN
44.Kh2 Rb8
+- (2.06) Depth: 16/36 00:00:01 5057kN
44.Kh2 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Rd8 48.Rb7 Kh7 49.Kg3 Bf8
+- (2.16) Depth: 17/40 00:00:02 9866kN
44.Kh2 Rb8 45.Ra6 Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qxb5 Qf4+ 49.g3 Qxe4 50.Ra2 Qxf5 51.Qxc4 e4 52.Qd4 e3 53.Re2 Qe5 54.Qxe5 Rxe5 55.c4 Ke7 56.g4
+- (1.75) Depth: 18/53 00:00:16 54168kN, bb=4
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Rc8 48.Rb7 h5 49.Kh2 Kh7 50.Rxb5 Rc7 51.Ra5 Rc8
+- (2.12) Depth: 18/53 00:00:25 85651kN, bb=9
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Rc8 48.Rb7 h5 49.Kh2 Kh7 50.Rxb5 Rc7 51.Kg3 Rc8 52.Kh4 Kh6
+- (2.33) Depth: 19/49 00:00:28 96254kN, bb=9
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Rc8 48.Rb7 h5 49.Kh2 Kh7 50.Rxb5 Rc7 51.Kg3 Rc8 52.Kh4 Kh6
+- (2.33) Depth: 20/49 00:00:30 105mN, bb=9
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Rc8 48.Rb7 h5 49.Kh2 Kh7 50.Rxb5 Rc7 51.Kg3 Rc8 52.Kh4 Kh6 53.Bd5 Re8 54.Bxc4 e4 55.Rb7 e3 56.Be2 Rb8 57.Bxh5 Rxb7
+- (2.30) Depth: 21/54 00:00:38 132mN, bb=9
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Kg8 47.Ra7 Rc8 48.Rb7 h5 49.Kh2 Kh7 50.Rxb5 Rc7 51.Kg3 Rc8 52.Kh4 Kh6
+- (2.31) Depth: 22/57 00:01:02 211mN, bb=20
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rc8 48.Rb7 h5 49.Kh2 h4 50.g3 Rc7 51.Rb8+ Ke7 52.Ra8 hxg3+ 53.Kxg3 Kf7 54.Kf3 Bf8 55.Kg4 Ke7 56.Rb8 Kf7
+- (2.63) Depth: 23/59 00:02:31 537mN, bb=39
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra5 Ke7 48.Rxb5 h5 49.Kh2 Rb8 50.Ra5 Bc7 51.Rc5 h4 52.g3 hxg3+ 53.Kxg3 Rd8 54.Rxc4 Rd1 55.b5 Rg1+ 56.Kh2 Rd1 57.Kg2 Rd2+
+- (2.59) Depth: 24/62 00:04:06 863mN, bb=817
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra8 Bf8 49.Ra5 Bd6
+- (2.73) Depth: 25/63 00:07:10 1535mN, bb=4357
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Kf8 48.Rb7 h5 49.Kh2 h4 50.g3 hxg3+ 51.Kxg3 Ke8 52.Rxb5 Ke7 53.Ra5 Bc7 54.Rc5 Rd8 55.Kg4 Rd6 56.Kh5 Rd8 57.Rxc4 Rh8+
+- (2.82) Depth: 26/64 00:13:34 2950mN, bb=6558
44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Kf8 48.Rb7 h5 49.Kh2 h4 50.g3 hxg3+ 51.Kxg3 Ke8 52.Rxb5 Ke7 53.Ra5 Bc7 54.Rc5 Rd8 55.Kg4 Rd6 56.Kh5 Rd8 57.Rxc4 Rh8+
+- (2.82) Depth: 27/67 00:21:18 4580mN, bb=88982
Top
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Deep Blue blunder

Post by bob »

kgburcham wrote:In this position Deep Blue played 44.Kf1. This move draws.
44.Kh1 wins. Too deep for Deep Blue.
Try your 8 way or four way overnight with large hash running Rybka 3.
Please post results.

