diep's inferior evaluation

[Uri Blass]

[Event "15th World Computer Chess Championship"]
[Site "Amsterdam, The Netherlands"]
[Date "2007.06.11"]
[Round "1"]
[White "Diep"]
[Black "Jonny"]
[Result "0-1"]
[PlyCount "150"]
[EventDate "2007.??.??"]

1. e4 {-00:00.001} e5 {-00:00.001} 2. Nf3 {-00:00.001} Nc6 {-00:00.001} 3. Bb5
{-00:00.001} a6 {-00:00.001} 4. Ba4 {-00:00.001} Nf6 {-00:00.001} 5. O-O {
-00:00.001} Be7 {-00:00.001} 6. Re1 {-00:00.001} b5 {-00:00.001} 7. Bb3 {
-00:00.001} O-O {-00:00.001} 8. c3 {-00:00.001} d6 {-00:00.001} 9. h3 {
-00:00.001} Na5 {-00:00.001} 10. Bc2 {-00:00.001} c5 {-00:00.001} 11. d4 {
-00:00.001} Qc7 {-00:00.001} 12. Nbd2 {-00:00.001} cxd4 {-00:00.001} 13. cxd4 {
-00:00.001} Bb7 {-00:00.001} 14. Nf1 {-00:00.001} Rac8 {-00:00.001} 15. Re2 {
-00:00.001} Nc6 {-00:00.001} 16. Ng3 {-00:00.001} exd4 {-00:00.001} 17. Nxd4 {
-00:00.001} d5 {-00:00.001} 18. Nxc6 {-00:00.001} Qxc6 {-00:00.001} 19. e5 {
-00:00.001} d4 {-00:00.001} 20. f3 {-00:00.001} Nd5 {-00:00.001} 21. Kh1 {
-00:00.001} Qc5 {-00:00.001} 22. Nf5 {-00:00.001} Nb4 {-00:00.001} 23. Bf4 {
-00:00.001} g6 {-00:00.001} 24. a3 {-00:00.001} Nd5 {-00:00.001} 25. Nxe7+ {
-00:00.001} Qxe7 {-00:00.001} 26. Bh6 {-00:00.001} Rfd8 {-00:00.001} 27. Bd3 {
-00:00.001} Nc7 {-00:00.001} 28. a4 {-00:00.001} bxa4 {-00:00.001} 29. Rxa4 {
-00:00.001} Ne6 {-00:00.001} 30. Qe1 {-00:00.001} Nc5 {-00:00.001} 31. Ra3 {
-00:00.001} Bd5 {-00:00.001} 32. Bf5 {-00:00.001} Rc6 {-00:00.001} 33. Qg3 {
-00:00.001} Re8 {-00:00.001} 34. b4 {-00:00.001} Be6 {-00:00.001} 35. Bc2 {
-00:00.001} Nd7 {-00:00.001} 36. Qf2 {-00:00.001} Rec8 {-00:00.001} 37. Ba4 {
-00:00.001} Rc3 {-00:00.001} 38. Rxc3 {-00:00.001} dxc3 {-00:00.001} 39. Qd4 {
-00:00.001} Nf8 {-00:00.001} 40. Bc2 {-00:00.001} Rc4 {-00:00.001} 41. Qe3 {
-00:00.001} Qd8 {-00:00.001} 42. Kh2 {-00:00.001} Rc8 {-00:00.001} 43. f4 {
-00:00.001} Qd7 {-00:00.001} 44. Bd3 {-00:00.001} Bc4 {-00:00.001} 45. Bxc4 {
-00:00.001} Rxc4 {-00:00.001} 46. Rc2 {-00:00.001} Qc7 {-00:00.001} 47. Bxf8 {
-00:00.001} Kxf8 {-00:00.001} 48. g3 {-00:00.001} Qc8 {-00:00.001} 49. g4 {
-00:00.001} Ke8 {-00:00.001} 50. Qd3 {-00:00.001} Qc7 {-00:00.001} 51. Qe3 {
-00:00.001} Ke7 {-00:00.001} 52. g5 {-00:00.001} Kf8 {-00:00.001} 53. Kg3 {
-00:00.001} Kg7 {-00:00.001} 54. Kh2 {-00:00.001} Kh8 {-00:00.001} 55. Kg3 {
-00:00.001} Kg8 {-00:00.001} 56. Qd3 {-00:00.001} Kg7 {-00:00.001} 57. Qe3 {
-00:00.001} h6 {-00:00.001} 58. gxh6+ {-00:00.001} Kxh6 {-00:00.001} 59. f5+ {
-00:00.001} Kg7 {-00:00.001} 60. f6+ {-00:00.001} Kh7 {-00:00.001} 61. Kh2 {
-00:00.001} Qc6 {-00:00.001} 62. Kg3 {-00:00.001} Qc8 {-00:00.001} 63. Qd3 {
-00:00.001} Qe6 {-00:00.001} 64. Qe2 {-00:00.001} Qd5 {-00:00.001} 65. Qe3 {
-00:00.001} Rd4 {-00:00.001} 66. Rxc3 {-00:00.001} Re4 {-00:00.001} 67. Qc5 {
-00:00.001} Qd2 {-00:00.001} 68. Rf3 {-00:00.001} Qg5+ {-00:00.001} 69. Kf2 {
-00:00.001} Qh4+ {-00:00.001} 70. Kg1 {-00:00.001} Re1+ {-00:00.001} 71. Kg2 {
-00:00.001} Re2+ {-00:00.001} 72. Rf2 {-00:00.001} Qe4+ {-00:00.001} 73. Kg1 {
-00:00.001} Re1+ {-00:00.001} 74. Rf1 {-00:00.001} Re3 {-00:00.001} 75. Qxe3 {
-00:00.001} Qxe3+ {-00:00.001 adjudicate 0-1} 0-1


