Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka analysing the famous Kasparov-Deep Blue position
- - By Sesse (****) Date 2007-06-27 22:18
Hi,

Just for fun, I started reading up on the famous Kasparov-Deep Blue matches. For those of you that don't know the story: Deep Blue eventually defeated Kasparov, but Kasparov accused IBM of cheating, having a human to help with the moves. One of the most well-known positions, from game 2 of the 1997 rematch (which Deep Blue won 3.5-2.5, after losing 4-2 in the original match), is this:

r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - 0 1


White (played by Deep Blue, the upgraded 1997 version) to move. Kasparov expected the computer to choose Qe6 (grabbing two pawns), but it instead played Be4!, leading to a resignation by Kasparov in under ten moves. I input the situation into Rybka 2.3.2a (multi-processor) on my dual-core Athlon 64 3800+ and let it churn for a while. (Note that Deep Blue was reportedly able to calculate 200 million nodes/sec, while Rybka on my machine does only about 102000 nodes/sec in this position. It probably doesn't help that I have to run the 32-bit version, as my server runs Linux and WINE is quite incapable of running 64-bit programs.)

In the beginning, Qb6 (the move Kasparov anticipated) comes out on top, with +0.81. Be4 is at that time rated second best, with +0.76, ie. close! -- actually it's rated the best for a very short time before Qb6 takes over. However, as soon as it's reaching 17 ply (after a minute or so), Qb6 drops to +0.25, and at 18 ply drops to +0.15 (fourth best). At 20 ply, Be4 is still rated highest by far, with +0.75, with Qe3 at +0.17. Qb6 is at +0.09, not even remotely close. For reference, the full PV at 21 ply (where I tired and quit) is: 37. Be4 Rcb8, 38. Ra6 Qd8, 39. Kh2 Kf8, 40. R6a5 Rxa5, 41. Rxa5 Rb6, 42. Kg3 Qb8, 43. Qa2, Qc7. Deep Blue's PV was 37. Be4 Rcb8, 38. g3 Qd8, 39. Kg2 Rxa2, 40. Rxa2 Bc7, 41. Qa7 Bb6
42. Qa6 Qd7, at least according to IBM. (It looks quite different to me, but I'm a relatively weak player, and you wouldn't expect two so different programs to come up with the exact same line anyway.)

Of course, this neither proves nor disproves Kasparov's accusations of Be4 coming from (or with the help of) a human, but at least it shows that a computer can indeed resist the temptation of Qb6 and find the move that shocked Kasparov and won the game. Assuming Rybka has not been hand-tuned to find the move in exactly this situation, that is :-)

/* Steinar */
Parent - - By Conwy (*) Date 2007-06-27 22:29
Hi, I checked my Rybka on the same position, with very similar results. I have an X2 4400+ which I can run at 4800 when I choose, so it's slightly faster than yours.

Note that the attractions of Be4 are starting to slide a little.

12.01   0:01   +0.84   1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 h5 5.Ra6 Rb8 (146.368) 138
13.01   0:01   +0.96   1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 h5 5.Ra6 Rb8 6.Qc6 (235.752) 134
14.01   0:02   +0.71   1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 Rb8 5.Qc6 Qxc6 6.dxc6 Bc7 7.Kh2 Bb6 (354.537) 135
15.01   0:04   +0.79   1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 Rb8 5.Qc6 Qxc6 6.dxc6 Bc7 7.Kh2 Bb6 (554.994) 138
16.01   0:09   +0.48   1.Qb6 Rxa2 2.Rxa2 Ra8 3.Ra5 Rxa5 4.bxa5 Qb8 5.Qxb8+ Bxb8 6.Kf2 Ba7+ 7.Kf3 Kf7 (1.412.346) 149
16.03   0:20   +0.72   1.Be4 Rcb8 2.Ra6 Qc8 3.Qa2 Rxa6 4.Qxa6 Qc7 5.Qc6 Kf8 6.Kf2 Ke7 7.Qa6 Kf7 (2.779.187) 138
17.01   0:25   +0.73   1.Be4 Rcb8 2.Ra6 Qc8 3.Qa2 Rxa6 4.Qxa6 Qc7 5.Kf2 Kf7 6.Qa7 Rb7 7.Qe3 Qd7 (3.314.395) 133
18.01   0:31   +0.73   1.Be4 Rcb8 2.Ra6 Qc8 3.Qa2 Rxa6 4.Qxa6 Qc7 5.Kf2 Kf7 6.Kg3 Qd7 7.Ra5 Kf8 (4.039.392) 132
19.01   0:37   +0.73   1.Be4 Rcb8 2.Ra6 Qc8 3.Qa2 Rxa6 4.Qxa6 Qc7 5.Kf2 Kf7 6.Kg3 Rb6 7.Qa7 Rb7 (4.874.386) 131
20.01   1:26   +0.73   1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Bc7 (11.009.968) 130
21.01   1:52   +0.72   1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Qa2 Qd7 (14.993.733) 136
22.01   2:34   +0.69   1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Qa2 Qd7 (21.664.158) 143
23.01   3:43   +0.69   1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Qa2 Qd7 (31.475.623) 144

Cheers,

Conwy
Parent - - By turbojuice1122 (Gold) Date 2007-06-27 22:35
Let's keep in mind that Deep Blue didn't calculate past ply 10 or 11, so the benefit of Be4 would not have been apparent in its horizon unless there was a bug.
Parent - By Alkelele (****) Date 2007-06-27 22:37

>Let's keep in mind that Deep Blue didn't calculate past ply 10 or 11...


It did, in selected lines. Of course! It's not like search extensions were not invented back then.
Parent - - By Uri Blass (*****) Date 2007-06-27 22:42
You cannot compare plies of different programs.
Rybka does a lot of pruning when deep blue did a lot of extensions.

I believe that deep blue was clearly weaker but nominal reported depth proves nothing.

