Test tourney result.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 12 April 2004 21:36:14:


P933 Test 10-04-2004 = 40/5, ponder=on.
DUAL-P3-933, 2004.04.11 - 2004.04.12
Score SOP 1 2 3 4 5 6 7 8 9
--------------------------------------------------------------------------------------
1: Fruit 1.0 7.0 / 9 45.0 30b+ 10w- 18b+ 5w- 21b+ 19w+ 9w+ 2b+ 3b+
2: Hermann 1.0.5 6.5 / 9 51.5 15w+ 8b+ 7w+ 4b- 10w+ 3b+ 13b= 1w- 11b+
3: BlackBishop 0.9.7g 6.5 / 9 49.5 9b= 17w+ 28b+ 16b+ 4w+ 2w- 8b+ 7b+ 1w-
4: Booot 3.2 6.5 / 9 47.5 12b+ 18w+ 29b+ 2w+ 3b- 8w= 10b+ 13w+ 5b-
5: Delphil 0.7d 6.5 / 9 45.5 19w+ 6w- 12b= 1b+ 29b+ 11w- 18b+ 20b+ 4w+
6: Matheus 2.3 6.5 / 9 44.0 21w+ 5b+ 16w= 10b- 9b+ 13w- 28b+ 17w+ 8w+
7: Uragano 0.73g 6.0 / 9 45.5 20b+ 14w+ 2b- 24w+ 8b- 18w+ 11b+ 3w- 13b+
8: NullMover 0.25 5.5 / 9 48.0 33b+ 2w- 14b+ 19b+ 7w+ 4b= 3w- 10w+ 6b-
9: Eeyore 1.36 5.5 / 9 47.0 3w= 27b+ 13w+ 11b- 6w- 12w+ 1b- 28w+ 17b+
10: Elephant 1.02 5.0 / 9 50.0 17b= 1b+ 22w+ 6w+ 2b- 16b+ 4w- 8b- 12w=
11: Fafis 1.0 5.0 / 9 45.0 31b- 25w+ 15b+ 9w+ 16w- 5b+ 7w- 19b+ 2w-
12: Tinker 4.47 5.0 / 9 44.5 4w- 23b+ 5w= 22w+ 13b- 9b- 29w+ 16b+ 10b=
13: Matacz 0.96 5.0 / 9 43.0 34b+ 29w= 9b- 28w+ 12w+ 6b+ 2w= 4b- 7w-
14: DrunkenMaster 0.9 5.0 / 9 40.5 25b+ 7b- 8w- 15w+ 18b- 23w+ 19b- 22w+ 20w+
15: Enigma 1.1.3 5.0 / 9 33.0 2b- 33w+ 11w- 14b- 20w- 34b+ 32w+ 23b+ 19w+
16: Kanguruh 1.712 4.5 / 9 43.0 26b+ 31w+ 6b= 3w- 11b+ 10w- 17b- 12w- 28b+
17: Neurosis 1.7b 4.5 / 9 42.5 10w= 3b- 32w= 23b+ 24b= 22w+ 16w+ 6b- 9w-
18: RDChess 3.23 4.0 / 9 47.0 23w+ 4b- 1w- 20b+ 14w+ 7b- 5w- 25b= 24w=
19: Smash 0.9-rc1 4.0 / 9 44.0 5b- 20w+ 21b+ 8w- 33w+ 1b- 14w+ 11w- 15b-
20: Bruja 1.2 4.0 / 9 39.5 7w- 19b- 25b+ 18w- 15b+ 33w+ 30b+ 5w- 14b-
21: AICE 0.63 4.0 / 9 37.5 6b- 30w+ 19w- 32b= 1w- 29b= 31w+ 24b= 25w=
22: KKFChess 2.6.1 4.0 / 9 36.5 27w+ 28w= 10b- 12b- 32w+ 17b- 24w= 14b- 30w+
23: Belzebub 0.65 4.0 / 9 33.5 18b- 12w- 30b+ 17w- 31b+ 14b- 26w+ 15w- 34b+
24: Parrot 040330 4.0 / 9 32.5 29b- 34w+ 31b+ 7b- 17w= 28w- 22b= 21w= 18b=
25: Zotron 4.4.6 4.0 / 9 30.5 14w- 11b- 20w- 34b+ 26w= 32b= 33b+ 18w= 21b=
26: Alfil 403.1 4.0 / 9 30.5 16w- 32b= 27w- 30w= 25b= 31b+ 23b- 33b+ 29w=
27: Piranha 0.5 4.0 / 9 28.0 22b- 9w- 26b+ 29w= 28b- 30w- 34w+ 32b= 33w+
28: Awesome 1.45 3.5 / 9 42.5 32w+ 22b= 3w- 13b- 27w+ 24b+ 6w- 9b- 16w-
29: Alex 1.52.09 3.5 / 9 42.0 24w+ 13b= 4w- 27b= 5w- 21w= 12b- 30b= 26b=
30: Hoplite 2.1 3.0 / 9 34.5 1w- 21b- 23w- 26b= 34w+ 27b+ 20w- 29w= 22b-
31: WESP 0.8 3.0 / 9 30.0 11w+ 16b- 24w- 33b- 23w- 26w- 21b- 34w+ 32b+
32: Hokus Pokus 0.43 2.5 / 9 36.0 28b- 26w= 17b= 21w= 22b- 25w= 15b- 27w= 31w-
33: Excelsior 2.32b 2.0 / 9 33.5 8w- 15b- 34b+ 31w+ 19b- 20b- 25w- 26w- 27b-
34: Fusch 1.09 0.0 / 9 34.0 13w- 24b- 33w- 25w- 30b- 15w- 27b- 31b- 23w-
--------------------------------------------------------------------------------------
153 games: +63 =26 -64
Next engines played with Polyglot:
Fuit 1.0 (lost 2 games on time but won the tourney anyway!!)
Piranha 0.5 (dont work with the latest "polyglot", played with ponder=off because of a ponder bug!!)
Fusch 1.09 (no problems)
Alfil 403.1 (worked only with /first(second)Protocolversion=1

