DarkThought at Test Games (1998-1999)

From March 1998 to January 1999 DARKTHOUGHT played a series of 280 test games against 20 strong PC chess programs. Prior to the games, the according version of DARKTHOUGHT had achieved good ratings for the well-known test suites BS-2830 (2720), BT-2630 (2521), and LCT-II (2610). All games were played with DARKTHOUGHT running on a 600MHz DEC Alpha-21164a LX164 workstation using 170MB RAM overall while the opponents enjoyed the service of our then fastest PC featuring an AMD K6/233MHz CPU and 64MB of RAM. We list the names and settings of the 20 PC chess programs in alphabetical order below.

• CHESSMASTER 4000 TURBO  --  [Windows/NT 4.0, 32MB hash (ttable size=25),
                                                                 sel=10, chess engine based on THE KING 2.0],

• COMET A.86/A.90  --  [MS-DOS 6.2, 52MB hash, sel=1, pb=on],

• CRAFTY 16.4  --  [Windows/NT 4.0, 32MB hash, no tablebases, engine in FRITZ 5.32, default play style],

• FRITZ 3.10  --  [Windows/NT 4.0, 16MB hash (max.), engine in FRITZ 5],

• FRITZ 5.00/5.03  --  [Windows/NT 4.0, 48MB hash, endgame databases enabled],

• FRITZ 5.32B  --  [Windows/NT 4.0, 48MB hash, default play style],

• GENIUS 3  --  [MS-DOS 6.2, 32MB hash, default play style (active)],

• GENIUS 5  --  [MS-DOS 6.2, 32MB hash, default play style (active)],

• HIARCS 6.0  --  [Windows/NT 4.0, 16MB hash (max.), engine in FRITZ 5],

• JUNIOR 4.6  --  [Windows/NT 4.0, 48MB hash, engine in FRITZ 5],

• JUNIOR 5.0  --  [Windows/NT 4.0, 48MB hash, default play style],

• MCHESS 5  --  [MS-DOS 6.2, 10MB hash (max.), default play style (normal)],

• MCHESS 7.1  --  [MS-DOS 6.2, 60MB hash (max.), default play style (normal)],

• MEPHISTO MEISTERSCHACH  --  [Windows/NT 4.0, hash automatic (24MB), standard
                                                                 style, no handicap, chess engine based on NIMZO 3.5],

• MEPHISTO SHREDDER  --  [Windows/NT 4.0, 24MB/16MB hash/eval, sel=12, endgame
                                                                 databases enabled, chess engine based on SHREDDER 1],

• NIMZO 98  --  [Windows/NT 4.0, 48MB hash, standard style, ``Paderborn'' update],

• NIMZO 99A  --  [Windows/NT 4.0, 48MB hash, NIMZO author's favourite play style
                                                                 (B=435, N=425, P=105, Q=1240, R=685, opp.=Junior)],

• REBEL 8  --  [MS-DOS 6.2, 60MB hash, default play style (strong, normal, sel=N)],

• REBEL 9  --  [MS-DOS 6.2, 60MB hash, default play style (strong, normal, sel=N)],

• SHREDDER 2  --  [Windows/NT 4.0, 24MB/16MB hash/eval, sel=12, endgame
                                                                 databases enabled].

The 266 test games were played from the so-called ``Nunn Positions'' #2 (ECO B89), #3 (ECO C19), #4 (ECO C97), #5 (ECO D36), #7 (ECO E15), #8 (ECO E98) and #9 (ECO A25) with all opening books and learning features of the programs disabled. In order to partly account for the hardware advantage of the 600MHz DEC Alpha-21164a (18.4 SPECint95) over the AMD K6/233MHz (9.6 SPECint95), DARKTHOUGHT played at a rate of 60 moves in 90 minutes while the PC programs received additional 30 minutes on their clocks (60 moves in 120 minutes). The test games ended 209.5:70.5 in favour of DARKTHOUGHT (74.82% winning percentage, see Table 1.16) which showed that it was not only good at solving test suites but also quite promising at standard tournament play.

 

Table 1.16: Strong PC Chess Programs vs. the 1998 Version of DarkThought.