I read about this game the following
in
http://www.grappa.univ-lille3.fr/icga/n ... .php?id=17

"He could not reach a sufficiently deep level of search in the tree and had to see that his opponent easily made 18 plies. His move 44. Bd3 changed the position from an equal position into a lost position."

I find that this move is result of a poor evaluation and not result of search that is not deep enough.

Toga easily avoid this move at depth 9 or bigger.

1: Diep - Jonny, 15th World Computer Chess Championship 2007
[D]2r2nk1/3q1p1p/p3b1pB/4P3/1P3P2/2p1Q2P/2B1R1PK/8 w - - 0 1

Analysis by Toga II 1.2.1a:

44.Bh6xf8 Kg8xf8
² (0.30) Depth: 1/3 00:00:00
44.Bc2-b1
² (0.34) Depth: 1/3 00:00:00
44.Bc2-d3
² (0.62) Depth: 1/6 00:00:00
44.Bc2-e4
± (0.77) Depth: 1/6 00:00:00
44.Qe3-g3
± (1.01) Depth: 1/7 00:00:00
44.Qe3-g3 Be6-d5
± (0.77) Depth: 2/9 00:00:00
44.Qe3-g3 Qd7-b5 45.Re2-e1 Qb5xb4
± (0.71) Depth: 3/9 00:00:00
44.Qe3-g3 Be6-f5 45.e5-e6 f7xe6 46.Bc2-b3
= (0.24) Depth: 4/13 00:00:00
44.Qe3-g3 Be6-f5 45.Re2-f2 Bf5xc2 46.Rf2xc2 Qd7-d4
² (0.42) Depth: 5/13 00:00:00
44.Qe3-g3 Be6-f5 45.Bh6-g5 Rc8-c4 46.Bg5-f6 Rc4xb4
² (0.33) Depth: 6/15 00:00:00 22kN
44.Qe3-g3 Be6-f5 45.Bh6-g5 Rc8-c4 46.Bg5-f6 Bf5xc2 47.Re2xc2 Nf8-e6
= (0.18) Depth: 7/15 00:00:00 43kN
44.Bc2-d3 Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Nf8-e6 47.Rc2xc3 Rc4xb4
² (0.26) Depth: 7/18 00:00:01 69kN
44.Bc2-d3 Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c7 47.Qe3-g3 Nf8-d7 48.Qg3-d3 Rc4-c6
= (0.23) Depth: 8/18 00:00:01 134kN
44.Bc2-d3 Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-e1 Kf8-g7 49.Qe1-e3 Kg7-f8
= (0.00) Depth: 9/23 00:00:01 300kN
44.Bh6-g5 Be6-d5 45.f4-f5 Qd7-b7 46.Bc2-d3 Qb7xb4 47.Bd3xa6 Rc8-a8 48.Ba6-d3
= (0.19) Depth: 9/23 00:00:01 514kN
44.Bh6-g5 Qd7-b7 45.Qe3-d4 Be6-c4 46.Bc2-e4 Qb7xb4 47.Re2-c2 Nf8-e6 48.Qd4-d7 Rc8-b8
= (0.14) Depth: 10/26 00:00:02 921kN