Uri
Parent - - By turbojuice1122 (Gold) Date 2007-06-28 02:45
It's true that nowadays nominal reported depth proves nothing (see, e.g. Junior and Hiarcs--Hiarcs is far stronger, but Junior searches 4-6 ply deeper per move), but in those days, it often gave a decent indication of whether moves were or were not beyond their search horizon.  As for the extensions, that's not something about which I know much, so I'll just have to trust you there.
Parent - By Uri Blass (*****) Date 2007-06-28 04:20
As far as I know deep blue did many extensions that rybka does not do like singular extensions.

deep blue was better than rybka in solving hard test positions like the nolot positions thanks to their extensions.

A test position that proves that rybka does not use singular extensions is nolot number 2
latest rybka is clearly better in it and can solve it at depth 14(this is a dramtic improvement relative to older versions of rybka) but deep thought could solve it at smaller depth.

nolot number 10 was also solved by deep thought in a short time

http://www.seanet.com/~brucemo/nolot/nolot02.htm

New game
r4rk1/pp1n1p1p/1nqP2p1/2b1P1B1/4NQ2/1B3P2/PP2K2P/2R5 w - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

1.Bg5-h6 Qc6-b5+ 2.Ke2-d1
  -+  (-1.58)   Depth: 5   00:00:00
1.Bg5-h6 Qc6-b5+ 2.Ke2-d1 Bc5-d4
  -+  (-1.83)   Depth: 6   00:00:00  9kN
1.Bg5-h6 Qc6-b5+ 2.Ke2-d1 Bc5-d4 3.Bh6xf8
  -+  (-1.83)   Depth: 7   00:00:00  27kN
1.Bg5-h6 Qc6-b5+ 2.Ke2-d2 Bc5-b4+ 3.Kd2-e3 Kg8-h8 4.Rc1-c7
  -+  (-1.79)   Depth: 8   00:00:01  74kN
1.Bg5-h6 Qc6-b5+ 2.Ke2-d2 Bc5-b4+ 3.Kd2-e3 Kg8-h8 4.Rc1-c7 Rf8-e8
  -+  (-1.79)   Depth: 9   00:00:01  102kN
1.Bg5-h6 Qc6-b5+ 2.Ke2-d2 Bc5-b4+ 3.Kd2-e3 Kg8-h8 4.Rc1-c7 Rf8-c8 5.Qf4xf7
  -+  (-1.65)   Depth: 10   00:00:04  287kN
1.Bg5-h6 Qc6-b5+ 2.Ke2-d2 Bc5-b4+ 3.Kd2-e3 Kg8-h8 4.Rc1-c7 Rf8-c8 5.e5-e6 Rc8xc7
  -+  (-2.15)   Depth: 11   00:00:07  433kN
1.a2-a4 Nb6xa4 2.Ne4-f6+ Nd7xf6 3.Bb3xa4 Qc6-a6+ 4.Qf4-c4 Bc5xd6 5.Qc4xa6 b7xa6 6.e5xd6 Kg8-g7 7.Rc1-c7 Rf8-c8
  µ  (-0.87)   Depth: 11   00:00:12  793kN
1.a2-a4 Nb6xa4 2.Ne4-f6+ Nd7xf6 3.Bb3xa4 Qc6-a6+ 4.Qf4-c4 Bc5xd6 5.Qc4xa6 b7xa6 6.e5xd6 Kg8-g7 7.Rc1-c7 Rf8-c8
  µ  (-0.87)   Depth: 12   00:00:14  912kN
1.a2-a4 Nb6xa4 2.Ne4-f6+ Nd7xf6 3.Bb3xa4 Qc6-a6+ 4.Qf4-c4 Bc5xd6 5.Qc4xa6 b7xa6 6.e5xd6 Kg8-g7 7.d6-d7 Ra8-b8
  ³  (-0.70)   Depth: 13   00:00:20  1277kN
1.a2-a4 Nb6xa4 2.Ne4-f6+ Nd7xf6 3.Bb3xa4 Qc6-a6+ 4.Qf4-c4 Bc5xd6 5.Qc4xa6 b7xa6 6.e5xd6 Kg8-g7 7.d6-d7 Ra8-b8
  ³  (-0.62)   Depth: 14   00:00:31  1878kN
1.Rc1xc5 Nd7xc5 2.Ne4-f6+ Kg8-h8 3.Qf4-h4 Qc6-b5+ 4.Ke2-e3 Qb5-d3+ 5.Ke3-f2
  +-  (1.47)   Depth: 14   00:00:39  2330kN
1.Rc1xc5 Qc6xc5 2.Ne4xc5 Nd7xc5 3.Bg5-h6 Nb6-d7 4.Bh6xf8 Ra8xf8 5.Bb3-c4 Nd7-b6 6.Qf4-d4 Nc5-e6 7.Bc4xe6 f7xe6
  +-  (3.36)   Depth: 14   00:01:06  4387kN
1.Rc1xc5 Qc6xc5 2.Ne4xc5 Nd7xc5 3.Bg5-h6 Nc5xb3 4.a2xb3 Nb6-d7 5.Bh6xf8 Ra8xf8 6.Qf4-e4 Rf8-e8 7.f3-f4 Nd7-c5
  +-  (3.57)   Depth: 15   00:01:11  4760kN
1.Rc1xc5 Qc6xc5 2.Ne4xc5 Nd7xc5 3.Bg5-h6 Nb6-d7 4.Bh6xf8 Ra8xf8 5.Bb3-c4 a7-a6 6.b2-b4 Nc5-e6 7.Bc4xe6 f7xe6
  +-  (3.61)   Depth: 16   00:01:20  5491kN
1.Rc1xc5 Qc6xc5 2.Ne4xc5 Nd7xc5 3.Bg5-h6 Nb6-d7 4.Bh6xf8 Ra8xf8 5.Bb3-c4 Nc5-e6 6.Bc4xe6 f7xe6 7.Qf4-c4 Nd7xe5
  +-  (3.65)   Depth: 17   00:01:30  6198kN

(,  28.06.2007)

New game
r1b2rk1/1p1nbppp/pq1p4/3B4/P2NP3/2N1p3/1PP3PP/R2Q1R1K w - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

