Author: Robert Hyatt
Date: 10:38:30 05/17/99
Go up one level in this thread
On May 17, 1999 at 04:35:43, Frederic Louguet wrote:
>Many people talk about DBjr and the project of a PCI DBjr card, but the fact is
>we don't know much about DBjr. So here is a little scoop and a few data about
>it. Three months ago, there was an important Computer Event in Paris, the FIHT.
>And on the IBM stand, it was possible to play against DB Junior via the internet
>! Which version of DBjr, I don't know. How many processors ? I don't know. In
>fact, all I know is that it was DB Jr. It was too good an opportunity to miss,
>so with a good friend, Ernest Bonnem, we did this : Ernest was manipulating on
>the stand, and I was at home sending him the moves by phone.
>
>We did 5 games, each at 10mn blitz. Manipulating was difficult and tiring for
>Ernest, because the DBjr interface was not very friendly, and there was no take
>back ! I played the first two games with Fritz 5.32, and the remaining three
>with my program Chess Wizard. The PC used was a PII overclocked at 504 Mhz, 256
>MB ram.
>
>The first game against Fritz was interrupted at the 11th move, because DBjr
>suddenly stopped to play, or more exactly, went into infinite time mode !
>Technical or transmission error, I don't know. Annotations are from Fritz.
>
what you played against wasn't 'deep blue junior' as we know it. The web-
based demo they run uses a single 'board' with DB processors, and is set to
search 1 second per move no matter how much time you use. Of that 1 second,
a good part is spent downloading the hardware evaluation.
No idea how strong it plays, but at less than a second per move, what you
are playing is obviously not a serious version of DB Junior at all... Their
'web-based' version is obviously much different from a real program, because
anything 'web-based' is by necessity 'stateless' meaning no continuity from
move to move since many can be playing it at one time. IE it will have no
idea about repetitions at all, other than what it can see in the current
position and the search it does from that position...
Just for the record...
Of course, beating the thing is still not easy. But the 'real' machine is so
much stronger..
Bob
>[Event "DBJr-Fritz Games"]
>[Site "FIHT/Louguet home"]
>[Date "1999.05.02"]
>[Round "1"]
>[White "Fritz 5.32"]
>[Black "DBJr"]
>[Result "*"]
>[ECO "B01"]
>[Annotator "Louguet,F"]
>[PlyCount "21"]
>
>{131072kB, General.ctg. Pentium II-504
>} 1. e4 d5 2. exd5 Qxd5 3. Nc3 Qe6+ 4.
>Be2 {0.16/12} 4... Nf6 5. Nf3 {0.38/11} 5... Bd7 6. d4 {0.41/11} 6... Qb6 7.
>O-O {0.25/10} 7... e6 8. Re1 {0.13/10} 8... Be7 9. Ne5 {0.25/11} 9... O-O 10.
>Nc4 {0.59/10} 10... Qc6 11. Qd3 {0.63/10} *
>
>
>The second game was an equal one. Fritz held its own quite well. It did not use
>tablebases in this game.
>
>[Event "DBJr-Fritz Games"]
>[Site "FIHT/Louguet home"]
>[Date "1999.05.02"]
>[Round "2"]
>[White "Fritz 5.32"]
>[Black "DBJr"]
>[Result "1/2-1/2"]
>[ECO "D02"]
>[Annotator "Louguet,F"]
>[PlyCount "133"]
>
>{131072kB, General.ctg. Pentium II-504
>} 1. d4 Nf6 2. c4 d5 3. Nf3 Bf5 4. cxd5
>{0.41/10} 4... Qxd5 5. Nc3 {0.41/10} 5... Qd6 6. Qb3 {0.34/11} 6... Qb6 7. Qxb6
>{0.28/11} 7... axb6 8. Bg5 {0.34/11} 8... Nbd7 9. Nd2 {0.41/12} 9... h6 10.
>Bxf6 {0.47/11} 10... exf6 11. e4 {0.44/11} 11... Bg6 12. Bd3 {0.47/11} 12...
>Bb4 13. Nd5 {0.31/11} 13... Bxd2+ 14. Kxd2 {0.44/12} 14... Kd8 15. Rac1 {
>0.28/10} 15... c6 16. Nc3 {0.31/11} 16... Ke7 17. Rhd1 {0.53/9} 17... Rhd8 18.