44.Bh6-g5 Qd7-b7 45.Qe3-d4 Be6-c4 46.Re2-f2 Qb7xb4 47.Rf2-f3 Nf8-e6 48.Qd4-d7 Rc8-b8 49.Bg5-e7 Qb4-b2
= (0.20) Depth: 11/27 00:00:03 1592kN
44.Bh6-g5 Qd7-b7 45.Bg5-h6 Qb7xb4 46.f4-f5 Be6-d5 47.Qe3-g3 Qb4-b8 48.Bh6-f4 Qb8-b5 49.Bc2-d3 Qb5-c5 50.Bf4-e3
= (0.00) Depth: 12/36 00:00:06 4428kN
44.Re2-f2 Be6-f5 45.g2-g4 Bf5xc2 46.Rf2xc2 Qd7-c6 47.f4-f5 Qc6-c4 48.Bh6xf8 Qc4-b3 49.Qe3-e4 Rc8xf8 50.Qe4-d3 Rf8-c8
= (0.06) Depth: 12/36 00:00:09 6317kN
44.Re2-f2 Be6-f5 45.g2-g4 Bf5xc2 46.Rf2xc2 Qd7-c6 47.f4-f5 Qc6-c4 48.Bh6xf8 Qc4-b3 49.Qe3-e4 Rc8-c4 50.Qe4-d3 Kg8xf8 51.Qd3-d6+ Kf8-g8
= (0.09) Depth: 13/36 00:00:11 8237kN
44.Re2-f2 Be6-f5 45.g2-g4 Bf5xc2 46.Rf2xc2 Qd7-c6 47.f4-f5 Qc6-c4 48.Bh6xf8 Kg8xf8 49.Qe3-h6+ Kf8-e7 50.e5-e6 g6xf5 51.g4xf5 f7xe6 52.Qh6xh7+ Ke7-e8 53.f5xe6 Qc4xe6 54.Qh7-h8+ Ke8-d7 55.Qh8-g7+ Qe6-e7 56.Qg7-g4+ Qe7-e6 57.Qg4-g7+
= (0.00) Depth: 14/41 00:00:19 14439kN
44.g2-g4 Qd7-b5 45.f4-f5 Be6-d5 46.Bc2-d3 Qb5xb4 47.Bd3xa6 Rc8-d8 48.Bh6-g5 Rd8-a8 49.Qe3-d3 Ra8xa6 50.Qd3xd5 g6xf5 51.g4xf5 Ra6-a5 52.Qd5-c6
= (0.13) Depth: 14/49 00:00:30 22676kN
44.g2-g4 Qd7-b5 45.f4-f5 Be6-d5 46.Bc2-d3 Qb5xb4 47.Bd3xa6 Rc8-d8 48.Re2-c2 Rd8-a8 49.Ba6-e2 Ra8-c8 50.Be2-f3 Qb4-b3 51.Bf3-e4 Bd5xe4 52.Qe3xe4 Nf8-d7
= (0.09) Depth: 15/49 00:00:46 34933kN
44.g2-g4 Qd7-b5 45.f4-f5 Be6-d5 46.Bc2-d3 Qb5xb4 47.Bd3xa6 Rc8-d8 48.Ba6-d3 Rd8-a8 49.Re2-c2 Nf8-d7 50.e5-e6 f7xe6 51.f5xe6 Nd7-f6 52.Bh6-g5 Qb4-d6+ 53.Bg5-f4 Qd6xe6 54.Qe3xe6+ Bd5xe6 55.Rc2xc3
² (0.41) Depth: 16/49 00:01:48 82633kN
44.g2-g4 Qd7-b5 45.f4-f5 Be6-d5 46.Bc2-d3 Qb5xb4 47.Bd3xa6 Rc8-c6 48.Ba6-d3 Qb4-b6 49.Qe3-g5 f7-f6 50.e5xf6 Rc6xf6 51.Qg5-f4 g6xf5 52.Qf4-g5+ Rf6-g6 53.Qg5xf5 Rg6xh6 54.Qf5-g5+ Kg8-h8 55.Qg5xd5 Qb6-d6+ 56.Qd5-e5+ Qd6xe5+ 57.Re2xe5
² (0.35) Depth: 17/49 00:02:58 135939kN

(, 12.06.2007)

After Bd3 it is easy to see positional advantage for black assuming that you have good evaluation.