1.Nd4-f5 Be7-f6
  ³  (-0.32)   Depth: 5   00:00:00
1.Nd4-f5 Be7-g5 2.Bd5-b3
  ³  (-0.41)   Depth: 6   00:00:00  14kN
1.Nd4-f5 Be7-g5 2.Qd1-h5 Bg5-f6
  ³  (-0.53)   Depth: 7   00:00:00  22kN
1.Bd5-b3 Be7-g5 2.Nc3-d5 Qb6-d8
  =  (-0.20)   Depth: 7   00:00:00  35kN
1.Bd5-b3 Qb6-c5 2.Nc3-d5 Be7-f6 3.c2-c3
  ³  (-0.28)   Depth: 8   00:00:01  45kN
1.Bd5-b3 Be7-g5 2.Nc3-d5 Qb6-a7 3.Rf1-f5 Bg5-f6
  ³  (-0.38)   Depth: 9   00:00:01  77kN
1.Bd5-b3 Be7-g5 2.Nc3-d5 Qb6-a7 3.Rf1-f5 Bg5-f6 4.c2-c3
  =  (-0.21)   Depth: 10   00:00:02  138kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-f6
  =  (-0.18)   Depth: 11   00:00:04  283kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-f6 5.Nc3-e2
  =  (-0.15)   Depth: 12   00:00:07  476kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-g5 5.Qe3-f2 Bg5-f6
  ³  (-0.28)   Depth: 13   00:00:11  703kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-f6 5.Nc3-e2 Bf6-e5 6.c2-c3
  =  (-0.11)   Depth: 14   00:00:22  1391kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-f6 5.Nc3-e2 Bf6-e5 6.c2-c3 Nh5-f6
  =  (-0.11)   Depth: 15   00:00:41  2536kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-f6 5.Nc3-e2 Bf6-e5 6.c2-c3 Nh5-f6 7.Ne2-f4
  =  (-0.11)   Depth: 16   00:01:04  4116kN
1.Bd5-b3 Nd7-f6 2.Qd1-d3 Qb6-c5 3.h2-h3 Nf6-h5 4.Qd3xe3 Be7-f6 5.Nc3-e2 Bf6-e5 6.c2-c3 Bc8-d7 7.Bb3-d5 Nh5-f6
  =  (-0.15)   Depth: 17   00:01:55  7515kN
1.Bd5xf7+ Rf8xf7 2.Rf1xf7 Be7-f6 3.Nc3-d5 Qb6xb2 4.Ra1-b1 Qb2-a2 5.Nd5xf6+ Kg8xf7 6.Nf6xd7 Bc8xd7 7.Rb1xb7 Qa2xa4
  =  (0.14)   Depth: 17   00:05:00  21173kN
1.Rf1xf7 Rf8xf7 2.Bd5xf7+ Kg8-h8 3.Nc3-d5 Qb6xb2 4.Nd5xe7 Nd7-e5 5.Ra1-b1 Bc8-g4 6.Rb1xb2 Bg4xd1 7.Bf7-d5 e3-e2
  ±  (0.94)   Depth: 17   00:05:44  23694kN

(,  28.06.2007)
Parent - By killkra (*) Date 2007-06-27 23:53
New version fruit 2.2.3 can find this move in 14 ply.
Parent - - By George Tsavdaris (****) Date 2007-06-27 23:20
Very good for Rybka, but maybe it is the previous move that was the most crucial.

In move 36 programs have to find 36.axb5! and not the tempting 36.Qb6
Does Rybka 2.3.2a finds this too after a logical time.....?

Here is the position:
r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - 0 36


And definitely the most crucial position is on move 44 where many programs play the wrong move 44.Kf1? while 44.Kh1! wins!
R7/1r3kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/6K1 w - - 0 44


Rybka - Kasparov would be the most interesting match for me. 2 monsters face each other! One with 400 eyes the other with 200! :-)
But Kasparov's days are over :-(  He retired and left us with sorrow.....
Parent - - By Sesse (****) Date 2007-06-27 23:56
To take them in opposite order: In the second position, Rybka starts off preferring Kh1, but then goes on to think Kf1 is marginally better up until 19 ply (+1.76 vs. +1.69 or something, Kh2 is between them). However, at 21 ply Kf1 drops to +1.46, and at 22 ply Kh2 increases to +2.01. (Again, this takes a minute or two on my machine.) I haven't checked if Kh1 still is the preferred beyond 23 ply (where Kh1 is +1.90 and Kf1 is +1.57).

The first position is open, and probably more complex (Rybka can get nowhere near 23 ply here). After 16 ply, which takes a minute or so, Qb6 is preferred, with +1.19, and axb5 is second with +0.70. (Kasparov is probably right that you need a good computer to understand this!) 17 ply only changes the scoring marginally, and getting there takes over three minutes. (Wonder how the win finder or the LK versions would do...) Given that, as you mention, this is the position leading up to the previously analysed one, and that needed 17 ply to get the analysis right, I'd guess >20 ply would be needed to solve this position, and in normal time controls that would probably be hopeless. Funny enough the PV for axb5 (as calculated to 19 ply) does list Be4 as the preferred next move, though; I wonder how that happens. I let it run to 20 ply, which took over ten minutes with MultiPV=3, hash size=1GB and otherwise mostly default settings (does MultiPV force the engine to consider supposedly inferior moves close so it's slowed down, or is it just an output setting?), and it still said +1.06 vs. +0.76, so it doesn't get the move right even after quite extensive calculation.