Problems found:
Lost on time:
Parrot, 1 game
Fruit, 2 games
Delphil 0.7e, 1 game
Crashed:
Matacz 0.96, at 1 game.
Piranha, Hermann and DrunkenMaster played with ponder=off
Delphil 0.7d replaced by 0.7e after round 2
LGpgnver:
File: P9331004.pgn
Level of warnings: 0
=============================================================================
Game 24: KKFChess 2.6.1-Awesome 1.45
1. c4 c5 2. Nc3 Nc6 3. g3 g6 4. Bg2 Bg7 5. e3 Bxc3 6. dxc3 Nf6 7. Ne2 Ne5
8. b3 d6 9. f4 Nc6 10. O-O Bf5 11. Bxc6+ bxc6 12. Bb2 Bh3 13. Rf3 Bg4 14.
Rf1 Ne4 15. Qc2 Bxe2 16. Qxe2 Qa5 17. Qd3 Nf6 18. e4 Rd8 19. Qe2 d5 20.
cxd5 cxd5 21. exd5 Rxd5 22. Qe3 Rd7 23. b4 cxb4 24. cxb4 Qf5 25. Qc5 Qxc5+
26. bxc5 Rc7 27. Rfc1 Rc6 28. Rab1 O-O 29. Bxf6 exf6 30. Rb7 Rfc8 31. Rxa7
Rxc5 32. Rxc5 Rxc5 33. a4 Rc1+ 34. Kg2 Rc2+ 35. Kg1 Rc1+ 36. Kg2 Rc2+ 37.
Kh3 Ra2 38. g4 Ra3+ 39. Kg2 Ra2+ 40. Kg3 Ra3+ 41. Kg2 Ra2+ 42. Kg3 Ra3+
43. Kf2 Ra2+ 44. Kg1 Rc2 45. h3 Rc4 46. Kg2 Rxf4 47. Kg3 Rd4 48. a5 Rd3+
49. Kg2 h5 50. Rb7 Ra3 51. Rb5 hxg4 52. hxg4 Kf8 53. Rd5 Ke7 54. Rb5 Ke6
55. Kf2 Kd6 56. Ke2 Ra4 57. Kf3 Ke6 58. Rc5 Kd7 59. Rb5 Ke8 60. Rd5 g5 61.
Rb5 Kf8 62. Kg3 Ke8 63. Rf5 Ra3+ 64. Kg2 Ke7 65. Rd5 Ra2+ 66. Kf3 Ra3+ 67.
Kf2 Ra1 68. Rc5 Ra4 69. Kf3 Rf4+ 70. Kg3 Kd7 71. Kh3 Ra4 72. Rb5 Ke8 73.
Kg3 Kd8 74. Kf3 Ke8 75. Kg3
Result :1/2-1/2 {Draw by repetition}
Warning level :3
Final position: 4k3/5p2/5p2/PR4p1/r5P1/6K1/8/8 b - - 29 75
Analyze result:* {no draw found} Claimed by??, no debugfile saved
=============================================================================
1 warnings/errors found
Best wishes,
Leo.



WBEC Ridderkerk homepage.