1: Diep - Jonny, 15th World Computer Chess Championship 2007
2r2nk1/3q1p1p/p3b1pB/4P3/1P3P2/2pBQ2P/4R1PK/8 b - - 0 1

Analysis by Toga II 1.2.1a:

44...c3-c2 45.Re2xc2 Rc8xc2 46.Bd3xc2
+- (1.94) Depth: 1/7 00:00:00
44...a6-a5 45.b4xa5
+- (1.54) Depth: 1/7 00:00:00
44...Be6-b3 45.Bd3xa6
± (0.85) Depth: 1/7 00:00:00
44...Be6-d5
² (0.38) Depth: 1/7 00:00:00
44...Be6-d5 45.Qe3-g3
² (0.56) Depth: 2/8 00:00:00
44...Be6-d5 45.Bd3xa6 Rc8-c7
² (0.66) Depth: 3/8 00:00:00
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Qe3-d3
= (0.16) Depth: 3/10 00:00:00
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Nf8-e6 47.Rc2xc3 Rc4xb4
² (0.26) Depth: 4/13 00:00:00
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Nf8-e6 47.Rc2xc3 Rc4xb4
² (0.26) Depth: 5/13 00:00:00
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Nf8-e6 47.Rc2xc3 Rc4xb4
² (0.26) Depth: 6/15 00:00:00 14kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c7 47.Qe3-g3 Nf8-d7 48.Qg3-d3 Rc4-c6
= (0.23) Depth: 7/19 00:00:00 51kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-c1 Kf8-g7 49.Qc1-e3 Kg7-f8
= (0.00) Depth: 8/21 00:00:01 118kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-c1 Kf8-g7 49.Qc1-e3 Qc6-c7 50.Kh2-g1
= (-0.14) Depth: 9/25 00:00:01 209kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-d3 Kf8-g7 49.Qd3-e3 Kg7-f8
= (0.00) Depth: 10/25 00:00:01 367kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-d3 h7-h5 49.Qd3-e3 h5-h4 50.Qe3-d3 Kf8-g7 51.Qd3-e3 Qc6-c7
= (-0.06) Depth: 11/30 00:00:01 924kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-d3 h7-h5 49.e5-e6 Qc6-c7 50.Kh2-g1 f7xe6 51.Qd3xg6 Rc4xb4 52.Qg6-f6+ Qc7-f7 53.Qf6-d8+ Kf8-g7 54.Qd8-g5+ Kg7-h7 55.Rc2xc3 Rb4xf4
= (-0.19) Depth: 12/33 00:00:02 2158kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-d3 Kf8-g7 49.Qd3-e3 Qc6-e4 50.Qe3xe4 Rc4xe4 51.Kh2-g3 Re4-c4 52.Kg3-f3 Rc4-c8 53.Kf3-e4
= (-0.09) Depth: 13/37 00:00:04 3602kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Qe3-d3 Kf8-g7 49.Qd3-e3 h7-h5 50.Qe3-g3 h5-h4 51.Qg3xh4 Rc4xb4 52.Qh4-g3 Rb4-b3 53.f4-f5
= (-0.20) Depth: 14/39 00:00:11 9036kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.g2-g4 Qc6-e4 49.Qe3xe4 Rc4xe4 50.Rc2xc3 Re4xf4 51.Rc3-a3 Rf4xb4 52.Ra3xa6 Rb4-b5 53.e5-e6 f7xe6 54.Ra6xe6 Kf8-f7 55.Re6-c6 Rb5-b2+ 56.Kh2-g3 Rb2-b3+ 57.Kg3-g2
= (-0.21) Depth: 15/39 00:00:20 16914kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.Kh2-g3 Kf8-g7 49.Kg3-h2 h7-h5 50.h3-h4 Qc6-e4 51.Qe3xe4 Rc4xe4 52.Kh2-g3 Re4-e3+ 53.Kg3-f2 Re3-d3 54.Kf2-e2 Rd3-g3 55.Ke2-d1
³ (-0.27) Depth: 16/41 00:00:39 33132kN
44...Be6-c4 45.Bd3xc4 Rc8xc4 46.Re2-c2 Qd7-c6 47.Bh6xf8 Kg8xf8 48.h3-h4 Qc6-c7 49.Kh2-g1 Qc7-c8 50.Qe3-d3 Qc8-c6 51.g2-g3 Qc6-b6+ 52.Kg1-g2 Qb6-b7+ 53.Kg2-h3 Qb7xb4 54.Qd3-d8+ Kf8-g7 55.Qd8-f6+ Kg7-g8 56.Qf6xa6 Qb4-b3 57.Qa6-a8+ Kg8-g7
³ (-0.36) Depth: 17/48 00:01:47 93586kN