/* Steinar */
Parent - - By Sesse (****) Date 2007-06-28 00:26
Oh, followup, I forgot to quit it:

  PV 1 (+0.94) (167284 kn,  83 kn/s, 22 ply):
  Qb6, Qe7, axb5, Rab8, Qxa6, e4, Qa7, Bc7, Ra6, Qe5, b6, Qh2+, Kf1, Bg3

  PV 2 (+0.71) (195788 kn,  84 kn/s, 22 ply):
  axb5, axb5, Be4, Qd8, Kh2, Rxa2, Qxa2, Rb8, Qa7, Rb6, Kg3, Bb8, Qa5, Kf7

So perhaps, if you get to 23 ply, it _might_ be able to find it. I'll let it stay overnight and see what happens, but I guess it'll probably just end in a server timeout (I do all my analysis via FICS, for various odd reasons)... :-)

/* Steinar */
Parent - By Sesse (****) Date 2007-06-28 11:53
FWIW, after 26 ply it still prefers Qb6, although the margin is small:

  PV 1 (+0.88) (2349202 kn,  89 kn/s, 26 ply):
  Qb6, Qe7, axb5, Rab8, Qxa6, e4, Qa7, Qe5, Qe3, Bc7, Qxe4, Qxc3, d6, Bb6+

  PV 2 (+0.71) (2596732 kn,  87 kn/s, 26 ply):
  axb5, axb5, Be4, Qd8, Kh2, Rxa2, Qxa2, Rb8, Qa7, Rb6, Kg3, Bb8, Qa5, Kf7

  PV 3 (+0.36) (2996342 kn,  83 kn/s, 26 ply):
  Qe1, a5, bxa5, Rxa5, axb5, Rxa2, Rxa2, Qxb5, Be4, Qb6+, Kh2, Bc5, Qd2, Be3

I guess neither are bad moves in the end. :-)

/* Steinar */
Parent - By vroger007 (**) Date 2007-06-28 19:39
UltraOptimistic rybka prefers Kh1 over Kf1 after 21 ply (maybe sooner, I was taking a shower while it calculated).
R7/1r3kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/6K1 w - - 0 1

Analysis by Rybka 2.3.2a:
1. +- (1.82): 44.Kh1 Rb8 45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Kf8 48.Kh2 Bc7 49.Rb7 Ke7 50.Rxb5 Rb8
2. +- (1.50): 44.Kf1 Rb8 45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Ke2 Kg8
3. ± (1.04): 44.Kh2 Rb8 45.Ra6 Qe3 46.Qxd6 Re8 47.Bf3 Qf4+ 48.Kg1 Qc1+ 49.Kf2 Qd2+ 50.Kf1 Qc1+
4. ± (0.97): 44.Qxb6 Rxb6 45.Ra7+ Kf8 46.Kf1 Bb8 47.Rd7 Bd6 48.Ke2 Ra6 49.Kd2 Kg8 50.Kc2 Kh7
5. -+ (-11.84): 44.Qc5 Bxc5+ 45.bxc5 Qxc5+ 46.Kh2 Qe3 47.Bf3 Ra7 48.Rxa7+ Qxa7 49.d6 Qd7 50.Bh5+ Kf8
Roger
Parent - - By Gaмßito (****) Date 2007-06-28 03:03
Hi,

It seems that it cannot find axb5 here, at least not in my hardware with some appropiate time.

r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

1.Qb6 Bc7
  ±  (1.30)   Depth: 5   00:00:00
1.Qb6 Bc7 2.Qe6+ Qxe6
  ±  (1.18)   Depth: 6   00:00:00  5kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rab8
  ±  (1.15)   Depth: 7   00:00:00  7kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rab8 4.axb5 axb5 5.g4
  ±  (1.20)   Depth: 8   00:00:00  9kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rab8 4.axb5 axb5 5.Kf2 Kf8
  ±  (1.24)   Depth: 9   00:00:01  15kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rab8 4.axb5 axb5 5.Kf2 Kf8 6.Kf3
  ±  (1.18)   Depth: 10   00:00:01  27kN
1.Qb6 Rd8 2.Be4 Be7 3.axb5 Rd6 4.Qa5 Bd8
  ±  (1.33)   Depth: 11   00:00:02  89kN
1.Qb6 Rd8 2.Be4 Be7 3.axb5 Rd6 4.Qa5 Bd8 5.Qa4
  ±  (1.33)   Depth: 12   00:00:04  113kN
1.Qb6 Rd8 2.Be4 Rac8 3.Qxa6 bxa4 4.Qxa4 Qh5 5.Qa7 Qg5 6.Qf2
  +-  (1.47)   Depth: 13   00:00:07  251kN
1.Qb6 Rd8 2.Be4 Rac8 3.Qxa6 bxa4 4.Qxa4 Qh5 5.Qa7 Qg5 6.Qf2 Rc7
  +-  (1.46)   Depth: 14   00:00:11  372kN
1.Qb6 Qf8 2.axb5 Rab8 3.Qxa6 Qd8 4.Be4 Rb6 5.Qa4 Bc7 6.Kh2 Qd7 7.Qc2
  +-  (1.46)   Depth: 15   00:00:26  945kN
1.Qb6 Qf8 2.axb5 a5 3.Rxa5 Rab8 4.Qf2 Bc7 5.Ra6 Rxb5 6.Be4 Qd8 7.Qe2 Bb6+
  ±  (1.40)   Depth: 16   00:00:38  1489kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Qe5 5.Qe3 Re8 6.b6 Qxf5 7.Ba4 Bf4
  ±  (1.15)   Depth: 17   00:01:28  3542kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Qe5 5.Qe3 Re8 6.b6 Qxf5 7.Ba4 Red8
  ±  (1.08)   Depth: 18   00:02:01  5205kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Qe5 5.Qe3 Re8 6.b6 Qxf5 7.Ba4 Red8
  ±  (1.12)   Depth: 19   00:02:51  7585kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Qe5 5.Qe3 Re8 6.b6 Qxf5 7.Ba4 Red8
  ±  (1.16)   Depth: 20   00:04:16  11331kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Qe5 5.Qe3 Re8 6.b6 Qxf5 7.Ba4 Red8
  ±  (1.16)   Depth: 21   00:06:40  17598kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Qe5 5.Qe3 Bc7 6.Qxe4 Qxc3 7.d6 Bb6+
  ±  (0.98)   Depth: 22   00:14:15  38942kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Bc7 5.Ra6 Qe5 6.b6 Bd6 7.b7 Re8
  ±  (1.01)   Depth: 23   00:27:20  76256kN
1.Qb6 Qe7 2.axb5 Rab8 3.Qxa6 e4 4.Qa7 Bc7 5.Ra6 Qe5 6.b6 Bd6 7.b7 Re8
  ±  (1.05)   Depth: 24   00:45:49  125mN

(, AMD 27.06.2007)

And yes, Kh1 was prefered by 2.3.2a easily.


