A Fascinating Game

[Harvey Williamson]

Hi All,

This game was just played at 60+15 on Playchess. The Hiarcs beta was on a Qx6700 quadcore at 2.66. Rybka was on a Qx6700 quadcore overclocked to 3.2. This game is not fascinating just because Hiarcs won but because of the difference in evaluations, particularly from about move 28. Hiarcs was always thinking it was better by around +1 while Rybka was saying =. Play through the moves it is very interesting.

Best Wishes,

Harvey

[Event "60m + 15s, rated"]
[Site "?"]
[Date "2007.08.10"]
[Round "?"]
[White "HiarcsX MP, HIARCS 11.51e MP"]
[Black "Fredis_cpu, Rybka 2.3.2a mp"]
[Result "1-0"]
[WhiteElo "2550"]
[BlackElo "2710"]
[PlyCount "203"]
[EventDate "2007.08.10"]
[EventType "rapid"]

1. e4 {0} c5 {0.00/0 0} 2. Nf3 {0} d6 {0.00/0 0} 3. d4 {0} cxd4 {0.00/0 0} 4.
Nxd4 {0} Nf6 {0.00/0 0} 5. Nc3 {0} a6 {0.00/0 0} 6. Be3 {0} e5 {0.00/0 0} 7.
Nb3 {0.01/1 0} Be6 {0.00/0 0} 8. Qd2 {0.01/1 0} Be7 {0.00/0 0} 9. f3 {0.01/1 0}
O-O {0.00/0 0} 10. O-O-O {0.01/1 0} Nbd7 {0.00/0 0} 11. g4 {0.01/1 0} b5 {
0.00/0 0} 12. g5 {0.01/1 0} b4 {0.00/0 0} 13. Ne2 {0.01/1 0} Ne8 {0.00/0 0} 14.
f4 {0.01/1 0} a5 {0.00/0 0} 15. f5 {0.01/1 0} a4 {0.00/0 0} 16. Nbd4 {0.01/1 0}
exd4 {0.00/0 0} 17. Nxd4 {0.01/1 0} b3 {0.00/0 0} 18. Kb1 {0.01/1 0} bxc2+ {
0.00/0 0} 19. Nxc2 {0.01/1 0} Bb3 {0.00/0 0} 20. axb3 {0.01/1 0} axb3 {0.00/0 0
} 21. Na3 {0.01/1 0} Ne5 {0.00/0 0} 22. h4 {0.01/1 0} Ra4 {0.00/0 0} 23. Bd4 {
0.01/1 0} Qa8 {0.00/0 0} 24. Qe3 {0.01/1 0} Kh8 {0.00/0 0} 25. Rg1 {0.48/16 223
} f6 {0.22/18 163} 26. Bb5 {(Le2) 0.46/16 0} Rb4 {0.01/17 30} 27. Be2 {
0.43/17 165} Nc7 {(Bd8) 0.00/18 106} 28. Bxe5 {(Tg3) 0.46/15 137} Rxe4 {
(Qxe4+) 0.00/16 29} 29. gxf6 {1.15/17 105} Rxe3 {0.02/21 3} 30. fxg7+ {
1.05/18 89} Kg8 {0.02/22 0} 31. gxf8=Q+ {1.06/18 146} Kxf8 {0.00/24 0} 32. Bg7+
{0.92/19 178} Ke8 {0.00/24 175} 33. Bh5+ {1.06/19 0} Kd7 {0.00/24 161} 34. f6 {
1.06/19 0} Qe4+ {0.00/23 197} 35. Ka1 {0.98/20 0} Qxh4 {0.00/23 459} 36. Bg4+ {
1.09/20 0} Ne6 {(Ke8) 0.00/23 633} 37. fxe7 {0.92/17 123} Kxe7 {0.08/23 163}
38. Bf5 {1.11/19 0} h5 {0.00/22 388} 39. Bh6 {1.16/19 0} Re5 {0.19/22 377} 40.
Rgf1 {1.50/20 0} Qa4 {0.21/21 101} 41. Bg6 {(Ld2) 1.64/19 0} Qg4 {
(Qb4) 0.00/20 33} 42. Rf7+ {1.93/19 129} Ke8 {0.18/21 53} 43. Rg7+ {1.95/19 77}
Kd8 {0.24/22 39} 44. Rxd6+ {1.96/19 80} Kc8 {0.24/23 0} 45. Rc6+ {2.05/19 125}
Kd8 {0.35/23 136} 46. Rg8+ {2.31/18 0} Ke7 {0.26/22 33} 47. Ra6 {2.34/17 77}
Re2 {0.22/21 102} 48. Ra7+ {2.09/16 25} Kd6 {0.19/21 96} 49. Bh7 {