(, 12.06.2007)


Uri

[Uri Blass]

I can add that I read in the link the following:

" Diep was better, even winning, in the middle game, but it was difficult. The attack should be handled precisely and with care. For a human being that is not difficult, but for a program (at least for the current programs) this is a difficult task."

I disagree with it.

I do not see an easy win for white in the game.
If the editor of http://www.grappa.univ-lille3.fr/icga/n ... .php?id=17
believes that there is an easy win("not difficult for a human being") he should give move of white and plan to do and humans can check if they can beat top programs with white based on the plan.

Uri

[Will Singleton]

I read about this game the following in
http://www.grappa.univ-lille3.fr/icga/n ... .php?id=17

"He could not reach a sufficiently deep level of search in the tree and had to see that his opponent easily made 18 plies. His move 44. Bd3 changed the position from an equal position into a lost position."

I find that this move is result of a poor evaluation and not result of search that is not deep enough.

Toga easily avoid this move at depth 9 or bigger.

My little program also avoids it quickly, but I wouldn't say my eval is as good or even close to Diep's. So the problem probably lies elsewhere, either a hash inconsistency or something else related to in-game data. Would Diep play the same blunder from a cold start?

[Andre van Ark]

Hi,

In the past I did a lot of Betatesting for Diep and my experience is the following:

Vincent has provided Diep with a huge amount off chessknowledge. The drawback off this approach is that Diep calculates very slow. Around 200 Kn/s. on my AMD Athlon(tm) 64 X2 Dual Core Processor 3800+, 2200 Mhz.

As a result of this, in crucial (tactical) positions Diep reached only 9 ~ 10 ply. So it was not able to reach a sufficiently deep level of search in the tree and lost because it did not see "a tacticall joke".

The question rises: What is the reason off Blass's inferior posting

[Event "WCCC 2004"]
[Site "Bar-Ilan University, Ramat-Gan"]
[Date "2004.??.??"]
[Round "6"]
[White "Movei"]
[Black "Diep"]
[Result "0-1"]
[ECO "B96"]
[PlyCount "72"]
[EventDate "2004.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Nc6 8. Nxc6
bxc6 9. Qf3 Qb6 10. O-O-O Rb8 11. b3 Nd7 12. Kb1 Nc5 13. Be2 Qc7 14. f5 e5 15.
Rhf1 a5 16. Ka1 a4 17. bxa4 h6 18. Bh4 Qa5 19. Bc4 Rb4 20. Bb3 Ba6 21. Rfe1 Be7
22. Bxe7 Kxe7 23. Qg3 Rhb8 24. Qxg7 Rxb3 25. f6+ Kd7 26. Qxf7+ Kc8 27. Ne2 R3b6
28. Qg8+ Kb7 29. Qg3 Bxe2 30. Rxe2 Qxa4 31. Rc1 Ka8 32. Re3 Qd4+ 33. Rc3 Nxe4
34. Qe3 Nxc3 35. Qe4 Rb1+ 36. Rxb1 Rxb1# 0-1

Best wishes,
André

[Daniel Mehrmann]

The question rises: What is the reason off Blass's inferior posting

Hi Andre,

i fully agree with you and asking the same.
DIEP is some sort of super knowledge engine. On my old Athlon XP 3200+ (only 32 bit) DIEP makes around 90.000 nps and reaching low depths.
But DIEP is tactical very very stable and positional very good. I always enjoyed the matches between DIEP and Homer. Compared between DIEP and Movei, Movei is just , what i'am calling, a "searcher".

