Finally a position programs do not have a clue

[Uri Blass]

The scores at small depths suggest that the hash is not clear because rybka is unable to see less than one pawn advantage at small depth with clear hash when your analysis show less than 1 pawn for black even at depth 6.

This is indeed strange. My analysis above was with 1 mb hash having loaded in the whole game and set white to find move 19 using infinite anlysis. If i just load the fen then it does not find Nd1 in a reasonable time. I reloaded the game and this time used 128 mb for hash and again it found Nd1 fairly quickly. So it seems that this is one of those positions that depends on whether you load the fen or the whole game maybe.

I do not agree.

my analysis was from game position but with clear hash that means without giving rybka to analyze different positions from the same game earlier.

Uri

[Omega-three]

I

do not agree.

my analysis was from game position but with clear hash that means without giving rybka to analyze different positions from the same game earlier.

Uri

This is what I did step by step.

1. Load the GUI
2. Paste game (having previously copied it from here)
3. Clicked move 18 (Rf7)
4. Clicked infinite analysis

This what I got with 128 mb hash :-

New game - Rybka 3 1-cpu 32-bit, 80'/40+40'/20+20'
r5k1/pp2prbp/1q1p2p1/n1pPn3/2P1B3/PPN3P1/1BQN1P2/5RK1 w - - 0 1

Analysis by Deep Rybka 3:

19.Na4 Qc7 20.Bxe5
-/+ (-1.40) Depth: 6 00:00:00 4kN
19.Na4 Qc7 20.Bxe5 Bxe5 21.b4
-/+ (-1.39) Depth: 7 00:00:00 5kN
19.Na4 Qc7 20.Bxe5 Bxe5 21.f4 Bd4+ 22.Kg2
-+ (-1.53) Depth: 8 00:00:00 15kN
19.Na4 Qc7 20.Bxe5 Bxe5 21.f4 Bd4+ 22.Kg2 Qd7 23.b4
-+ (-1.57) Depth: 9 00:00:00 26kN
19.Na4 Qc7 20.Bxe5 Bxe5 21.f4 Bd4+ 22.Kg2 Qd7 23.b4 cxb4 24.axb4 Be3 25.bxa5 Bxd2 26.Nc5 dxc5 27.Qxd2
-+ (-1.52) Depth: 10 00:00:01 68kN
19.Na4 Qc7 20.Bxe5 Bxe5 21.f4 Bd4+ 22.Kg2 a6 23.Nf3 Bf6 24.Nc3 Rb8
-+ (-1.50) Depth: 11 00:00:02 194kN
19.Nd1
-/+ (-1.32 !) Depth: 11 00:00:04 293kN
19.Nd1
-/+ (-1.12 !) Depth: 11 00:00:07 498kN
19.Nd1 a6 20.Ne3 Rc8 21.Re1
-/+ (-1.13) Depth: 11 00:00:09 617kN
19.Nd1
-/+ (-0.93 !) Depth: 12 00:00:17 1099kN
19.Nd1 Raf8 20.Ne3
-/+ (-0.90) Depth: 12 00:00:21 1280kN
19.Nd1 Bh6 20.f4 Nd7
-/+ (-0.89) Depth: 13 00:00:35 2166kN
19.Nd1 Bh6 20.f4 Nd7 21.Nf2 Nf6 22.Bf3 Bg7 23.Bc3 Rc8 24.Nh3 Bh6 25.Nf2
-/+ (-0.80) Depth: 14 00:00:58 3562kN
19.Nd1 Bh6 20.f4[] Nd7 21.Nf2 Nf6 22.Bf3 Raf8 23.Bc3 Rc8 24.Re1 Bg7 25.Kg2 Qc7 26.Nh3 Bh6 27.Nf2 Rcf8
-/+ (-0.71) Depth: 15 00:01:43 6352kN
19.Nd1
=/+ (-0.51 !) Depth: 16 00:05:25 22530kN

Above was with 128 mb hash.

[Omega-three]

Hi Uri,

I think I may have an answer for the discrepancy. I repeated the above but this time set Rybka to only use 1 core and this is what I got :-

New game - Rybka 3 1-cpu 32-bit, 80'/40+40'/20+20'
r5k1/pp2prbp/1q1p2p1/n1pPn3/2P1B3/PPN3P1/1BQN1P2/5RK1 w - - 0 1

Analysis by Deep Rybka 3: 1 core only

19.Na4 Qc7 20.Bxe5 Bxe5
-+ (-1.44) Depth: 6 00:00:00 1kN
19.Re1 Raf8
-/+ (-1.36) Depth: 9 00:00:00 5kN
19.Bg2
-/+ (-1.16 !) Depth: 9 00:00:00 11kN
19.Bg2 Raf8 20.Nce4 h6 21.Nc3 Nxb3 22.Nxb3 Nxc4
-/+ (-1.14) Depth: 9 00:00:00 18kN
19.Bg2 Rd8 20.Nce4 h6 21.Nc3 g5 22.Na4 Qc7
-/+ (-1.09) Depth: 10 00:00:02 90kN
19.Bg2 Rd8 20.Nce4 h6 21.Nc3 g5 22.Na4 Qc7 23.Bxe5
-/+ (-1.09) Depth: 11 00:00:02 107kN
19.Bg2 Rd8 20.Nce4 h6 21.Nc3 g5 22.Na4 Qc7 23.Bxe5
-/+ (-1.09) Depth: 12 00:00:03 145kN
19.Bg2 Rb8 20.Na4 Qd8
-/+ (-1.18) Depth: 13 00:00:15 665kN
19.Bg2 Nxb3 20.Nxb3[] Nxc4 21.Bc1
-/+ (-1.08) Depth: 14 00:00:26 1140kN
19.Bg2 Nxb3 20.Nxb3[] Nxc4 21.Bc1
-/+ (-1.08) Depth: 15 00:00:31 1379kN
19.Bg2 Nxb3 20.Nxb3[] Nxc4[] 21.Bc1 Qa6 22.Nd2 Nxa3 23.Qb3 Bxc3 24.Qxc3 Nb5[] 25.Qe3 Nd4 26.Nf3 Nxf3+ 27.Bxf3 Qc4 28.Bb2 Raf8 29.Rc1 Qb4 30.Bc3
-/+ (-1.21) Depth: 16 00:01:19 3669kN
19.Bg2 Nxb3 20.Nxb3[] Nxc4[] 21.Bc1 Qa6 22.Nd2 Nxa3 23.Qb3 Bxc3 24.Qxc3 Nb5[] 25.Qe3 Nd4 26.Nf3 Nxf3+ 27.Bxf3 Qc4 28.Bb2 Raf8 29.Rc1 Qb4 30.Bc3
-/+ (-1.21) Depth: 17 00:03:07 8945kN
19.f4
-/+ (-1.01 !) Depth: 17 00:05:40 15742kN