(Lf8+) 2.08/16 0} Re1+ {0.12/15 11} 50. Nb1 {2.08/17 63} Qd4 {0.35/19 46} 51.
Ra6+ {2.18/16 65} Kc7 {0.40/20 41} 52. Re8 {2.18/17 32} Nc5 {0.55/20 122} 53.
Ra7+ {2.49/19 0} Kb6 {0.62/20 7} 54. Rxe1 {2.35/21 108} Kxa7 {0.58/24 12} 55.
Be3 {2.67/20 78} Qe5 {0.58/24 6} 56. Bf2 {2.64/21 76} Qg5 {0.51/23 42} 57. Nc3
{(Te4) 2.67/20 44} Kb7 {0.38/17 8} 58. Bb1 {(Se4) 2.67/19 61} Kc8 {
(Kc6) 0.46/18 38} 59. Rg1 {(Ld4) 2.71/18 79} Qe5 {(Qe7) 0.51/15 11} 60. Rg8+ {
2.95/18 113} Kb7 {0.32/21 22} 61. Rg6 {(Sb5) 3.03/20 52} Ne6 {0.37/20 24} 62.
Be4+ {2.96/19 48} Kc8 {0.29/20 12} 63. Rg8+ {(Ld5) 2.41/17 89} Kd7 {0.26/19 0}
64. Rg1 {(Tg3) 2.42/18 92} Qf6 {(Qf4) 0.17/16 12} 65. Rd1+ {2.58/17 90} Ke8 {
(Kc8) 0.25/21 29} 66. Be3 {(Lg1) 2.41/16 63} h4 {0.01/15 10} 67. Bc6+ {
(Sd5) 2.45/16 42} Kf7 {(Kf8) 0.06/19 43} 68. Bd5 {(Lb5) 2.52/17 84} h3 {
0.11/17 15} 69. Bd4 {(Lxb3) 2.53/17 91} Qf5 {(Qf4) 0.10/19 24} 70. Bg1 {
(Te1) 2.60/18 78} Kg7 {(Qe5) 0.15/18 31} 71. Bxb3 {2.53/18 90} Ng5 {0.16/20 49}
72. Bd5 {(Lh2) 2.50/18 0} Kg6 {(Nf3) 0.29/19 132} 73. b4 {2.91/17 60} h2 {
(Qg4) 0.31/21 33} 74. Bxh2 {3.21/17 48} Qf8 {0.31/21 5} 75. Bb3 {
(Kb2) 3.37/17 65} Qxb4 {0.23/15 6} 76. Kb2 {(Td6+) 3.20/17 83} Nf3 {0.23/18 27}
77. Nd5 {(Lg1) 3.21/17 39} Qb5 {(Qc5) 0.19/15 28} 78. Bc7 {(Ld6) 3.21/15 43}
Kh5 {(Kg5) 0.21/17 18} 79. Nf4+ {(Sc3) 3.24/16 72} Kh4 {(Kg4) 0.25/18 44} 80.
Rd8 {(Td5) 3.28/16 54} Qb4 {0.23/15 7} 81. Ne6 {3.38/15 30} Kh3 {
(Qe4) 0.27/17 10} 82. Rg8 {3.40/17 90} Kh4 {0.32/20 6} 83. Bg3+ {3.53/17 50}
Kh5 {0.31/21 4} 84. Bf4 {3.53/18 51} Qe4 {(Kh4) 0.31/21 32} 85. Rg2 {3.57/17 51
} Qb4 {0.31/21 13} 86. Be3 {(Lg5) 3.85/17 39} Qb8 {(Qb5) 0.33/16 33} 87. Nf4+ {
4.13/17 52} Kh6 {0.37/20 18} 88. Rg6+ {3.89/17 40} Kh7 {0.38/22 0} 89. Re6 {
4.10/17 37} Qh8+ {0.46/20 2} 90. Kb1 {4.59/17 38} Qb8 {0.47/20 20} 91. Rb6 {
(Kc2) 4.84/17 27} Qe5 {0.44/17 10} 92. Bc1 {4.84/17 37} Qa5 {0.78/19 13} 93.
Rb7+ {4.53/15 30} Kh6 {0.85/19 23} 94. Bd5 {4.78/15 0} Nd2+ {0.85/19 45} 95.
Kc2 {6.65/16 0} Nf1 {1.03/19 90} 96. Nd3+ {(Se2+) 6.69/15 0} Kg6 {0.88/17 22}
97. Bf7+ {6.90/16 0} Kf5 {(Kf6) 1.10/18 31} 98. Bb3 {(Tb4) 7.36/15 35} Kg6 {
1.30/16 19} 99. Rf7 {7.60/15 27} Nh2 {6.55/18 5} 100. Bd2 {7.78/15 41} Qb5 {
7.24/18 0} 101. Nf4+ {7.98/15 34} Kg5 {7.39/17 0} 102. Ng2+ {
Fredis_cpu,Rybka 2.3.2a mp gibt auf (Lag: Av=0.27s, max=2.8s) 8.42/15 33} 1-0

[mackgra]

Hi Harvey,

Yes, very interesting indeed! Looks like larry Kaufman will have to rejig a few of those rybka piece values for the next version!

