A few games with DoubleCheck 3.3

Mike S. · Post by **Mike S.** » Mon Apr 23, 2012 5:05 am

1m+1s, Atom N455/1.66 GHz
32 Bit, 1 thread each, 64 MB hash
Balanced-16.ctg openings
ponder off, Tbs. for GUI on
Fritz 13, Windows 7 32-bit

Code: Select all

DCh33   2012

DoubleCheck 3.3 JA   - Crafty 23.02            5.0 - 7.0    +3/=4/-5    41.67%
DoubleCheck 3.3 JA   - Fritz 5.32              5.5 - 6.5    +4/=3/-5    45.83%
DoubleCheck 3.3 JA   - Fruit 05/11/03          2.5 - 9.5    +1/=3/-8    20.83%
DoubleCheck 3.3 JA   - Shredder Classic 4u     2.5 - 9.5    +1/=3/-8    20.83%
DoubleCheck 3.3 JA   - Texel 1.01 32-bit       3.5 - 8.5    +3/=1/-8    29.17%

Total&#58; 19.0/60 = 31.7%

Sometimes, new engines give me such a good first impression that I have a tendency to choose too strong test opponents.

But the match loss against Fritz 5.32 was very narrow. OTOH, even Texel won clearly. But this were only 5x12 games.

Since ponder on-tests are becoming popular, I think future DoubleCheck versions will be interesting for the large rating list tests, if the ponder feature would be added.

I wonder if DoubleCheck would benefit from a more flexible time consumption, in games. Although, in general an engine's time consumption don't interest me much because it is meaningless in analysis mode. But it may lead to underestimation of an engine's quality if that is based on practical game results, only.

This is from a 1+1 game against Fruit 051103. The blue line is DoubleCheck's time consumption, and the black line is Fruit's, for ~90 moves.

(I saw from previous versions that it looks somewhat different, but not much more flexible, at 3m+2s. Maybe it is much better at long TCs.; I haven't tested that.)

One example game:

[Event "DCh33 "]
[Site "?"]
[Date "2012.04.23"]
[Round "9.1"]
[White "DoubleCheck 3.3 JA"]
[Black "Crafty 23.02"]
[Result "1-0"]
[ECO "D15"]
[Annotator "0.61;0.72"]
[PlyCount "151"]
[TimeControl "60+1"]