[Rafael Peña]

Geschrieben von:/Posted by: Rafael Peña at 12 April 2004 22:45:05:
Als Antwort auf:/In reply to: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 12 April 2004 21:36:14:

P933 Test 10-04-2004 = 40/5, ponder=on.
DUAL-P3-933, 2004.04.11 - 2004.04.12
Score SOP 1 2 3 4 5 6 7 8 9
--------------------------------------------------------------------------------------
11: Fafis 1.0 5.0 / 9 45.0 31b- 25w+ 15b+ 9w+ 16w- 5b+ 7w-

Hi Leo!
please send me the pgn file, i want to see fafis 1.0 games =) looks more stronger than old fafis .
(any problem with fafis??)
thank's
Rafael Peña

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 12 April 2004 23:40:11:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Rafael Peña at 12 April 2004 22:45:05:

P933 Test 10-04-2004 = 40/5, ponder=on.
DUAL-P3-933, 2004.04.11 - 2004.04.12
Score SOP 1 2 3 4 5 6 7 8 9
--------------------------------------------------------------------------------------
11: Fafis 1.0 5.0 / 9 45.0 31b- 25w+ 15b+ 9w+ 16w- 5b+ 7w-

Hi Leo!
please send me the pgn file, i want to see fafis 1.0 games =) looks more stronger than old fafis .
(any problem with fafis??)
thank's
Rafael Peña

WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 13 April 2004 10:36:35:
Als Antwort auf:/In reply to: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 12 April 2004 21:36:14:

Next engines played with Polyglot:
Fuit 1.0 (lost 2 games on time but won the tourney anyway!!)
Piranha 0.5 (dont work with the latest "polyglot", played with ponder=off because of a ponder bug!!)
Fusch 1.09 (no problems)
Alfil 403.1 (worked only with /first(second)Protocolversion=1

Just to make sure,
*NEVER USE PONDERING WITH POLYGLOT*
It just does not work.
Incompatibility between xboard and UCI protocols, hard work to fix (not planned soon).
I should add an error message when pondering is used.
If pondering was not used, it might be a performance problem.
Let me know, I am working on raw performance at the moment.
Fabien.

[Andreas Herrmann]

Geschrieben von:/Posted by: Andreas Herrmann at 13 April 2004 11:16:14:
Als Antwort auf:/In reply to: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 12 April 2004 21:36:14:

