Computer Chess Club Archives




Subject: Re: Request for Analysis--DiepVsTiger12.0 Spanish Champ

Author: Vincent Diepeveen

Date: 13:13:23 12/30/99

Go up one level in this thread

On December 30, 1999 at 15:19:10, Stephen A. Boak wrote:

>I'd like to seem some analysis by strong human players (Vincent?) of the Diep
>vs. Tiger12.0 1999 Spanish Computer Champ game won by Tiger.  Pro's and Con's,
>please--a balanced review.

D:\diep\spain99>type ct6.pgn
[Event "Sp 99"]
[Site "?"]
[Date "29.12.99"]
[Round "6"]
[White "Diep"]
[Black "Chess Tiger"]
[Result "0-1"]

1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 e6 5. e3 Nbd7 6. Bd3 dxc4 7. Bxc4 b5
8. Bd3 a6 9. e4 c5 10. e5 cxd4 11. Nxb5 Nxe5 12. Nxe5 axb5 13. Qf3 Bb4+
14. Ke2 Rb8 15. Qg3 Qd6 16. Nf3 Qxg3 17. hxg3 Bd6 18. Nxd4 Bd7 19. Bg5 Be5
20. Be3 h6 21. f4 Bd6 22. Nf3 b4 23. Bd4 Bc6 24. Rac1 Bd5 25. Bxf6 gxf6
26. Bc4 Bb7 27. Bb3 Ke7 28. Kf2 Rbc8 29. Rxc8 Rxc8 30. Rxh6 Rc1 31. Rh5 Rb1
32. Nd2 Rxb2 33. Ke2 f5 34. Rh7 Bc5 35. Kd1 Bxg2 36. Rh2 Bc6 37. Re2 Bd4
38. Bc4 Ba4+ 39. Nb3 Rb1+ 40. Kd2 Ba7 41. Nc1 Kd6 42. Bd3 Ra1 43. Bc2 Bb5
44. Rh2 Bd4 45. Rh1 Ke7 46. Rd1 Bf2 47. Bd3 Bxd3 48. Kxd3 Bxg3 49. Rf1 Bxf4
50. Rxf4 Rxc1 51. Rxb4 e5 52. Rb7+ Kf6 53. Rb6+ Kg5 54. Rb5 e4+
55. Ke3 Rc3+ 56. Kd4 Rd3+ 57. Kc4 Rd1 58. a4 e3 59. Re5 Kf4 60. Re7 Kf3
61. Rxf7 f4 62. Rb7 e2 63. Re7 e1=Q 64. Rxe1 Rxe1 65. a5 Ke4 66. a6 f3
67. a7 Ra1 68. a8=Q+ Rxa8 69. Kb5 f2 0-1

My judgement: black is more than ok after whites move 13.Qf3?

Diep's last bookmove was 17.hxg3

ECO edition 2 here at my home gives 18.Bf4 as great advantage for white.
However dudes missed something during analyzes of that game.
They missed that after 18.Bf4,Bxf4 19.gf4,Bb7!!
This is a peanut move for programs. However i didn't see it either
i have to admit. I would very unlikely have seen it in a normal game too.
It's incredibly well however. Just sacraficing the pawn at b5 is brilliant.
Book continues with Bd7. No doubt that this is good for white.

So logically all programs play Nxd4 there

In the game a position comes on the board where black can attack and attack
and put pressure and play with its 2 strong bishops. Of course that's
chanceless. One day you make either a mistake or simply the black position
wins. It's too easy for black and too tough for white to defend.

I didn't receive any log file yet from DIEP, but as soon as i get it i
can see what depth it got. As it only had very little hashtables (used to
like a lot more nowadays) it didn't search very deep too in that endgame
i expect.

Of course all modern theory knows that only 13.Bxb5 is ok.
The tournament books of Kure (fritz6,nimzo) and Noomen (tiger,rebel)
know this too.

Not a single player nowadays plays the variation Qf3 anymore.
Statistics in wide book from DIEP:
 # move freq games  score learn ==> from harddisk
 7 d1b3   0       2      1  0
10 d1f3  12      28     14  0
27 d3b5  60      57     33  0
46 e1g1  27      37     23  0

So Bb5 had 60% chance to get played,
Qf3 had only 12% that it was played.

You can call that bad luck. But i wouldn't call that
bad luck. If you have for several moves a chance that a bad line
gets played then it's near to dead sure it happens.

Now please don't call my book bad.

Jeroen and Alex just did last years a great job and no one comes near
to their job for now. Not even a grandmaster of 2600+ can write a for
programs better book than them (altermann has made the junior book if
i understood well).

>What did each program do well, and what did each program not do so well?

>When did each program leave book, and what was the score shown by each at that

When i run at my PII450 with 80mb hash then diep says for Nxd4 0.05 as
a score (positive for white).

>What does the strong human player say about the opening score?

Human doesn't have openingsscore. It just says white is playing a defensive
position the whole game. So one day you either collapse or position gets
too bad. Programs nowadays play at a too well level to assume them to
play this as bad as a human would.

>How about the positional 'understanding' shown by the moves of each program?

I have no log file of that. I'm dead sure that there is still a lot to
improve in programs endgame. This was however a classical case where the
'modern' automatic generated book is simply outpowered by the excellent
performance of Jeroen in this case.

>Thanks in advance.

This page took 0.02 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.