{Intel(R) Atom(TM) CPU N455 @ 1.66GHz 1662 MHz W=13.2 plies; 124kN/s;
Balanced-16.ctg B=11.5 plies; 501kN/s; 23 TBAs; Balanced-16.ctg} 1. d4 {B 0}
d5 {B 0} 2. c4 {B 0} c6 {B 0} 3. Nf3 {B 0} Nf6 {B 0} 4. Nc3 {B 0} a6 {B 0} 5.
c5 {B 0} Bf5 {B 0} 6. Bf4 {B 0} Nbd7 {B 0} 7. e3 {B 0} e6 {B 0} 8. Bd3 {0.61/
12 3} Bxd3 {0.72/12 3} 9. Qxd3 {0.55/13 3} Be7 {0.56/12 4} 10. O-O {0.52/12 3}
Nh5 {0.55/12 4 (0-0)} 11. Be5 {0.53/11 3 (Bd6)} O-O {0.53/12 3} 12. Rfd1 {0.54/
11 3 (Bd6)} Nhf6 {0.26/13 3} 13. Rac1 {0.48/11 2 (Ng5)} Nxe5 {0.19/13 3} 14.
Nxe5 {0.40/13 2} Qc7 {0.20/13 3} 15. Nf3 {0.41/12 2 (Rc2)} Ng4 {0.10/12 3
(Rad8)} 16. h3 {0.53/13 2 (Rc2)} Nh6 {0.06/12 3 (Nf6)} 17. g3 {0.60/13 2 (e4)}
Nf5 {-0.13/12 3} 18. Kg2 {0.59/12 2 (Kh1)} Rad8 {-0.13/10 3} 19. a4 {0.59/12 2
(Rd2)} Bf6 {-0.27/11 3 (Rfe8)} 20. b4 {0.62/11 2} Rfe8 {-0.16/11 6} 21. b5 {0.
60/13 2} axb5 {-0.03/11 3 (Qa5)} 22. axb5 {0.61/12 2} Ra8 {-0.05/11 2 (b6)} 23.
Ra1 {0.71/12 2} Red8 {0.11/10 3 (Qd7)} 24. b6 {0.81/10 2} Qb8 {0.40/13 3} 25.
Ne2 {0.76/11 2 (Rxa8)} Be7 {0.27/10 2 (g6)} 26. Qc3 {0.84/11 2 (Nf4)} Bf6 {0.
36/10 1 (g6)} 27. Rxa8 {0.95/12 2 (Nf4)} Qxa8 {0.39/11 0} 28. Ra1 {0.93/14 2}
Qc8 {0.51/13 2} 29. Nf4 {0.90/14 2} Qb8 {0.51/13 2 (g6)} 30. Nh5 {0.93/13 2
(Ra7)} Be7 {0.50/14 3} 31. Ra7 {0.93/14 2} Kf8 {0.50/13 1 (g6)} 32. Nf4 {0.92/
12 2 (Qb4)} Bf6 {0.54/12 3 (g6)} 33. h4 {0.97/12 2 (Nd3)} Qc8 {0.53/10 2 (g6)}
34. h5 {0.94/11 2} Kg8 {0.59/10 1} 35. Nd3 {0.83/11 2} Be7 {0.56/11 1 (g6)} 36.
Qb4 {0.93/12 2 (g4)} Nh6 {0.58/10 1} 37. Qe1 {0.89/13 2} Ng4 {0.69/11 2 (f6)}
38. Qh1 {0.97/13 2} Re8 {0.70/13 3 (h6)} 39. Qh3 {0.90/13 2} Nh6 {0.70/12 1}
40. Kg1 {0.92/12 2 (g4)} Kh8 {0.70/10 1 (Qb8)} 41. Kh2 {0.92/13 2 (Nf4)} Rd8 {
0.70/12 2} 42. Kg2 {0.91/12 2} Kg8 {0.67/12 2} 43. g4 {1.04/14 2 (Nf4)} f6 {0.
82/11 1} 44. Qg3 {1.02/14 2} Rd7 {0.78/12 1} 45. Kf1 {1.04/13 2 (g5)} Nf7 {0.
77/11 2} 46. Ke2 {1.06/15 2 (Kg1)} Qe8 {0.71/11 2 (Nh6)} 47. g5 {1.31/12 2
(Qh3)} Kf8 {1.28/10 1 (Qc8)} (47... fxg5 $2 48. Nfe5 (48. Nde5)) 48. g6 {1.61/
13 2} hxg6 {2.35/12 2} 49. Nf4 {1.96/12 2} Bd8 {2.38/12 1 (e5)} 50. Nxg6+ {2.
46/13 2} Kg8 {2.38/4 0} 51. Qb8 {2.60/15 2} Nh8 {3.27/12 2} 52. Rxb7 {2.69/14 2
} Rxb7 {4.91/14 2} 53. Qxb7 {4.09/15 2} Nxg6 {5.40/15 2} 54. hxg6 {4.42/12 2}
Kf8 {5.73/14 1} 55. Qb8 {6.20/13 2 (Qc8)} Bxb6 {8.72/14 3} 56. Qd6+ {9.60/12 2}
Qe7 {8.72/14 1} 57. Qh2 {13.15/13 1 (Qxe7+)} Qa7 {13.02/12 1} 58. Qh8+ {13.49/
13 1} Ke7 {13.02/4 0} 59. Qxg7+ {13.71/14 1 (cxb6)} Ke8 {13.27/12 2} 60. Qg8+ {
13.71/13 1} Ke7 {13.63/12 2 (Kd7)} 61. Qf7+ {14.01/12 1} Kd8 {13.63/4 0} 62.
Qxa7 {14.22/13 1 (cxb6)} Bxa7 {16.77/14 2} 63. g7 {14.27/13 1} Kc7 {17.39/14 2
(Bb8)} 64. g8=Q {17.20/14 1} Bb8 {17.74/13 2 (Kb7)} 65. Qf7+ {17.46/12 1} Kd8 {
17.74/12 1 (Kc8)} 66. Qxf6+ {20.28/11 1 (Qf8+)} Kc8 {21.00/12 3} 67. Qxe6+ {25.
69/12 1} Kd8 {21.00/12 1} 68. Qg8+ {25.69/11 1 (Qxc6)} Kc7 {21.15/12 1} 69. Qe8
{31.46/14 1 (Qf7+)} Kb7 {21.04/12 1} 70. Qd7+ {#13/10 1} Bc7 {#6/12 1} 71. Ne5
{#13/11 1} Kb8 {#5/12 2} 72. Nxc6+ {#9/8 0} Kb7 {#4/12 2} 73. f4 {#10/8 0 (Kd3)
} Ka6 {#3/11 1} 74. Qxc7 {#8/7 0} Kb5 {#2/4 0} 75. Qb6+ {#6/5 0 (Qb7+)} Kc4 {
#1/10 1 (Ka4)} 76. Qb4# {0.01/1 0} 1-0

[D]1q1r2k1/Rp2bppp/1Pp1p3/2Pp1n1N/3P4/2Q1PNPP/5PK1/8 b - - 0 31

lucasart · Post by **lucasart** » Mon Apr 23, 2012 1:50 pm

Mike S. wrote:
This is from a 1+1 game against Fruit 051103. The blue line is DoubleCheck's time consumption, and the black line is Fruit's, for ~90 moves.

This is normal. DoubleCheck's strategy in time allocation is plain and stupid, given by the following formula:

Code: Select all

time_for_move = min&#40;remaining_time/40 + increment, remaining_time - 25ms&#41;

Improving this (playing "easy moves" quicker leaving more time for other moves) is on my todo list. However, so far I haven't found anything that gives me a significant elo increase in self-play, so I kept it as it is.

But thanks for your interest in DoubleCheck

A few games with DoubleCheck 3.3

A few games with DoubleCheck 3.3

Re: A few games with DoubleCheck 3.3