	DT #2	DT #3	DT #4	DT #5	DT #7	DT #8	DT #9	Sum
CO	- -	+ -	- -	- -	- -	- -	- -	1.0:13.0
CM	= -	- -	= -	= +	= -	+ -	= =	5.0:9.0
CR	= =	+ -	= -	= =	- =	= -	= -	5.0:9.0
F3	= =	- -	- -	- -	- -	- -	= =	2.0:12.0
F5	= =	= -	= -	+ -	= =	+ -	- =	5.5:8.5
FB	- -	= -	+ -	- =	= =	= -	- -	3.5:10.5
G3	- =	- =	- =	+ -	= =	= -	= =	5.0:9.0
G5	= =	- =	- -	= -	- -	= -	- -	2.5:11.5
H6	- -	+ -	+ -	- -	= =	- -	= +	4.5:9.5
J4	= =	- -	- -	- =	- -	- =	= =	3.0:11.0
J5	- -	- =	+ =	- =	- +	= -	= -	4.5:9.5
M5	- =	+ -	= -	- =	- -	= -	- -	3.0:11.0
M7	- =	= -	+ -	= -	- =	+ -	= -	4.5:9.5
MM	- -	- -	= =	= -	- -	- -	- -	1.5:12.5
MS	- -	- -	- -	- -	= -	= -	- =	1.5:12.5
N8	+ -	+ -	- -	+ -	= -	- -	- =	4.0:10.0
N9	= -	= -	= =	- -	+ -	- =	- +	4.5:9.5
R8	+ -	+ -	= =	- =	- -	= -	- -	4.0:10.0
R9	+ -	+ -	- -	- -	- -	- -	- =	2.5:11.5
S2	- -	+ -	- -	= =	= -	= -	- =	3.5:10.5
Sum	10.5:29.5	11.5:28.5	10.0:30.0	10.5:29.5	9.0:31.0	8.5:31.5	10.5:29.5	70.5:209.5

 

The below ``Jakarta Revanche'' game against MEPHISTO SHREDDER adds to this promise. The game was started from the position after White's move 14. d5! as played by SHREDDER against DARKTHOUGHT in the 6th round of the 14th WMCC (Jakarta, October 1996). Nearly all computer-chess periodicals published this very position after the championship and asserted the hopelessness of Black's position. But within roughly a minute of computing time our new version of DARKTHOUGHT found a brilliant defense for Black (see moves 16-23) that leads to nearly equal play. MEPHISTO SHREDDER remained clueless about the tricky maneuvre until very late and then even managed to lose the drawish endgame which emerged after move 34.

Mephisto Shredder - DarkThought: [0 - 1]
(Revanche Game of 6th Round at 14th WMCC '96, March 1998)

1. d4 d5 2. c4 e6 3. Nc3 c5 4. cxd5 exd5 5. Nf3 Nc6 6. g3 c4 7. Bg2 Bb4 8. O-O Nge7 9. a3 Ba5 10. e4 dxe4 11. Nxe4 O-O 12. Qe2 b5 13. Rd1 Bf5 14. d5 Nxd5 15. Nh4 Be6 16. Ng5 Re8 17. Qh5 Nf6 18. Rxd8 Raxd8 19. Qe2 Bg4 20. Qxe8+ Rxe8 21. Bxc6 Re1+ 22. Kg2 Bd2 23. Bxd2 Rxa1 24. Bc3 a6 25. Bb7 h6 26. Bxf6 hxg5 27. Bxg5 Rb1 28. Bxa6 Rxb2 29. Nf3 Bxf3+ 30. Kxf3 c3 31. Bc1 Rb1 32. Be3 c2 33. Ke2 c1=Q 34. Bxc1 Rxc1 35. Bxb5 g5 36. h3 Rh1 37. Bd7 Kf8 38. a4 Ke7 39. Bf5 Kd6 40. Kf3 f6 41. Kg2 Ra1 42. Bc2 Ke5 43. Bb3 Kd4 44. h4 gxh4 45. gxh4 Ke5 46. h5 Kf5 47. Bc2+ Kg5 48. Kf3 Kxh5 49. Kf4 Ra2 50. Bd1+ Kg6 51. f3 Rd2 52. Bb3 Rb2 53. Bd1 Rb4+ 54. Ke3 Kg5 55. a5 f5 56. a6 f4+ 57. Kf2 Rb2+ 58. Ke1 Ra2 59. Be2 Kh4 60. Bc4 Ra4 61. Bb5 Ra3 62. Kf2 Rc3 63. Be2 Rc2 64. Ke1 Kg3 65. Kd1 Rc5 66. Ke1 Rc7 67. Bd1 Rc1 68. Kd2 Rc6 69. Be2 Kf2 70. Bd3 Kxf3 71. a7 Rc8 72. Bc4 Rd8+ 73. Ke1 Ke4 74. Bf7 Ra8 75. Kf2 Rxa7 0-1