Comparing DIEP with Toga/Fruit is just a joke. These are completely different engines. Toga/Fruit lives from his search and well tuned tiny evaluation.
DIEP lives from his very big chess knowledge which is not tuned.

Best,
Daniel

[Uri Blass]

Hi,

In the past I did a lot of Betatesting for Diep and my experience is the following:

Vincent has provided Diep with a huge amount off chessknowledge. The drawback off this approach is that Diep calculates very slow. Around 200 Kn/s. on my AMD Athlon(tm) 64 X2 Dual Core Processor 3800+, 2200 Mhz.

As a result of this, in crucial (tactical) positions Diep reached only 9 ~ 10 ply. So it was not able to reach a sufficiently deep level of search in the tree and lost because it did not see "a tacticall joke".

The question rises: What is the reason off Blass's inferior posting

[Event "WCCC 2004"]
[Site "Bar-Ilan University, Ramat-Gan"]
[Date "2004.??.??"]
[Round "6"]
[White "Movei"]
[Black "Diep"]
[Result "0-1"]
[ECO "B96"]
[PlyCount "72"]
[EventDate "2004.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Nc6 8. Nxc6
bxc6 9. Qf3 Qb6 10. O-O-O Rb8 11. b3 Nd7 12. Kb1 Nc5 13. Be2 Qc7 14. f5 e5 15.
Rhf1 a5 16. Ka1 a4 17. bxa4 h6 18. Bh4 Qa5 19. Bc4 Rb4 20. Bb3 Ba6 21. Rfe1 Be7
22. Bxe7 Kxe7 23. Qg3 Rhb8 24. Qxg7 Rxb3 25. f6+ Kd7 26. Qxf7+ Kc8 27. Ne2 R3b6
28. Qg8+ Kb7 29. Qg3 Bxe2 30. Rxe2 Qxa4 31. Rc1 Ka8 32. Re3 Qd4+ 33. Rc3 Nxe4
34. Qe3 Nxc3 35. Qe4 Rb1+ 36. Rxb1 Rxb1# 0-1

Best wishes,
André

Depth 9 was enough here for Toga to see that Bd3 is worse and certainly diep could get more than depth 9(it could not get depth 18 but every program today when it use good hardware in tournament time control get at least depth 12 even with no reduction tricks except null move pruning with R=3 that vincent is using)

I think that vincent is arrogant when he claims that his program has the best evaluation and I do not like this behaviour and this is the reason for my post.

It can be hash problem(I do not know what is the problem) but my impression is that diep's problem was not some tactics that you can see at bigger depth but evaluation.

Of course bigger depth may help diep to see more because bigger depth helps also with bad evaluation but if diep needs depth that is bigger than 12 to see that Bd3 is not the best move it has serious problems.

Uri

[Andre van Ark]

Hi Daniel,

Nice to see that you, as a programmer off a chessengine, shares my opinion about Diep! Since I am dad, I am not seriously involved in betatesting anymore, but I use my "old" Diepversions for the analysis off the games I played at my local chessclub.

A match Diep - Homer should be able produce very interesting games.

Talking about Homer: the evaluation of your newest private version was very stable and reliable during the kibitzing off yesterdays game "Shreddi - Zappi".

That promises a lot for the coming CSVN-tournament that will be hold in october

Best wishes,
André

[tiger]

Hi,

In the past I did a lot of Betatesting for Diep and my experience is the following:

Vincent has provided Diep with a huge amount off chessknowledge. The drawback off this approach is that Diep calculates very slow. Around 200 Kn/s. on my AMD Athlon(tm) 64 X2 Dual Core Processor 3800+, 2200 Mhz.

As a result of this, in crucial (tactical) positions Diep reached only 9 ~ 10 ply. So it was not able to reach a sufficiently deep level of search in the tree and lost because it did not see "a tacticall joke".