R7/1r3kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/6K1 w - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

1.Kf1 Rb8
  ±  (1.27)   Depth: 5   00:00:00
1.Kf1 Rb8 2.Qd7+ Kg8
  ±  (1.24)   Depth: 6   00:00:00  1kN
1.Kf1 Rb8 2.Qd7+ Kg8 3.Ra7
  ±  (1.30)   Depth: 7   00:00:00  2kN
1.Kf1 Rb8 2.Qd7+ Kg8 3.Ra7 Bf8
  ±  (1.27)   Depth: 8   00:00:00  3kN
1.Kf1 Rb8 2.Qd7+ Kg8 3.Ra7 Bf8 4.Qe6+ Kh7
  ±  (1.27)   Depth: 9   00:00:00  5kN
1.Kf1 Rb8 2.Ra6 Qxc6 3.dxc6 Bc7 4.Ra7 Rc8
  ±  (1.36)   Depth: 10   00:00:00  10kN
1.Kf1 Rb8 2.Ra6 Qxc6 3.dxc6 Rc8 4.Ra7+ Kf8 5.Rb7 Rc7
  +-  (1.73)   Depth: 11   00:00:00  19kN
1.Kf1 Rb8 2.Ra6 Qxc6 3.dxc6 Rc8 4.Ra7+ Rc7 5.Ra5 Ke7 6.Ke2 Kd8 7.Ke3
  +-  (1.71)   Depth: 12   00:00:01  29kN
1.Kf1 Rb8 2.Ra6 Qxc6 3.dxc6 Rc8 4.Ra7+ Rc7 5.Ra8 Ke7 6.Ke2 h5
  +-  (1.76)   Depth: 13   00:00:01  47kN
1.Kf1 Rb8 2.Ra6 Qxc6 3.dxc6 Rc8 4.Ra7+ Rc7 5.Ra8 Ke7 6.Ke2 h5 7.Ra5
  +-  (1.76)   Depth: 14   00:00:01  67kN
1.Kf1 Rb8 2.Ra6 Qe3 3.Qxd6 Re8 4.h4 h5 5.Bf3 Qc1+ 6.Kf2 Qd2+ 7.Be2 Qf4+
  +-  (1.43)   Depth: 15   00:00:04  215kN
1.Kh2 Rb8 2.Ra6 Qxc6 3.dxc6 Re8 4.Ra7+ Re7 5.Ra5 Rc7 6.Kg3 Ke7 7.Kf3 Ke8
  +-  (1.84)   Depth: 15   00:00:06  319kN
1.Kh2 Rb8 2.Ra6 Qxc6 3.dxc6 Rc8 4.Ra7+ Kg8 5.Rb7 Kf8 6.Rxb5 Ke7 7.Kg3 Bc7
  +-  (1.84)   Depth: 16   00:00:08  386kN
1.Kh2 Rb8 2.Qd7+ Kg8 3.Ra7 Bf8 4.Qe6+ Kh7 5.Qxb6 Rxb6 6.Kg3 Bd6 7.Kg4 Bb8
  ±  (1.14)   Depth: 17   00:00:23  1127kN
1.Kf1 Rb8 2.Ra6 Qe3 3.Qxd6 Re8 4.h4 h5 5.Bf3 Qc1+ 6.Kf2 Qd2+ 7.Be2 Qf4+
  ±  (1.25)   Depth: 17   00:00:30  1532kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Kh2 Bc7 6.Rb7 Ke7 7.Rxb5 Rd8
  +-  (1.63)   Depth: 17   00:00:33  1650kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Kh2 Bc7 6.Rb7 Ke7 7.Rxb5 Rd8
  +-  (1.65)   Depth: 18   00:00:35  1748kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.Kh2 Bb6 7.Rb7 Be3
  +-  (1.70)   Depth: 19   00:00:39  1943kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.Kh2 Bb6 7.Rb7 Bc7
  +-  (1.84)   Depth: 20   00:00:52  2576kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.Kh2 h5 7.Kg3 Bb6
  +-  (1.84)   Depth: 21   00:01:04  3239kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.Kh2 Bb6 7.Rb7 Bc7
  +-  (1.80)   Depth: 22   00:01:24  4459kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.Kh2 Bb6 7.Rb7 Bc7
  +-  (1.85)   Depth: 23   00:02:14  7000kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.Kh2 Bb6 7.Rb7 Bc7
  +-  (1.85)   Depth: 24   00:03:01  9754kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.g3 Bb6 7.Rb7 Bc7
  +-  (2.02)   Depth: 25   00:06:17  19374kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Kf8 4.Ra7 Rc8 5.Rd7 Bc7 6.g3 Bb6 7.Rb7 Bc7
  +-  (2.07)   Depth: 26   00:10:49  31748kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Re8 4.Ra5 Bc7 5.Rxb5 Rb8 6.Rc5 Ke7 7.Rxc4 Bd6
  +-  (2.06)   Depth: 27   00:18:31  54446kN
1.Kh1 Rb8 2.Ra6 Qxc6 3.dxc6 Re8 4.Ra5 Bc7 5.Rxb5 Rb8 6.Rc5 Ke7 7.Rxc4 Bd6
  +-  (2.10)   Depth: 28   00:29:32  87305kN

(, AMD 27.06.2007)

Regards,
Gambito.
Parent - By Uri Blass (*****) Date 2007-06-28 05:07 Edited 2007-06-28 05:10
There is a reason for that
I believe that
axb5 of deep blue was not the best move and Qb6 was stronger.

Uri
Parent - - By Uri Blass (*****) Date 2007-06-28 05:09
Why do you think that program have to find 36.axb5
I believe 36.Qb6 is simply a better move.

I remember that some analysis in the past showed that white has an advantage with 36.Qb6
Parent - By Gaмßito (****) Date 2007-06-28 08:21
Yes, I also believe it. I was only checking that line with Rybka and of course it prefers 36. Qb6 as the stronger and not Deep Blue move 36. axb5.

After 36. Qb6, black seems to have a dangerous attack with Qe7 37. axb5 Rab8 38. Qxa6 and now e4! threatening Qe5 and Qh2+.