Piece imbalances as you may know is a personal hobby horse of mine and playing through the game brought the following thoughts:

1) Hiarcs initially played a king side attack which we know it excels at over all chess engines

2) The two bishops plus a rook vs queen and two pawns for rybka in my mind favours Hiarcs by at least a pawn. There is absolutely no piece redundancy with white pieces (no duplication of role) with the three pieces coordinating wonderfully together. In contrast the two extra black pawns weren't advanced at all.

3) The two bishops are paticularly dominant in the endgame anyway, especially with pawns on both wings. This alone can account for an extra pawn, (which hiarcs did pick up later on).

4) Even quite late on on the endgame Rybka did not think it was in trouble but anyone can see that 4 white pieces (2 bishops, a knight and a rook) easily win through sheer numbers against a queen and a knight.


Hope these comments will prove useful!

keep up the good work,

Graham

[Mark Uniacke]

Interesting game on powerful hardware, thanks Harvey. Especially nice to see Hiarcs overcoming Rybka on faster quadcore hardware.

I agree with Graham who is clearly knows his stuff in these situations

A beautifully played game from Hiarcs, Rybka really looked all at sea in this game with big difference of eval opinions between the programs.

The difference of opinion really began at move 28/29

[fen]q4r1k/2n1b1pp/3p1p2/4BPP1/4r2P/Np2Q3/1P2B3/1K1R2R1 w - - 0 29[/fen]

29. gxf6 is the beginning of a long combination to a very unbalanced material situation of Rook & two Bishops against Queen and two pawns. At move 29 the eval difference is more than a pawn with Hiarcs thinking white is ahead 1.15 and Rybka thinking it is level.

By move 40 the dust has settled and the unbalanced material is clear:

[fen]8/4k3/3pn2B/4rB1p/7q/Np6/1P6/K2R2R1 w - - 0 40[/fen]

It surprises me that Rybka is so calm about the dangers of this position for black. The bishop pair in this position are very strong.

Later we reach an unbalanced position without pawns:
[fen]8/8/6k1/6n1/1q6/1BN5/7B/K2R4 w - - 0 76[/fen]

Now the databases have R+2B v Q as predominantly a draw although there are wins for the R & 2Bs. I don't know if it is the extra knights on the board or the active approach used by Hiarcs but the end comes quite quickly for black here!

I think it is very dangerous of Rybka to be so complacent about this material imbalance especially in the open position. Many chess players might be misled by its evals

[fen]8/5R2/8/1q4k1/5N2/1B6/2KB3n/8 w - - 0 102[/fen]

The power of the bishop pair in an open position in this game was very telling

Best wishes,
Mark

[Ted Summers]

Harvey,

That was one hell of a brilliant game. I will have to put that in game my Hiarcs greatest games collection.

Thanks Much!!

[Sebastian Boehme]

Wow guys what an ending!

Isn't that sweet how Hiarcs outplayed the smelly fish? Just: Wow!

Seeing this endgame as human (especially an endgame with no pawns), my humble mind tells me this is a draw somehow. Good the engine proved me wrong.

Best regards,
Sebastian

[BigBen]

Hi,
When do we get our hands on this Beta 11.51 thingy

Regards

[mclane]

one game alone makes no summer.

i would be pleased more to get 11.2 into my palm or my mobile.

[Steve B]

one game alone makes no summer.

i.

i agree

it is time for the two great Houses to join in battle

House Hiarcs vs House Rybka

On Playchess

The Spice ..Must Flow

Regards From Dune
Steve

[Master HIARCS]

HI!

GREATFULL!

HIARCS is simply the BEST


Wish all the best

[Peter Grayson]

An excellent win for Hiarcs. Rybka is heavily focused on and very good at protecting and advancing pawns where other engines fail to see their potential, but here Rybka needs awareness of King exposure and Hiarcs shows Rybka is sadly lacking in this department.

This knowledge gap works both ways ..... try playing through as White with Rybka and this becomes very clear when after 45.Rc6+ .. Kd8, Rybka's analysis
46.Rd6+ Kc8
= (0.00) Depth: 20 00:00:22 7275kN, tb=1
not only shows zero but the move continuation is very short where it is expecting a repetition of checks for the draw!

This game further confirms that when Rybka is taken out of its comfort zone it can fall apart quite quickly and be made to look quite ordinary. However, knowing it is one thing, getting there is a lot harder!

PeterG