[d] R7/1r3kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/6K1 w - - 0 44
It will only matter if you can force Kf1 and then let Rybka find the winning line. Most programs would probably prefer Kh1 for king safety, but for the wrong reason... Crafty finds Kh1 quickly but it is based on king safety more than anything else...
Top
Dirt
Posts: 2851
Joined: Wed Mar 08, 2006 10:01 pm
Location: Irvine, CA, USA

Re: Deep Blue blunder

Post by Dirt »

bob wrote:It will only matter if you can force Kf1 and then let Rybka find the winning line.
Since Kf1 is supposed to be a draw, I think you must mean "It will only matter if you can force Kf1 and then let Rybka find the drawing line." Finding a winning line after Kh1 looks like it may be comparatively easy.
Top
User avatar
Dr.Wael Deeb
Posts: 9773
Joined: Wed Mar 08, 2006 8:44 pm
Location: Amman,Jordan

Re: Deep Blue blunder

Post by Dr.Wael Deeb »

It looks like a trivial position buddy,why did you think that it will take Rybka or any other chess program a whole night to find the winning move :!: :?:
In extream criticize mood regards,
Dr.Wael Deeb

P.S.A free advice:check the position with your friends ultra-mega powerful hardware before posting it as world shaking tough position....
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
Top
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Deep Blue blunder

Post by bob »

Dr.Wael Deeb wrote:It looks like a trivial position buddy,why did you think that it will take Rybka or any other chess program a whole night to find the winning move :!: :?:
In extream criticize mood regards,
Dr.Wael Deeb

P.S.A free advice:check the position with your friends ultra-mega powerful hardware before posting it as world shaking tough position....
The main question is can a computer find that Kh1 is better than Kf1 by determining that Kf1 leads to a forced draw, or do they find it purely by serendipity, which is not so interesting. This is like fine #70 in that regard. Nobody gets this by evaluation. The "normal" move is Kb2 and leads to a dead draw, the surprise is Kb1 that wins. Computers get it by searching deep enough to see winning a pawn, whereas if you limit the depth to 15 plies or less, only a rare program will choose the right move. Finding the queen move and the proving it leads to a draw after Kf1 is a deep challenge in the DB position.
Top
TSP

Re: Deep Blue blunder

Post by TSP »

I am leaving to go to a funeral, so don't have time to check pv out but here is the Rybka analysis I have after forcing Kf1.

[+0.00] d=22 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Rf5 Kg4 13.Rb5 Rxh5 14.Rxb4 Kf3 (0:16.48)
[+0.00] d=21 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Rf5 Kg4 13.Rb5 Rxh5 14.Rxb4 Kf3 (0:07.42)
[+0.00] d=20 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Rf5 Kg4 13.Rb5 Rxh5 14.Rxb4 (0:03.27)
[+0.00] d=19 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Rf5 Kg4 13.Rb5 Rxh5 14.Rxb4 (0:00.54)
[-0.14] d=18 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Re5 Rh6 (0:00.01)
[-0.14] d=17 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Re5 Rh6 (0:00.00)
[-0.14] d=16 1...Qb8 2.Re2 Rd8 3.f3 Qb5 4.Qxb5 cxb5 5.fxg4 Rxd7 6.gxh5 Kxg3 7.Rf2 Rh7 8.Rxf6 Rxh5 9.Ra6 Rh7 10.Ra5 b4 11.h5 Kf3 12.Re5 Rh6 (0:00.00)
[-1.17] d=4 1...Rb8 2.Rxg4 (0:00.00)
[-1.17] d=3 1...Rb8 2.Rxg4 (0:00.00)
[-1.17] d=2 1...Rb8 2.Rxg4 (0:00.00)
Top
Dirt
Posts: 2851
Joined: Wed Mar 08, 2006 10:01 pm
Location: Irvine, CA, USA

Re: Deep Blue blunder

Post by Dirt »

TSP wrote:I am leaving to go to a funeral, so don't have time to check pv out but here is the Rybka analysis I have after forcing Kf1.
...
[-1.17] d=2 1...Rb8 2.Rxg4 (0:00.00)
After forcing Kf1 I get this position:
[d]R7/1r3kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/5K2 b - - 1 1
1...Rb8 looks normal, but 2.Rxg4 doesn't look possible.
Top

Return to “Computer Chess Club: General Topics”