But note that it seems that even after 36. axb5 played by Deep Blue, its position seems to be enough good. I think that Deep Blue did a last and a clearly mistake in their last moves, with 45. Ra6. Note that Rybka can see easily the very strong black reply move (46. Qe3!) after 45. Ra6 and this is why it changes quickly their thoughts and does not want to play 45. Ra6 anymore. It prefers clearly 45. Qd7+ and it still says, white has a strong position. I am not making some deep analisys here to verify if 45. Qd7+ can even win this game for white or not, but Rybka thinks it gives some chances. How good is really this move?

But also note that Rybka is avoiding the simplest 45. Qxb6, that gives white a winning game. I think that this line is not very difficult to see for a strong player:

45. Qxb5 Rxb5 46. Ra7+ Kf8 47. Kf2 Bb8 48. Rd7 Bd6 49. Rd8+ Kf7 50. Ra8 Bb8 51. Kg3 Ke7 52. Kh4! Kf7 53. Kh5 and white goes directly for a victory.

Here Rybka thinks 45. Qd7!

Rr6/5kp1/1qQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/5K2 w - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

45.Qd7+ Kg8 46.Ra7
  ±  (1.30)   Depth: 5   00:00:01
45.Qd7+ Kg8 46.Ra7 Bf8
  ±  (1.27)   Depth: 6   00:00:01  1kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.d6
  +-  (1.46)   Depth: 7   00:00:01  3kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.d6 Qe3
  +-  (1.52)   Depth: 8   00:00:01  4kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Ke2
  ±  (1.20)   Depth: 9   00:00:01  10kN
45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Kf8 48.Rb7 Rc7
  +-  (1.68)   Depth: 9   00:00:01  25kN
45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 h5 48.Ke2 Rc8 49.Ke3
  +-  (1.74)   Depth: 10   00:00:01  31kN
45.Ra6 Qxc6 46.dxc6 Kf8 47.Ra7 Rc8 48.Rb7 h5 49.Kf2
  +-  (1.81)   Depth: 11   00:00:02  41kN
45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Ke7 49.Kf2 Rc8 50.Kf3
  +-  (1.86)   Depth: 12   00:00:02  69kN
45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Ke7 49.Kf2 Rc8 50.Kf3 Bc7
  +-  (1.80)   Depth: 13   00:00:03  84kN
45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Ke7 49.Kf2 Rc8 50.Kf3 Bc7 51.Rxb5 Ra8
  +-  (1.85)   Depth: 14   00:00:04  128kN
45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Ke7 49.Kf2 Rc8 50.Kf3 Bc7 51.Rxb5 Rd8
  +-  (1.88)   Depth: 15   00:00:06  191kN
45.Ra6 Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra8 Bf8 49.Ke2 h5 50.Ra5 Bd6 51.Rxb5 Ke7
  +-  (2.10)   Depth: 16   00:00:10  373kN
45.Ra6 Qe3 46.Qxd6 Re8 47.h4 h5 48.Bf3 Qc1+ 49.Kf2 Qd2+ 50.Be2 Qf4+ 51.Ke1 Qc1+
  ±  (1.03)   Depth: 17   00:00:45  2063kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Bh5 Kg8 51.Ke2 Rb8
  +-  (1.42)   Depth: 17   00:00:49  2304kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Bh5 Kg8 51.Ke2 Rb8
  +-  (1.54)   Depth: 18   00:00:54  2598kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Bh5 Kg8 51.Ke2 e4
  ±  (1.39)   Depth: 19   00:01:08  3358kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Bh5 Kg8 51.Ke2 e4
  ±  (1.39)   Depth: 20   00:01:21  4181kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Bh5 Kg8 51.Ke2 e4
  ±  (1.39)   Depth: 21   00:01:45  5629kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Bh5 Kg8 51.Ke2 e4
  ±  (1.40)   Depth: 22   00:02:21  7835kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qe6+ Kh7 48.Qxb6 Rxb6 49.Bf3 Bd6 50.Ke2 Kg8 51.Ke3 Kf8
  +-  (1.54)   Depth: 23   00:03:27  11638kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qf7+ Kh7 48.Ke2 Rd8 49.Qb7 Rb8 50.Qxb6 Rxb6 51.Bf3 Bd6
  +-  (1.54)   Depth: 24   00:05:49  19396kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qf7+ Kh7 48.Ke2 Rd8 49.Qb7 Rb8 50.Qxb6 Rxb6 51.Bf3 Bd6
  +-  (1.72)   Depth: 25   00:10:07  32473kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qf7+ Kh7 48.Ke2 Rd8 49.Qb7 Rb8 50.Qxb6 Rxb6 51.Bf3 Bd6
  +-  (1.69)   Depth: 26   00:15:07  49378kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qf7+ Kh7 48.Ke2 Rd8 49.Qb7 Rb8 50.Qxb6 Rxb6 51.Bf3 Bd6
  +-  (1.65)   Depth: 27   00:27:58  91315kN
45.Qd7+ Kg8 46.Ra7 Bf8 47.Qf7+ Kh7 48.Ke2 Rd8 49.h4 Kh8 50.Qb7 Rb8 51.Qxb6 Rxb6
  +-  (1.50)   Depth: 28   01:02:15  198mN

(, AMD 28.06.2007)

After 45. Ra6!?, Rybka understand black Queen move is enough strong, although it cannot not see the draw, at least not in my machine with an appropriate time.