P933 Test 10-04-2004 = 40/5, ponder=on.
DUAL-P3-933, 2004.04.11 - 2004.04.12
Score SOP 1 2 3 4 5 6 7 8 9
--------------------------------------------------------------------------------------
1: Fruit 1.0 7.0 / 9 45.0 30b+ 10w- 18b+ 5w- 21b+ 19w+ 9w+ 2b+ 3b+
2: Hermann 1.0.5 6.5 / 9 51.5 15w+ 8b+ 7w+ 4b- 10w+ 3b+ 13b= 1w- 11b+
3: BlackBishop 0.9.7g 6.5 / 9 49.5 9b= 17w+ 28b+ 16b+ 4w+ 2w- 8b+ 7b+ 1w-
4: Booot 3.2 6.5 / 9 47.5 12b+ 18w+ 29b+ 2w+ 3b- 8w= 10b+ 13w+ 5b-
5: Delphil 0.7d 6.5 / 9 45.5 19w+ 6w- 12b= 1b+ 29b+ 11w- 18b+ 20b+ 4w+
6: Matheus 2.3 6.5 / 9 44.0 21w+ 5b+ 16w= 10b- 9b+ 13w- 28b+ 17w+ 8w+
7: Uragano 0.73g 6.0 / 9 45.5 20b+ 14w+ 2b- 24w+ 8b- 18w+ 11b+ 3w- 13b+
8: NullMover 0.25 5.5 / 9 48.0 33b+ 2w- 14b+ 19b+ 7w+ 4b= 3w- 10w+ 6b-
9: Eeyore 1.36 5.5 / 9 47.0 3w= 27b+ 13w+ 11b- 6w- 12w+ 1b- 28w+ 17b+
10: Elephant 1.02 5.0 / 9 50.0 17b= 1b+ 22w+ 6w+ 2b- 16b+ 4w- 8b- 12w=
11: Fafis 1.0 5.0 / 9 45.0 31b- 25w+ 15b+ 9w+ 16w- 5b+ 7w- 19b+ 2w-
12: Tinker 4.47 5.0 / 9 44.5 4w- 23b+ 5w= 22w+ 13b- 9b- 29w+ 16b+ 10b=
13: Matacz 0.96 5.0 / 9 43.0 34b+ 29w= 9b- 28w+ 12w+ 6b+ 2w= 4b- 7w-
14: DrunkenMaster 0.9 5.0 / 9 40.5 25b+ 7b- 8w- 15w+ 18b- 23w+ 19b- 22w+ 20w+
15: Enigma 1.1.3 5.0 / 9 33.0 2b- 33w+ 11w- 14b- 20w- 34b+ 32w+ 23b+ 19w+
16: Kanguruh 1.712 4.5 / 9 43.0 26b+ 31w+ 6b= 3w- 11b+ 10w- 17b- 12w- 28b+
17: Neurosis 1.7b 4.5 / 9 42.5 10w= 3b- 32w= 23b+ 24b= 22w+ 16w+ 6b- 9w-
18: RDChess 3.23 4.0 / 9 47.0 23w+ 4b- 1w- 20b+ 14w+ 7b- 5w- 25b= 24w=
19: Smash 0.9-rc1 4.0 / 9 44.0 5b- 20w+ 21b+ 8w- 33w+ 1b- 14w+ 11w- 15b-
20: Bruja 1.2 4.0 / 9 39.5 7w- 19b- 25b+ 18w- 15b+ 33w+ 30b+ 5w- 14b-
21: AICE 0.63 4.0 / 9 37.5 6b- 30w+ 19w- 32b= 1w- 29b= 31w+ 24b= 25w=
22: KKFChess 2.6.1 4.0 / 9 36.5 27w+ 28w= 10b- 12b- 32w+ 17b- 24w= 14b- 30w+
23: Belzebub 0.65 4.0 / 9 33.5 18b- 12w- 30b+ 17w- 31b+ 14b- 26w+ 15w- 34b+
24: Parrot 040330 4.0 / 9 32.5 29b- 34w+ 31b+ 7b- 17w= 28w- 22b= 21w= 18b=
25: Zotron 4.4.6 4.0 / 9 30.5 14w- 11b- 20w- 34b+ 26w= 32b= 33b+ 18w= 21b=
26: Alfil 403.1 4.0 / 9 30.5 16w- 32b= 27w- 30w= 25b= 31b+ 23b- 33b+ 29w=
27: Piranha 0.5 4.0 / 9 28.0 22b- 9w- 26b+ 29w= 28b- 30w- 34w+ 32b= 33w+
28: Awesome 1.45 3.5 / 9 42.5 32w+ 22b= 3w- 13b- 27w+ 24b+ 6w- 9b- 16w-
29: Alex 1.52.09 3.5 / 9 42.0 24w+ 13b= 4w- 27b= 5w- 21w= 12b- 30b= 26b=
30: Hoplite 2.1 3.0 / 9 34.5 1w- 21b- 23w- 26b= 34w+ 27b+ 20w- 29w= 22b-
31: WESP 0.8 3.0 / 9 30.0 11w+ 16b- 24w- 33b- 23w- 26w- 21b- 34w+ 32b+
32: Hokus Pokus 0.43 2.5 / 9 36.0 28b- 26w= 17b= 21w= 22b- 25w= 15b- 27w= 31w-
33: Excelsior 2.32b 2.0 / 9 33.5 8w- 15b- 34b+ 31w+ 19b- 20b- 25w- 26w- 27b-
34: Fusch 1.09 0.0 / 9 34.0 13w- 24b- 33w- 25w- 30b- 15w- 27b- 31b- 23w-
--------------------------------------------------------------------------------------
153 games: +63 =26 -64
Next engines played with Polyglot:
Fuit 1.0 (lost 2 games on time but won the tourney anyway!!)
Piranha 0.5 (dont work with the latest "polyglot", played with ponder=off because of a ponder bug!!)
Fusch 1.09 (no problems)
Alfil 403.1 (worked only with /first(second)Protocolversion=1