>a4 {0.63/11} 18... Kf8 19. Ke3 {0.56/9} 19... Re8 20. Kf4 {0.59/10} 20... Red8
>21. Bc2 {0.59/10} 21... Ra5 22. b4 {0.66/9} 22... Rh5 23. h3 {0.72/10} 23...
>Ke7 24. Rd3 {0.75/9} 24... Nf8 25. g4 {0.69/9} 25... Ne6+ 26. Kg3 {0.59/13}
>26... Rxh3+ 27. Kxh3 {0.75/14} 27... Nf4+ 28. Kg3 {0.72/12} 28... Nxd3 29. Bxd3
>{0.97/12} 29... Rxd4 30. Rd1 {0.59/10} 30... Rxb4 31. Kf4 {0.59/12} 31... Rb3
>32. Rc1 {0.53/12} 32... h5 33. gxh5 {0.56/11} 33... Bxh5 34. Ke3 {0.47/11}
>34... Bg4 35. Kd2 {0.56/10} 35... g6 36. Bc2 {0.69/10} 36... Rb4 37. Ke3 {
>0.59/10} 37... Rb2 38. f3 {0.53/9} 38... Be6 39. f4 {0.38/11} 39... Kd6 40. Bd1
>{0.38/10} 40... Rh2 41. Bf3 {0.41/11} 41... Rb2 42. Be2 {0.41/9} 42... Rb3 43.
>Bd3 {0.34/11} 43... Ra3 44. Kd2 {0.28/11} 44... Bb3 45. Bc2 {0.28/11} 45... Bc4
>46. Nd1 {0.31/11} 46... Bb3 47. Bxb3 {0.47/10} 47... Rxb3 48. Ne3 {0.50/11}
>48... Rb4 49. Rc4 {0.34/12} 49... Rb3 50. Ke2 {0.38/11} 50... Rb2+ 51. Kd3 {
>0.41/10} 51... Ke6 52. f5+ {0.44/9} 52... Ke7 53. Nd1 {0.31/10} 53... Rb1 54.
>Nc3 {0.44/10} 54... Rb2 55. Rd4 {0.38/12} 55... gxf5 56. exf5 {0.78/11} 56...
>Rf2 57. Rb4 {0.63/12} 57... b5 58. axb5 {0.78/10} 58... Rxf5 59. bxc6 {0.69/9}
>59... bxc6 60. Rc4 {0.72/10} 60... Rf1 61. Rxc6 {1.50/9} 61... Re1 62. Nd5+ {
>2.22/9} 62... Kd7 63. Rxf6 {2.22/9} 63... Rd1+ 64. Ke4 {2.38/12} 64... Re1+ 65.
>Kf5 {2.34/12} 65... Rd1 66. Rxf7+ {2.84/9} 66... Kd6 67. Nf4 {3.13/12} 1/2-1/2
>
>
>Then, I could not help playing DBjr with my program Chess Wizard, actual French
>computer chess champion. The version which played was the one which earned a
>2656 elo perf in the Aubervilliers Open one week earlier (an elo perf means
>nothing, okay, no need to remind me that :)). The first game was equal, and in
>the end DBjr repeated the position, clearly wanting a draw. Unfortunately,
>Wizard had a nasty bug in the repetition code which kicked in a the most
>inapropriate time. It had only about 30 seconds left at its clock, and did not
>accept the draw, and lost. Always funny. Sorry for not having all the moves, but
>the end of the log (about 10 moves) was eaten by a HD problem, and I did not
>discover it until recently.
>
>[Event "Wizard vs DBjr 10mn"]
>[Site "FIHT/Louguet Home"]
>[Date "1999.05.02"]
>[Round "1"]
>[White "Deep Blue Jr"]
>[WhiteElo ""]
>[Black "Chess Wizard"]
>[BlackElo ""]
>[Result "1-0"]
>
>1. Nf3 Nf6 2. e3 g6 3. d4 Bg7 4. Nbd2 O-O 5. e4 d6 6. Bd3 c5 7. c3 cxd4 8.
>cxd4 e5 9. d5 Na6 10. Bxa6 bxa6 11. O-O Bd7 12. Qb3 Rc8 13. Re1 Qb6 14.