1r6/5kp1/RqQb1p1p/1p1PpP2/1Pp1B3/2P4P/6P1/5K2 b - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

45...Qxc6 46.dxc6 Bc7
  ±  (1.30)   Depth: 5   00:00:01
45...Qxc6 46.dxc6 Bc7 47.Ra7
  ±  (1.30)   Depth: 6   00:00:01  1kN
45...Qxc6 46.dxc6 Bc7 47.Ra7 Rc8
  ±  (1.35)   Depth: 7   00:00:01  1kN
45...Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7
  ±  (1.35)   Depth: 8   00:00:01  2kN
45...Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7
  +-  (1.44)   Depth: 9   00:00:01  4kN
45...Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7 49.Ke2 Kd6
  +-  (1.68)   Depth: 10   00:00:02  33kN
45...Qxc6 46.dxc6 Bc7 47.Ra7 Rc8 48.Rb7 Ke7 49.Ke2 Kd6
  +-  (1.51)   Depth: 11   00:00:02  37kN
45...Qxc6 46.dxc6 Re8 47.Ra7+ Re7 48.Ra5 Rc7 49.Rxb5 Ke7 50.Ra5 Rc8 51.Ke2
  +-  (1.88)   Depth: 12   00:00:03  100kN
45...Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Ke7 49.Ke2 Rc8 50.Rxb5 Bc7 51.Rd5
  +-  (1.74)   Depth: 13   00:00:04  138kN
45...Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Ke7 49.Ke2 Rc8 50.Rxb5 Bc7 51.Rd5 Rd8
  +-  (1.74)   Depth: 14   00:00:06  154kN
45...Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Rc8 49.Rxb5 Ke7 50.Ke2 Rb8 51.Ra5 Bc7 52.Rc5
  +-  (1.80)   Depth: 15   00:00:07  199kN
45...Qxc6 46.dxc6 Rc8 47.Ra7+ Rc7 48.Ra5 Rc8 49.Ke2 Ke7 50.Kf3 Bc7 51.Rxb5 Rd8 52.Rc5
  +-  (1.93)   Depth: 16   00:00:09  274kN
45...Qe3 46.Qxd6 Re8 47.h4 h5 48.Bf3 Qc1+ 49.Kf2 Qd2+ 50.Be2 Qf4+ 51.Kg1 Qe3+ 52.Kh2
  ±  (1.35)   Depth: 16   00:00:42  1838kN
45...Qe3 46.Qxd6 Re8 47.h4 h5 48.Bf3 Qc1+ 49.Kf2 Qd2+ 50.Be2 Qf4+ 51.Kg1 Qe3+ 52.Kh2
  ±  (0.86)   Depth: 17   00:00:54  2405kN
45...Qe3 46.Qxd6 Re8 47.h4 h5 48.Bf3 Qc1+ 49.Kf2 Qd2+ 50.Be2 Qf4+ 51.Kg1 Qe3+ 52.Kh2
  ±  (0.90)   Depth: 18   00:01:09  3124kN
45...Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qc6 Qxe4 49.d6 Qxf5+ 50.Kg1 Rd7 51.Qxb5 Qe6 52.Qc5
  ²  (0.60)   Depth: 19   00:02:00  5800kN
45...Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qc6 Qf4+ 49.Bf3 Qc1+ 50.Kf2 Qd2+ 51.Be2 Qf4+ 52.Kg1
  ²  (0.58)   Depth: 20   00:02:52  8737kN
45...Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qc6 Qf4+ 49.Bf3 Qc1+ 50.Kf2 Qd2+ 51.Be2 Qf4+ 52.Kg1
  ²  (0.67)   Depth: 21   00:04:15  13149kN
45...Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qc6 Qf4+ 49.Bf3 Qc1+ 50.Kf2 Qd2+ 51.Be2 Qf4+ 52.Kg1
  ²  (0.67)   Depth: 22   00:06:07  19317kN
45...Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qc6 Qf4+ 49.Bf3 Qc1+ 50.Kf2 Qd2+ 51.Be2 Qf4+ 52.Kg1
  ²  (0.67)   Depth: 23   00:09:13  29470kN
45...Qe3 46.Qxd6 Re8 47.Qd7+ Re7 48.Qc6 Qxe4 49.d6 Qd3+ 50.Kg1 Re8 51.Qd7+ Kf8 52.Kh2
  =  (0.11)   Depth: 24   00:28:22  93265kN

(, AMD 28.06.2007)

45. Qe3 46. Qxd6 Re8 47 h4! h5 48. Bf3 Qc1+ 49. Kf2 Qd2+ 50. Be2 Qf4+ 51. Kg1 Qe3+ 52. Kf1 Qc1+ and of course it is a draw by repetition.

Regards,
Gambito.
Parent - - By Gaмßito (****) Date 2007-06-28 01:15
Rybka 2.3.2a clings to Be4 in my ancient hardware and it doesn't loose it.

r1r1q1k1/6p1/3b1p1p/1p1PpP2/1Pp5/2P4P/R1B2QP1/R5K1 w - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