The question rises: What is the reason off Blass's inferior posting

[Event "WCCC 2004"]
[Site "Bar-Ilan University, Ramat-Gan"]
[Date "2004.??.??"]
[Round "6"]
[White "Movei"]
[Black "Diep"]
[Result "0-1"]
[ECO "B96"]
[PlyCount "72"]
[EventDate "2004.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Nc6 8. Nxc6
bxc6 9. Qf3 Qb6 10. O-O-O Rb8 11. b3 Nd7 12. Kb1 Nc5 13. Be2 Qc7 14. f5 e5 15.
Rhf1 a5 16. Ka1 a4 17. bxa4 h6 18. Bh4 Qa5 19. Bc4 Rb4 20. Bb3 Ba6 21. Rfe1 Be7
22. Bxe7 Kxe7 23. Qg3 Rhb8 24. Qxg7 Rxb3 25. f6+ Kd7 26. Qxf7+ Kc8 27. Ne2 R3b6
28. Qg8+ Kb7 29. Qg3 Bxe2 30. Rxe2 Qxa4 31. Rc1 Ka8 32. Re3 Qd4+ 33. Rc3 Nxe4
34. Qe3 Nxc3 35. Qe4 Rb1+ 36. Rxb1 Rxb1# 0-1

Best wishes,
André

Depth 9 was enough here for Toga to see that Bd3 is worse and certainly diep could get more than depth 9(it could not get depth 18 but every program today when it use good hardware in tournament time control get at least depth 12 even with no reduction tricks except null move pruning with R=3 that vincent is using)

I think that vincent is arrogant when he claims that his program has the best evaluation and I do not like this behaviour and this is the reason for my post.

It can be hash problem(I do not know what is the problem) but my impression is that diep's problem was not some tactics that you can see at bigger depth but evaluation.

Of course bigger depth may help diep to see more because bigger depth helps also with bad evaluation but if diep needs depth that is bigger than 12 to see that Bd3 is not the best move it has serious problems.

Uri

I fully agree with you Uri.


// Christophe

[tiger]

Hi,

In the past I did a lot of Betatesting for Diep and my experience is the following:

Vincent has provided Diep with a huge amount off chessknowledge. The drawback off this approach is that Diep calculates very slow. Around 200 Kn/s. on my AMD Athlon(tm) 64 X2 Dual Core Processor 3800+, 2200 Mhz.

As a result of this, in crucial (tactical) positions Diep reached only 9 ~ 10 ply. So it was not able to reach a sufficiently deep level of search in the tree and lost because it did not see "a tacticall joke".

The question rises: What is the reason off Blass's inferior posting

[Event "WCCC 2004"]
[Site "Bar-Ilan University, Ramat-Gan"]
[Date "2004.??.??"]
[Round "6"]
[White "Movei"]
[Black "Diep"]
[Result "0-1"]
[ECO "B96"]
[PlyCount "72"]
[EventDate "2004.??.??"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4 Nc6 8. Nxc6
bxc6 9. Qf3 Qb6 10. O-O-O Rb8 11. b3 Nd7 12. Kb1 Nc5 13. Be2 Qc7 14. f5 e5 15.
Rhf1 a5 16. Ka1 a4 17. bxa4 h6 18. Bh4 Qa5 19. Bc4 Rb4 20. Bb3 Ba6 21. Rfe1 Be7
22. Bxe7 Kxe7 23. Qg3 Rhb8 24. Qxg7 Rxb3 25. f6+ Kd7 26. Qxf7+ Kc8 27. Ne2 R3b6
28. Qg8+ Kb7 29. Qg3 Bxe2 30. Rxe2 Qxa4 31. Rc1 Ka8 32. Re3 Qd4+ 33. Rc3 Nxe4
34. Qe3 Nxc3 35. Qe4 Rb1+ 36. Rxb1 Rxb1# 0-1

Best wishes,
André

So what? Diep is 2 to 4 times slower than other programs and it translates into 10 plies of search less than other programs?

The fact that Diep has a better evaluation than other programs is unproven. It is just a claim of its author.

The fact that Diep is unable to search deep has been both proven and used many time as an excuse by its author, but what does it show? That Vincent has not been able to write a decent search.


// Christophe

[Andre van Ark]

I think that vincent is arrogant when he claims that his program has the best evaluation and I do not like this behaviour and this is the reason for my post.

I think it is arrogant to judge an engine what you don't own and never will own

Best wishes,
André