Problems found:
Lost on time:
Parrot, 1 game
Fruit, 2 games
Delphil 0.7e, 1 game
Crashed:
Matacz 0.96, at 1 game.
Piranha, Hermann and DrunkenMaster played with ponder=off
Delphil 0.7d replaced by 0.7e after round 2
LGpgnver:
File: P9331004.pgn
Level of warnings: 0
=============================================================================
Game 24: KKFChess 2.6.1-Awesome 1.45
1. c4 c5 2. Nc3 Nc6 3. g3 g6 4. Bg2 Bg7 5. e3 Bxc3 6. dxc3 Nf6 7. Ne2 Ne5
8. b3 d6 9. f4 Nc6 10. O-O Bf5 11. Bxc6+ bxc6 12. Bb2 Bh3 13. Rf3 Bg4 14.
Rf1 Ne4 15. Qc2 Bxe2 16. Qxe2 Qa5 17. Qd3 Nf6 18. e4 Rd8 19. Qe2 d5 20.
cxd5 cxd5 21. exd5 Rxd5 22. Qe3 Rd7 23. b4 cxb4 24. cxb4 Qf5 25. Qc5 Qxc5+
26. bxc5 Rc7 27. Rfc1 Rc6 28. Rab1 O-O 29. Bxf6 exf6 30. Rb7 Rfc8 31. Rxa7
Rxc5 32. Rxc5 Rxc5 33. a4 Rc1+ 34. Kg2 Rc2+ 35. Kg1 Rc1+ 36. Kg2 Rc2+ 37.
Kh3 Ra2 38. g4 Ra3+ 39. Kg2 Ra2+ 40. Kg3 Ra3+ 41. Kg2 Ra2+ 42. Kg3 Ra3+
43. Kf2 Ra2+ 44. Kg1 Rc2 45. h3 Rc4 46. Kg2 Rxf4 47. Kg3 Rd4 48. a5 Rd3+
49. Kg2 h5 50. Rb7 Ra3 51. Rb5 hxg4 52. hxg4 Kf8 53. Rd5 Ke7 54. Rb5 Ke6
55. Kf2 Kd6 56. Ke2 Ra4 57. Kf3 Ke6 58. Rc5 Kd7 59. Rb5 Ke8 60. Rd5 g5 61.
Rb5 Kf8 62. Kg3 Ke8 63. Rf5 Ra3+ 64. Kg2 Ke7 65. Rd5 Ra2+ 66. Kf3 Ra3+ 67.
Kf2 Ra1 68. Rc5 Ra4 69. Kf3 Rf4+ 70. Kg3 Kd7 71. Kh3 Ra4 72. Rb5 Ke8 73.
Kg3 Kd8 74. Kf3 Ke8 75. Kg3
Result :1/2-1/2 {Draw by repetition}
Warning level :3
Final position: 4k3/5p2/5p2/PR4p1/r5P1/6K1/8/8 b - - 29 75
Analyze result:* {no draw found} Claimed by??, no debugfile saved
=============================================================================
1 warnings/errors found
Best wishes,
Leo.

Hi Leo,
have you got my email about BB 0.9.7i before some weeks? BB 0.9.7i should be slightly stronger.
best wishes
Andreas

[Andreas Herrmann]

Geschrieben von:/Posted by: Andreas Herrmann at 13 April 2004 11:49:26:
Als Antwort auf:/In reply to: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 12 April 2004 21:36:14:


Score SOP 1 2 3 4 5 6 7 8 9
--------------------------------------------------------------------------------------
...
3: BlackBishop 0.9.7g 6.5 / 9 49.5 9b= 17w+ 28b+ 16b+ 4w+ 2w- 8b+ 7b+ 1w-
4: Booot 3.2 6.5 / 9 47.5 12b+ 18w+ 29b+ 2w+ 3b- 8w= 10b+ 13w+ 5b-
5: Delphil 0.7d 6.5 / 9 45.5 19w+ 6w- 12b= 1b+ 29b+ 11w- 18b+ 20b+ 4w+
...
7: Uragano 0.73g 6.0 / 9 45.5 20b+ 14w+ 2b- 24w+ 8b- 18w+ 11b+ 3w- 13b+
...
17: Neurosis 1.7b 4.5 / 9 42.5 10w= 3b- 32w= 23b+ 24b= 22w+ 16w+ 6b- 9w-
18: RDChess 3.23 4.0 / 9 47.0 23w+ 4b- 1w- 20b+ 14w+ 7b- 5w- 25b= 24w=
...
--------------------------------------------------------------------------------------

Not sure, if there were more Pascal engines in the participating list. But great to see, that there are more and more new engines also written in Pascal/Delphi like Booot and Delphil. In the next days i will upload also some versions from Holmes (including the strongest version 0.8.29e) for free on my new homepage. This version of Holmes is only 40 to 50 ELO weaker than my current BlackBishop versions.

best wishes
Andreas




wbholmes.de

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 13 April 2004 17:45:59:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Andreas Herrmann at 13 April 2004 11:16:14:

Hi Leo,
have you got my email about BB 0.9.7i before some weeks? BB 0.9.7i should be slightly stronger.
best wishes
Andreas

Hi Andreas,
Yes I have it here, I saw yesterday that I had not used the latest BB, I am sorry but next test it will play!
Best wishes,
Leo.


WBEC Ridderkerk homepage.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 13 April 2004 17:54:24:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Fabien Letouzey at 13 April 2004 10:36:35:

Next engines played with Polyglot:
Fuit 1.0 (lost 2 games on time but won the tourney anyway!!)
Piranha 0.5 (dont work with the latest "polyglot", played with ponder=off because of a ponder bug!!)
Fusch 1.09 (no problems)
Alfil 403.1 (worked only with /first(second)Protocolversion=1

Just to make sure,
*NEVER USE PONDERING WITH POLYGLOT*
It just does not work.
Incompatibility between xboard and UCI protocols, hard work to fix (not planned soon).
I should add an error message when pondering is used.
If pondering was not used, it might be a performance problem.
Let me know, I am working on raw performance at the moment.
Fabien.

What do you mean, the engines use the CPU for 100% when its the opponents move!?
What are they doing then?
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 13 April 2004 17:59:38:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 13 April 2004 17:54:24:

*NEVER USE PONDERING WITH POLYGLOT*
It just does not work.

What do you mean, the engines use the CPU for 100% when its the opponents move!?
What are they doing then?
Leo.

Yes I know; as far as the implementation is concerned, I believe the code is working.
BUT, engines ponder with incorrect clock information (hence the losses on time)
This is due to a limitation in both xboard and UCI protocols; see my thread about pondering above.
I could *estimate* the amount of time that should be removed, but it's not a substitute for the real clock info that xboard has (and does not send until the opponent replies).
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 13 April 2004 18:21:31:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Fabien Letouzey at 13 April 2004 17:59:38:

*NEVER USE PONDERING WITH POLYGLOT*
It just does not work.

What do you mean, the engines use the CPU for 100% when its the opponents move!?
What are they doing then?
Leo.

Yes I know; as far as the implementation is concerned, I believe the code is working.
BUT, engines ponder with incorrect clock information (hence the losses on time)
This is due to a limitation in both xboard and UCI protocols; see my thread about pondering above.
I could *estimate* the amount of time that should be removed, but it's not a substitute for the real clock info that xboard has (and does not send until the opponent replies).
Fabien.

About how much time is the 'difference'?
I see by far the most timeloses with Fruit, I think you (Fruit) have to little time left at the clock when he plays the latest (40th in my case) move before the time control?
Thats risky anyway, you should play the last move with 2 or better 3 seconds left on the clock, I think Fruit do it with 1 (or less) second on the clock?
I have for example not see any timelose from Shredder 7.04, thats another test which I will report later this evening.
Winboard is allways sendig the time left with each move so it is only the last move before time control, right?
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 13 April 2004 18:31:20:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 13 April 2004 18:21:31:

About how much time is the 'difference'?
I see by far the most timeloses with Fruit, I think you (Fruit) have to little time left at the clock when he plays the latest (40th in my case) move before the time control?
Thats risky anyway, you should play the last move with 2 or better 3 seconds left on the clock, I think Fruit do it with 1 (or less) second on the clock?
I have for example not see any timelose from Shredder 7.04, thats another test which I will report later this evening.

This is due to a limitation in both xboard and UCI protocols; see my thread about pondering above.
I could *estimate* the amount of time that should be removed, but it's not a substitute for the real clock info that xboard has (and does not send until the opponent replies).

Winboard is allways sendig the time left with each move so it is only the last move before time control, right?

The current PolyGlot uses the clock info that was sent last.
That means it does not take into account the last move played by the engine.
So the difference can be quite large (unbounded).
Yes you are right, it keeps only one second.
Keep in mind most users stick with blitz time control (even 1 0 to test engines!!!). I wanted to keep things simple so I decided for 1s always.
Unless network play is involved, this should be enough, right?
A possible explanation could be that PolyGlot eats too much CPU time, I now have a version that should be faster (Dann will make it available if it seems OK).
Was the test you are talking about done with ponder off?
Because of the way pondering works on UCI, I have to send clock info before WinBoard sends the updated value. So one move was not taken into account (the engine move before pondering)
WinBoard does send the updated info after engine moves, only after opponent moves. In other words, after the engine plays a move I don't know how much time is left.
Fabien.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 13 April 2004 18:34:50:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Fabien Letouzey at 13 April 2004 18:31:20:

WinBoard does send the updated info after engine moves, only after opponent moves. In other words, after the engine plays a move I don't know how much time is left.

Please read "WinBoard does *not* ...".
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 13 April 2004 18:48:21:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Fabien Letouzey at 13 April 2004 18:31:20:

About how much time is the 'difference'?
I see by far the most timeloses with Fruit, I think you (Fruit) have to little time left at the clock when he plays the latest (40th in my case) move before the time control?
Thats risky anyway, you should play the last move with 2 or better 3 seconds left on the clock, I think Fruit do it with 1 (or less) second on the clock?
I have for example not see any timelose from Shredder 7.04, thats another test which I will report later this evening.

This is due to a limitation in both xboard and UCI protocols; see my thread about pondering above.
I could *estimate* the amount of time that should be removed, but it's not a substitute for the real clock info that xboard has (and does not send until the opponent replies).

Winboard is allways sendig the time left with each move so it is only the last move before time control, right?

The current PolyGlot uses the clock info that was sent last.
That means it does not take into account the last move played by the engine.
So the difference can be quite large (unbounded).
Yes you are right, it keeps only one second.
Keep in mind most users stick with blitz time control (even 1 0 to test engines!!!). I wanted to keep things simple so I decided for 1s always.
Unless network play is involved, this should be enough, right?
A possible explanation could be that PolyGlot eats too much CPU time, I now have a version that should be faster (Dann will make it available if it seems OK).
Was the test you are talking about done with ponder off?
Because of the way pondering works on UCI, I have to send clock info before WinBoard sends the updated value. So one move was not taken into account (the engine move before pondering)
WinBoard does send the updated info after engine moves, only after opponent moves. In other words, after the engine plays a move I don't know how much time is left.
Fabien.

1 seconds is to little, believe me sooner or later the engine(s) lose on time for one or the other reason! (and I can know it ) )
I have looked very carefull at the windows(2000Pro) taskmanager and polyglot did not use cpu time during the game, what I see is two polyglot.exe's loaded for each engine!?
No, I use always ponder=on (dual CPU)!
But I can set it of for individual engines, I did it with Fruit on the other test after 2 games but have not checked it yet for timeloses or other problems.
Why should you know that, winboard sends the time left for the rest of the game (or until next timecontrol) just before it sends the opponents move so you should know how much time is left!? (maby I miss something?)
This is what winboard do:
13828 second: time 239965
otim 239868
13828 >second: d4f5
13875 first : time 239868
otim 239960
13875 >first : d8f6
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 13 April 2004 18:55:30:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 13 April 2004 18:48:21:

Yes you are right, it keeps only one second.
Keep in mind most users stick with blitz time control (even 1 0 to test engines!!!). I wanted to keep things simple so I decided for 1s always.
Unless network play is involved, this should be enough, right?
A possible explanation could be that PolyGlot eats too much CPU time, I now have a version that should be faster (Dann will make it available if it seems OK).
Was the test you are talking about done with ponder off?
Because of the way pondering works on UCI, I have to send clock info before WinBoard sends the updated value. So one move was not taken into account (the engine move before pondering)
WinBoard does send the updated info after engine moves, only after opponent moves. In other words, after the engine plays a move I don't know how much time is left.

1 seconds is to little, believe me sooner or later the engine(s) lose on time for one or the other reason! (and I can know it ) )
I have looked very carefull at the windows(2000Pro) taskmanager and polyglot did not use cpu time during the game, what I see is two polyglot.exe's loaded for each engine!?
No, I use always ponder=on (dual CPU)!
But I can set it of for individual engines, I did it with Fruit on the other test after 2 games but have not checked it yet for timeloses or other problems.
Why should you know that, winboard sends the time left for the rest of the game (or until next timecontrol) just before it sends the opponents move so you should know how much time is left!? (maby I miss something?)

Yes, I will trust your experience.
I will think about making the "time reserve" dependent on time control.
Thanks!
Yes, this is due to how CYGWIN1.DLL emulates Unix multitasking.
One PolyGlot "clone" should be "dead" though, and not do anything.
Yes I think you are missing how UCI pondering works:
The engine must be sent correct clock information when *starting* pondering (just after the engine made its move). Also the information cannot be updated when the opponent replies the expected move!!! So they need to be correct first time. Unfortunately time info will be sent by WinBoard only after the opponent move.
Really unfortunate incompatibility between xboard and UCI!
Maybe you can post an extract of PolyGlot log, and I will explain things with an example.
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 13 April 2004 19:28:34:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Fabien Letouzey at 13 April 2004 18:55:30:

Yes you are right, it keeps only one second.
Keep in mind most users stick with blitz time control (even 1 0 to test engines!!!). I wanted to keep things simple so I decided for 1s always.
Unless network play is involved, this should be enough, right?
A possible explanation could be that PolyGlot eats too much CPU time, I now have a version that should be faster (Dann will make it available if it seems OK).
Was the test you are talking about done with ponder off?
Because of the way pondering works on UCI, I have to send clock info before WinBoard sends the updated value. So one move was not taken into account (the engine move before pondering)
WinBoard does send the updated info after engine moves, only after opponent moves. In other words, after the engine plays a move I don't know how much time is left.

1 seconds is to little, believe me sooner or later the engine(s) lose on time for one or the other reason! (and I can know it ) )
I have looked very carefull at the windows(2000Pro) taskmanager and polyglot did not use cpu time during the game, what I see is two polyglot.exe's loaded for each engine!?
No, I use always ponder=on (dual CPU)!
But I can set it of for individual engines, I did it with Fruit on the other test after 2 games but have not checked it yet for timeloses or other problems.
Why should you know that, winboard sends the time left for the rest of the game (or until next timecontrol) just before it sends the opponents move so you should know how much time is left!? (maby I miss something?)

Yes, I will trust your experience.
I will think about making the "time reserve" dependent on time control.
Thanks!
Yes, this is due to how CYGWIN1.DLL emulates Unix multitasking.
One PolyGlot "clone" should be "dead" though, and not do anything.
Yes I think you are missing how UCI pondering works:
The engine must be sent correct clock information when *starting* pondering (just after the engine made its move). Also the information cannot be updated when the opponent replies the expected move!!! So they need to be correct first time. Unfortunately time info will be sent by WinBoard only after the opponent move.
Really unfortunate incompatibility between xboard and UCI!
Maybe you can post an extract of PolyGlot log, and I will explain things with an example.
Fabien.

Will follow later!
For now, what means this:
"parse_bestmove(): missing argument" its from List/polyglot, whats the error here?
I just checked the other test tourney, 5 timeloses found, 1 by Fruit and 4 by List 5.12, none for SOS 4 and Shredder 7.04. (7 games each, Fruit played 5 games with ponder off without timeloses)
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 14 April 2004 10:29:04:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 13 April 2004 19:28:34:

For now, what means this:
"parse_bestmove(): missing argument" its from List/polyglot, whats the error here?
I just checked the other test tourney, 5 timeloses found, 1 by Fruit and 4 by List 5.12, none for SOS 4 and Shredder 7.04. (7 games each, Fruit played 5 games with ponder off without timeloses)

Bug in List, syntax error (it sends "bestmove xxx ponder" instead of "bestmove xxx").
Unfortunately it occurs in almost every game (usually near the end).
I don't feel like writing a work around; ask the author to fix it.
Don't use pondering for now. I'm going to write a "crippled pondering" as described in a thread above; it won't have time-out problems.
Hopefully it will be ready this (european) evening.
Please only report losses on time with ponder off, I told you pondering does not work at all!!! Any loss on time?
Fabien.

[Andreas Herrmann]

Geschrieben von:/Posted by: Andreas Herrmann at 14 April 2004 16:16:52:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 13 April 2004 17:45:59:

Hi Leo,
have you got my email about BB 0.9.7i before some weeks? BB 0.9.7i should be slightly stronger.
best wishes
Andreas

Hi Andreas,
Yes I have it here, I saw yesterday that I had not used the latest BB, I am sorry but next test it will play!
Best wishes,
Leo.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 14 April 2004 18:10:47:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Fabien Letouzey at 14 April 2004 10:29:04:

For now, what means this:
"parse_bestmove(): missing argument" its from List/polyglot, whats the error here?
I just checked the other test tourney, 5 timeloses found, 1 by Fruit and 4 by List 5.12, none for SOS 4 and Shredder 7.04. (7 games each, Fruit played 5 games with ponder off without timeloses)

Bug in List, syntax error (it sends "bestmove xxx ponder" instead of "bestmove xxx").
Unfortunately it occurs in almost every game (usually near the end).
I don't feel like writing a work around; ask the author to fix it.
Don't use pondering for now. I'm going to write a "crippled pondering" as described in a thread above; it won't have time-out problems.
Hopefully it will be ready this (european) evening.
Please only report losses on time with ponder off, I told you pondering does not work at all!!! Any loss on time?
Fabien.

This means if I can remove "ponder" from that line its fixed, right?
Again, I use _always_ ponder=ON.
Only engines with "ponder" problems are set to ponder=OFF, in case of Fruit I have played 5 games with ponder=off and no timeloses!
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 14 April 2004 18:17:30:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 14 April 2004 18:10:47:

Again, I use _always_ ponder=ON.
Only engines with "ponder" problems are set to ponder=OFF, in case of Fruit I have played 5 games with ponder=off and no timeloses!
Leo.

OK I will say it one last time, then I give up.
About PolyGlot (all versions so far)+ pondering:
1) It does not work at all
2) never use it with any engine
3) even if it "looks OK", engines use really broken clock information (info from one move too early)
Remember I don't rely on looks, I know exactly what's going on.
I think UCI2WB has exactly the same problem (maybe never got noticed), log files would prove this easily.
I have written a work around (which I call crippled pondering), it should be available in a few hours.
Fabien.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 14 April 2004 18:20:14:
Als Antwort auf:/In reply to: Re: Test tourney result. geschrieben von:/posted by: Leo Dijksman at 14 April 2004 18:10:47:

For now, what means this:
"parse_bestmove(): missing argument" its from List/polyglot, whats the error here?

Bug in List, syntax error (it sends "bestmove xxx ponder" instead of "bestmove xxx").

This means if I can remove "ponder" from that line its fixed, right?

Yes, but don't remove it from lines with "bestmove xxx ponder yyy" which are necessary for pondering.
Just out of curiosity, how are you going to do it?
Fabien.