Author: Shep
Date: 00:50:43 08/05/99
Go up one level in this thread
On August 04, 1999 at 12:40:05, blass uri wrote:
>Please post some of these positions when chessmaster sees a win score 10-20
>moves before its opponent realizes that it is in trouble.
>
>I do not think that chessmaster sees 20 plies more than the opponent so the
>problem must be the evaluation function.
Of course! :)
>There is probably something wrong in the evaluation of other programs and it is
>important to look at these positions to see what is the problem.
You can check out the positions around move 40 in this game (posted here
earlier) below. I suppose CM 6000 would also see a similar advantage here. 6666
showed +1,5 to +2,2 most of the time while Genius wavered between +0.4 and -0.6.
The point in the endgame itself is something different. I think many programs
tend to "see" long checking sequences that push the actual loss over the search
horizon. I believe that they go like "Oops, looks like I'm losing... No wait, I
can give 10 checks in a row, maybe it is perpetual check? Well, I'll rate this
as -0.05 then".
Chessmaster seems to be one of the few (the only?) program who does not suffer
from this problem.
That's why Genius believed it could keep the king in check constantly while CM
"knew" the checking would not go on forever (although he could not possible
"calculate" it that far ahead).
---
Shep
[Event "Test game, 60 min/game"]
[Site "?"]
[Date "1999.07.27"]
[Round "4"]
[White "Genius 6.0"]
[Black "Chessmaster 6666"]
[Result "0-1"]
[ECO "B09"]
[WhiteElo "2645"]
[BlackElo "2700"]
1. e4 d6 2. d4 Nf6 3. Nc3 g6 4. f4 Bg7 5. Nf3 O-O 6. Be3 Nbd7 7. Qd2 Nb6 8.
Bd3 c6 9. O-O Ng4 10. f5 Nxe3 11. Qxe3 e5 12. fxg6 hxg6 13. dxe5 dxe5 14. Qc5
Be6 15. a3 Nd7 16. Qe3 Qe7 17. Rad1 Nb6 18. Nd2 Rad8 19. Be2 Kh7 20. Qf2 Bh6
21. Nb3 Nc4 22. Bxc4 Bxc4 23. Rfe1 b6 24. Nd2 Be6 25. Nf3 f6 26. Rf1 Bc4 27.
Ne2 Rfe8 28. Kh1 Kg7 29. b3 Be6 30. Rxd8 Rxd8 31. h3 Kh7 32. Qh4 Rf8 33. b4
Bc4 34. Re1 a5 35. c3 Ra8 36. Nd2 Bb5 37. Nf1 axb4 38. axb4 Kg7 39. Qg4 Bc4
40. Nfg3 Bg5 41. Rd1 Be6 42. Qf3 Bb3 43. Rf1 Bc4 44. Rd1 Ra2 45. h4 Bb3 46.
Rd3 Ra1+ 47. Ng1 Bf4 48. Qg4 Be6 49. Nf5+ Bxf5 50. exf5 e4 51. Rd1 Rxd1 52.
Qxd1 gxf5 53. Qd4 Bc7 54. Ne2 f4 55. Qd2 Be5 56. Nxf4 Bxf4 57. Qxf4 e3 58.
Qg4+ Kf8 59. Qe2 Qe5 60. c4 Qd4 61. g4 Qd2 62. Qf3 Kg7 63. Qe4 Kf7 64. Qh7+
Kf8 65. Qh6+ Ke7 66. Qf4 Qd1+ 67. Kg2 e2 68. Qc7+ Ke6 69. Qxc6+ Ke5
{Finally, Genius realizes he's going down. Eval drops from -0.6 to -5.
CM was showing a significant advantage since move 40 already.} 70. Qb5+
Kf4 71. Qf5+ Ke3 72. Qf2+ Kd3 73. Qf5+ Kd2 74. Qf2 Kc1 75. Qxf6 e1=Q 76. Qa1+
Kc2 77. Qxd1+ 0-1
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.