1.Qb6 Rxa2
  ±  (0.71)   Depth: 5   00:00:00
1.Qb6 Rxa2 2.Rxa2 Bc7
  ²  (0.66)   Depth: 6   00:00:00  3kN
1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7
  ²  (0.60)   Depth: 7   00:00:00  4kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rxa2 4.Rxa2 Kf8 5.Kf2
  ²  (0.54)   Depth: 8   00:00:00  8kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rxa2 4.Rxa2 Kf8 5.Kf1 Ke7
  ²  (0.58)   Depth: 9   00:00:00  20kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rxa2 4.Rxa2 Kf8 5.Kf1 Ke7 6.Ra7
  ²  (0.63)   Depth: 10   00:00:00  26kN
1.Qb6 Bc7 2.Qe6+ Qxe6 3.fxe6 Rxa2 4.Rxa2 Kf8 5.Kf1 Ke7 6.Ra7 Kd6
  ±  (0.76)   Depth: 11   00:00:01  48kN
1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 h5 5.Ra6 Rb8
  ±  (0.84)   Depth: 12   00:00:02  88kN
1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 Rb8 5.Qc6 Qxc6 6.dxc6 Bc7 7.Kh2
  ±  (0.84)   Depth: 13   00:00:03  126kN
1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 Rb8 5.Qc6 Qxc6 6.dxc6 Bc7 7.Kh2 Bb6
  ±  (0.74)   Depth: 14   00:00:05  205kN
1.Qb6 Rxa2 2.Rxa2 Bc7 3.Qb7 Bd6 4.Be4 Rb8 5.Qc6 Qxc6 6.dxc6 Bc7 7.Kh2 Bb6
  ±  (0.79)   Depth: 15   00:00:07  325kN
1.Qb6 Rxa2 2.Rxa2 Ra8 3.Ra5 Rxa5 4.bxa5 Qb8 5.Qxb8+ Bxb8 6.Be4 Kf7 7.Kf2 Ke7
  =  (0.20)   Depth: 16   00:00:20  965kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Kf7 4.Qa7+ Qc7 5.Qe3 Rb8 6.Kg3 Qd7 7.Ra7 Rb7
  ±  (0.73)   Depth: 16   00:01:30  4123kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Kf7 4.Qa7+ Qc7 5.Qe3 Rb8 6.Kg3 Qd7 7.Ra7 Rb7
  ±  (0.73)   Depth: 17   00:01:40  4624kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Bc7 5.Qc5 Rb6 6.Kg3 Bd6 7.Qe3 Rb7
  ±  (0.73)   Depth: 18   00:01:59  5401kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Bc7 5.Qc5 Rb6 6.Kg3 Bd6 7.Qe3 Rb7
  ±  (0.73)   Depth: 19   00:02:24  6427kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Bc7 5.Qc5 Rb6 6.Kg3 Bd6 7.Qe3 Rb7
  ±  (0.73)   Depth: 20   00:03:10  8480kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Bc7 5.Qc5 Rb6 6.Kg3 Bd6 7.Qe3 Rb7
  ±  (0.73)   Depth: 21   00:04:15  11437kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Bc7
  ±  (0.72)   Depth: 22   00:06:29  18618kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Bc7
  ±  (0.72)   Depth: 23   00:10:12  29780kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Bc7
  ±  (0.72)   Depth: 24   00:17:37  51653kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Bc7
  ±  (0.72)   Depth: 25   00:31:00  89485kN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Qd6
  ±  (0.72)   Depth: 26   00:53:19  151mN
1.Be4 Qd8 2.Kh2 Rxa2 3.Qxa2 Rb8 4.Qa7 Rb6 5.Kg3 Bb8 6.Qa5 Kf7 7.Kg4 Qd6
  ±  (0.72)   Depth: 27   01:36:22  278mN

(, AMD 27.06.2007)

Regards,
Gambito.
Parent - By Conwy (*) Date 2007-06-28 06:29
The nolot position:

r4rk1/pp1n1p1p/1nqP2p1/2b1P1B1/4NQ2/1B3P2/PP2K2P/2R5 w - -


Sherdder 10.1:

9/31   0:01   -1.59++  1.Bf6 Qb5+ (1.545.042) 969
  9/31   0:02   -0.25   1.Bf6 Qb5+ 2.Ke1 Be3 3.Qxe3 Nxe5 4.Qe2 Qa5+ 5.Kf2 Rac8 (2.009.319) 973
10/32   0:03   -0.10   1.Bf6 Qb5+ 2.Ke1 Bb4+ 3.Kf2 Nd5 4.Bxd5 (3.097.705) 986
11/37   0:06   -0.01   1.Bf6 Qb5+ 2.Ke1 Qa5+ 3.Kf1 Qa6+ 4.Ke1 Bd4 5.Qh6 (7.035.507) 1016
12/39   0:11   -0.01   1.Bf6 Qb5+ 2.Ke1 Qa5+ 3.Kf1 Qa6+ 4.Ke1 Bd4 5.Qh6 Bg1 (11.625.728) 1030
13/43   0:20   +0.01   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ (21.955.840) 1055
13/46   0:25   +0.02++  1.Rxc5 (26.953.376) 1060
13/46   0:28   +0.37++  1.Rxc5 (29.871.443) 1064
13/46   0:46   +4.73   1.Rxc5 Qxc5 2.Nxc5 Nxc5 3.Bh6 Nxb3 4.Bxf8 Rxf8 5.axb3 Nd7 6.Qe4 Rd8 7.Ke3 a5 8.b4 (51.902.315) 1124
14/44   0:49   +4.98++  1.Rxc5 Qxc5 (55.392.758) 1128
14/44   0:49   +4.98   1.Rxc5 Qxc5 (55.999.683) 1130

UCI_AnalyseMode set to true
  8.00   0:01   -1.79   1.Bh6 Qb5+ 2.Kd2 Bb4+ 3.Ke3 Kh8 4.Rc7 (132.678) 129
  9.00   0:01   -1.79   1.Bh6 Qb5+ 2.Kd2 Bb4+ 3.Ke3 Kh8 4.Rc7 Rfe8 (182.756) 127
10.00   0:03   -1.65   1.Bh6 Qb5+ 2.Kd2 Bb4+ 3.Ke3 Kh8 4.Rc7 Rfc8 5.Qxf7 (455.610) 129
10.00   0:07    0.00   1.Bf6 Qb5+ 2.Ke1 Bb4+ 3.Kf2 Nxf6 4.Nxf6+ Kg7 5.Nh5+ Kg8 6.Nf6+ Kg7 7.Nh5+ Kg8 (893.134) 122
11.00   0:08    0.00   1.Bf6 Qb5+ 2.Ke1 Bb4+ 3.Kf2 Nxf6 4.Nxf6+ Kg7 5.Nh5+ Kg8 6.Nf6+ Kg7 7.Nh5+ Kg8 (1.037.812) 118
12.01   0:13    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (1.530.037) 115
13.01   0:16    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (1.848.750) 112
14.01   0:22    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (2.439.016) 108
15.01   0:33    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (3.490.330) 105
16.01   0:53    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (5.463.246) 103
17.01   1:28    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (8.853.313) 102
18.01   2:36    0.00   1.Bf6 Qb5+ 2.Ke1 Nxf6 3.Nxf6+ Kg7 4.Nh5+ gxh5 5.Qg5+ Kh8 6.Qf6+ Kg8 7.Qg5+ Kh8 (15.576.239) 102

Not yet. it seems..

Conwy
Parent - By Vasik Rajlich (Silver) Date 2007-06-29 13:52
The critical position is one move earlier, 36. axb5 instead of 36. Qb6. 36. Qb6 wins the house and the only reason to reject it is because of a very speculative and desperate-looking counterattack. (I forget the moves now.)

I suspect Deep Blue had some bug in its evaluation function which caused it to reject 36. Qb6.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka analysing the famous Kasparov-Deep Blue position

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill