Toga - Deep Sjeng continued at 20 minutes + 10 seconds

[Eelco de Groot]

I ran some more matches of Toga Blueberry settings against Deep Sjeng 2.5. I can not directly compare the results with the first match because I changed more than one parameter, but the results have to stand on their own.

What I changed was the timecontrol, I went from game in 15 minutes + ten seconds to game in twenty minutes + ten seconds. It does not seem a lot different but it is still almost ⅓ extra time.
I also wanted to test some other Toga settings, the Toga Checkov settings for which I don't have any other testresults, so I don't really know how strong they are compared to Toga Blueberry 0.94a.

Deep Sjeng 2.5 seems to do better in these matches so far, it could be that Toga Checkov 1 and Toga Checkov 2 are a lot weaker than earlier settings or weaker at this timecontrol, but the main thing is I think that Deep Sjeng is playing better the longer the timecontrol is, of course a very useful quality and better than an engine that only plays Blitz. In some of the matches we have seen in the earlier thread, the difference could well be a 100 points or more going from Blitz to 40 moves in 40 minutes, but of course we could use a lot more tests to confirm that. I only have a few matches played at this timecontrol of 20 +10 so the data is not yet significant, just a first impression. Tested on single core Athlon 2009 MHz.


First Run "A", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+1 =0 -1]

First Run "B", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Deep Sjeng 2.5 wins 1½-½ [+1 =1 -0]

Second Run "A", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3 Checkov wins 1½-½ [+1 =1 -0]

Second Run "B", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+0 =2 -0]

Intermediate result Toga 1.3 "Checkov" 1 - Deep Sjeng 2.5: 4-4


First Run "A", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Deep Sjeng 2.5 wins 2-0 [+2 =0 -0]

First Run "B", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga Checkov 2 wins 1½-½ [+1 =1 -0]

Second Run "A", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga Checkov 2 wins 1½-½ [+1 =1 -0]

Second Run "B", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+0 =2 -0]

Intermediate result Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5: 4-4

Toga 1.3 "Checkov" 2 is almost the same as Toga 1.3 "Checkov" 1 but the difference is Material which is set at 100 in version 2 and at 95 in version 1.

[Eelco de Groot]

Some more matches, Toga 1.3x4 was added and at the moment has the lead in the matches against Deep Sjeng.

I also wanted to report that, very occasionally, Deep Sjeng can lose on time, at least I have seen this twice now in all the matches. It happens very rarely therefore I think it is a difficult bug to catch. The last time it happened was just now, and similar to the first time: Deep Sjeng has made its move, this is reported in the Shredder GUI move window, but somehow the move I think was not transferred to the GUI. Shredder GUI then stops the match and comes with a pop-up that the engine has lost on time. In this particular case Deep Sjeng had a lost position but I don't know if that has anything to do with it. I could not see anything strange about the last position or game. This was the game:

[Event "20 minuten per partij + 10 seconden per zet"]
[Site "Engine Match"]
[Date "2007.05.15"]
[Round "1"]
[White "Toga II 1.3x4 C"]
[Black "Deep Sjeng 2.5"]
[Result "*"]

1. d4 {Boek 0s} d5 {Boek 0s} 2. Nf3 {Boek 0s} c6 {Boek 0s}
3. c4 {Boek 0s} Nf6 {Boek 0s} 4. Nc3 {Boek 0s} dxc4 {Boek
0s} 5. a4 {Boek 0s} Bf5 {Boek 0s} 6. Ne5 {Boek 0s} e6 {Boek
0s} 7. f3 {Boek 0s} Bb4 {-0.07/13 38s} 8. e4 {Boek 0s} Bxe4
{+0.54/13 44s} 9. fxe4 {Boek 0s} Nxe4 {+0.65/12 18s}
10. Bd2 {Boek 0s} Qxd4 {0.00/13 43s} 11. Nxe4 {Boek 0s}
Qxe4+ {+0.04/13 1:17m} 12. Qe2 {Boek 0s} Bxd2+ {+0.04/13
1:13m} 13. Kxd2 {Boek 0s} Qd4+ {Boek 0s} 14. Kc2 {Boek 0s}
Na6 {-0.33/13 1:04m} 15. Rd1 {Boek 0s} Qc5 {-0.36/13 45s}
16. Kb1 {0.00/13 33s} c3 {-0.27/13 32s} 17. Rc1 {+0.23/14
1:19m} Nb4 {-0.24/14 48s} 18. Rxc3 {+0.15/14 32s} Qd5
{-0.25/15 59s} 19. Qc4 {+0.24/14 37s} a5 {-0.24/15 47s}
20. Qxd5 {+0.32/15 35s} exd5 {-0.16/15 29s} 21. Rb3
{+0.25/16 50s} g6 {-0.27/16 42s} 22. g4 {+0.21/16 44s} Rf8
{-0.24/16 29s} 23. Rh3 {+0.31/17 42s} f6 {-0.35/18 35s}
24. Nd3 {+0.27/18 49s} Rf7 {-0.37/18 47s} 25. Nxb4
{+0.53/17 38s} axb4 {-0.42/18 30s} 26. b3 {+0.50/18 41s}
Ke7 {-0.45/17 1:10m} 27. Be2 {+0.55/16 47s} Kd6 {-0.48/17
39s} 28. Rf1 {+0.62/17 31s} Re8 {-0.50/16 19s} 29. Bd3
{+0.56/17 26s} Re5 {-0.51/16 16s} 30. a5 {+0.59/17 44s} Rg5
{-0.37/16 28s} 31. Rf4 {+0.58/17 37s} c5 {-0.37/16 28s}
32. Rhf3 {+0.54/16 25s} Ke6 {-0.42/15 15s} 33. Kc2
{+0.63/16 25s} h5 {-0.34/15 19s} 34. Re3+ {+0.69/15 28s}
Kd6 {-0.55/14 18s} 35. h3 {+0.66/16 34s} Rf8 {-0.52/15 15s}
36. Re2 {+0.47/16 1:34m} hxg4 {-0.39/16 14s} 37. hxg4
{+0.60/16 21s} Rf7 {-0.42/16 22s} 38. Re8 {+0.87/15 25s} f5
{-0.37/16 30s} 39. Rd8+ {+0.82/15 31s} Ke5 {-0.39/16 29s}
40. Rf2 {+0.98/16 24s} Rxg4 {-0.41/16 17s} 41. Re2+
{+1.00/16 32s} Kf4 {-0.51/15 6s} 42. Rxd5 {+1.00/16 27s}
Rc7 {-0.55/15 22s} 43. Kb2 {+1.05/16 28s} Kf3 {-0.58/15
16s} 44. Rc2 {+1.34/16 28s} Rg2 {-0.76/17 33s} 45. Rxg2
{+1.34/17 17s} Kxg2 {-0.81/17 9s} 46. Bb5 {+1.26/18 26s}
Kg3 {-0.83/17 19s} 47. Rd7 {+2.38/19 16s} Rxd7 {-1.95/19
16s} 48. Bxd7 {+2.63/21 26s} Kf4 {-1.95/19 16s} 49. Bc6
{+2.66/20 16s} Ke5 {-2.33/19 15s} 50. Bxb7 {+2.67/20 27s}
Kd6 {-2.55/19 18s} 51. Ba6 {+2.84/20 18s} f4 {-1.88/16 12s}
52. Kc2 {+3.96/19 26s} g5 {-2.57/16 8s} 53. Kd3 {+3.96/19
24s} g4 {-2.55/17 22s} 54. Bc8 {+5.16/20 23s} g3 {-2.42/13
9s} 55. Bb7 {+8.57/19 23s} *

Last position and output from Deep Sjeng 2.5:

[D]8/1B6/6p1/P1p1kp2/1p6/1P6/1K6/8 b - -


16/40 0:07 -3.60 55...Kc7 56.Le4 Kc8 57.Kc4 Kb8 58.Kxc5 Ka7 59.Kxb4 Ka6 60.Lf3 Ka7 61.Kc4 Kb8 62.b4 Kc7 63.a6 (5.609.619) 734 edb:1.472
beste zet: Kd6-c7 tijd: 0:07.922 min n/s: 727.999 nodes: 5.698.777 edb: 1.527

Deep Sjeng played 46...Kg3? at the last minute in this position, but unfortunately it loses quickly after White's reply 47.Rd7 which Deep Sjeng does see in the next ply.

[D]8/1pr5/6p1/PBpR1p2/1p6/1P6/1K4k1/8 b - -

Engine: Deep Sjeng 2.5 (128 MB)
gemaakt door Gian-Carlo Pascutto

11/24 0:27 -0.71 46...f4 47.Kc2 Kf3 48.Ld3 Kg4
49.Lxg6 f3 50.Kd2 Kf4 51.Tf5+ Kg3
52.Tg5+ Kf4 53.Tg1 Td7+ 54.Ld3 (288.202) 7

12/24 0:27 -0.71 46...f4 47.Kc2 Kf3 48.Ld3 Kg4
49.Lxg6 f3 50.Kd2 Kf4 51.Tf5+ Kg3
52.Tg5+ Kf4 53.Tg1 Td7+ 54.Ld3 (471.780) 11

13/30 0:29 -0.51 46...f4 47.Kc2 Kh3 48.Kd3 Kh4 49.Ke4 g5
50.Kf3 Tf7 51.Txc5 g4+ 52.Kf2 f3
53.Tc4 (1.356.574) 36

14/30 0:31 -0.68 46...f4 47.Kc2 Kh3 48.Ld3 f3 49.Kd2 Kg3
50.Tg5+ Kf4 51.Txg6 Tf7 52.Te6 f2
53.Ke2 Kg4 54.Kf1 (2.174.764) 66

15/31 0:34 -0.83 46...f4 47.Kc2 Kh3 48.Tg5 f3 49.Kd2 f2
50.Ke2 Tf7 51.Kf1 Kh4 52.Txg6 Tf3
53.Lc4 Tf8 54.Tg7 (3.298.659) 93

16/35 0:44 -0.80 46...f4 47.Kc2 f3 48.Tg5+ Kf2 49.Kd3 Te7
50.Txc5 Te3+ 51.Kc4 Ke1 52.Kxb4 Te4+
53.Ka3 f2 54.b4 b6 55.Tc1+ Kd2
56.Tc8 bxa5 57.bxa5 (7.910.878) 178 edb:6

17/37 1:00 -0.85 46...f4 47.Kc2 f3 48.Tg5+ Kf2 49.Kd3 Te7
50.Kd2 Te1 51.Txc5 Tb1 52.Lc4 Tb2+
53.Kd3 Ke1 54.Te5+ Te2 55.Txe2+ fxe2
56.Ld5 (14.979.377) 245 edb:42

17/37 1:02 -0.83 46...Kg3 47.Kc2 Kf4 48.Td7 Txd7
49.Lxd7 Ke3 50.Lc8 f4 51.Lxb7 (16.007.673) 254 edb:55

18/38 1:38 -1.23-- 46...Kg3 (31.242.086) 314 edb:132

18/38 1:41 -1.95 46...Kg3 47.Td7 Txd7 48.Lxd7 Kf4
49.Lc8 Ke5 50.Lxb7 Kd6 (32.498.648) 317 edb:167

18/42 2:21 -0.78 46...f4 47.Kc2 f3 48.Tg5+ Kf2 49.Kd2 Tc8
50.Ld3 Tc6 51.Lxg6 c4 52.bxc4 Txc4
53.Tb5 Tc6 54.Ld3 Ta6 55.Lc4 Td6+
56.Ld3 Ta6 57.Lc4 Td6+ 58.Ld3 (50.046.471) 354 edb:288


New matchmaterial Toga vs Deep Sjeng, if anyone wants to see the games send me a P.M.

Eelco

Third Run "A", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3 Checkov wins 1½-½ [+1 =1 -0]


Third Run "B", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+0 =2 -0]


Fourth Run "A", two games Toga 1.3 "Checkov" 0.93g9k18b - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+1 =0 -1]

Intermediate result Toga 1.3 "Checkov" 1 - Deep Sjeng 2.5: 7½-6½



Third Run "A", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga Checkov 2 wins 1½-½ [+1 =1 -0]


Third Run "B", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga Checkov 2 wins 1½-½ [+1 =1 -0]


Fourth Run "A", two games Toga 1.3 "Checkov" 2 - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+1 =0 -1]

Intermediate result Toga 1.3 Checkov 2 - Deep Sjeng 2.5: 8-6



First Run "A", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+0 =2 -0]


First Run "B", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+0 =2 -0]


Second Run "A", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3x4 Default wins 1½-½ [+1 =1 -0]


Second Run "B", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

A tie 1-1 [+1 =0 -1]


Third Run "A", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3x4 Default wins 1½-½ [+1 =1 -0]


Third Run "B", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3x4 Default wins 1½-½ [+1 =1 -0]


Fourth Run "A", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3x4 Default wins 2-0 [+2 =0 -0]


Fourth Run "B", two games Toga 1.3x4 Default (c) - Deep Sjeng 2.5, Shredder GUI, Shredder9.bkt for both programs, 20 minutes per game with 10 seconds Fischer increment. 128 Mb hash for both programs. No bitbases for Toga.

Toga 1.3x4 Default wins 1½-½ [+1 =1 -0]

Intermediate result Toga 1.3x4 Default - Deep Sjeng 2.5 11-5

[Eelco de Groot]

Last position and output from Deep Sjeng 2.5:

[D]8/1B6/6p1/P1p1kp2/1p6/1P6/1K6/8 b - -


16/40 0:07 -3.60 55...Kc7 56.Le4 Kc8 57.Kc4 Kb8 58.Kxc5 Ka7 59.Kxb4 Ka6 60.Lf3 Ka7 61.Kc4 Kb8 62.b4 Kc7 63.a6 (5.609.619) 734 edb:1.472
beste zet: Kd6-c7 tijd: 0:07.922 min n/s: 727.999 nodes: 5.698.777 edb: 1.527

Sorry diagram for last position was incorrect, should be:
[D]8/1B6/3k4/P1p5/1p3p2/1P1K2p1/8/8 b - -