>Qxb6 axb6 15. b3 Bb5 16. Ba3 Rfd8 17. Rac1 Bf8 18. h4 Rxc1 19. Rxc1 h6 20.
>Rc3 Ne8 21. g4 Bd7 22. g5 b5 23. Bb4 f6 24. Ba5 Rb8 25. b4 fxg5 26. hxg5 h5
>27. Kg2 Kf7 28. Kg1 Rc8 29. Rxc8 Bxc8 30. Kh2 Bg4 31. Kg3 Be7 32. Bb6 Bd7
>33. Kh4 Ng7 34. Ne1 Bg4 35. f3 Bc8 36. Nd3 Ne8 37. Nb3 Bf8 38. Na5 Kg7 39.
>Nc6 Kh7 40. f4 exf4 41. Nxf4 Bg4 42. Kg3 Ng7 { a few moves later DB repeated the
>position but Wizard lost, (a bug in the 3 rep.code when time left < 1 mn. Fixed
>since :) }
>
>
>The second game was dynamic and interesting, but my friend (who was beginning to
>feel very tired, which is understandable given the conditions) made a mistake in
>entering a move, and there was no take back for DB.
>
>[Event "Wizard vs DBjr 10mn"]
>[Site "FIHT/Louguet Home"]
>[Date "1999.05.02"]
>[Round "2"]
>[White "Chess Wizard"]
>[WhiteElo ""]
>[Black "Deep Blue Jr"]
>[BlackElo ""]
>[Result "*"]
>
>1. e4 d5 2. exd5 Qxd5 3. Nc3 Qa5 4. d4 Nf6 5. Nf3 Bf5 6. Bd2 Na6 7. Bc4 e6
>8. Nd5 Qa4 9. b3 Qd7 10. Ne5 Qd8 11. Bb5+ c6 12. Nxf6+ gxf6 13. Nxc6 Qb6
>14. Nxa7+ Ke7 15. c3 Rxa7 16. Bc4 Rg8 17. Qf3 h5 18. O-O Bg4 19. Qd3 Bf5
>20. Qf3 Bg4 21. Qd3 Nc7 22. Rfe1 {here mistake from the operator, and no
>take back (!) for DB, so could not continue }
>
>
>So we did a final game, which was won by Chess Wizard.
>
>[Event "Wizard vs DBjr 10mn"]
>[Site "FIHT/Louguet Home"]
>[Date "1999.05.02"]
>[Round "3"]
>[White "Chess Wizard"]
>[WhiteElo ""]
>[Black "Deep Blue Jr"]
>[BlackElo ""]
>[Result "1-0"]
>1. e4 Nc6 2. Nf3 e5 3. Bb5 Nf6 4. O-O Bc5 5. Nxe5 Nxe5 6. d4
>a6 7. Ba4 b5 8. Bb3 Bxd4 9. Qxd4 d6 10. f4 c5 11. Qd2 Nc6 12. e5
>dxe5 13. fxe5 Nxe5 14. Qxd8+ Kxd8 15. Bg5 c4 16. Rxf6 Kd7 17. Rf1
>cxb3 18. axb3 f6 19. Be3 Bb7 20. Nc3 Rhe8 21. Bc5 Kc6 22. b4
>Nc4 23. b3 Nb6 24. Rac1 Nd7 25. Bd4 Rad8 26. Na2 Re4 27. Bf2
>Ne5 28. Bc5 a5 29. Ra1 Rd2 30. Rf2 Ree2 31. Rc1 Kd7 32. bxa5
>Ng4 33. a6 Bc6 34. Rxe2 Rxe2 35. h3 Rxg2+ 36. Kf1 Nh2+ 37. Ke1
>Nf3+ 38. Kf1 Nd2+ 39. Ke1 Nxb3 40. Rd1+ Ke8 41. Bf2 Na5 42. Nb4
>Rh2 43. Nxc6 Nxc6 44. Rd6 Rxh3 45. Rxc6 Kd7 46. a7 Ra3 47. Rc3
>Ra2 48. Bb6 h5 49. Rc7+ Kd6 50. Rxg7 f5 51. Rg8 Ra1+ 52. Ke2
>Ra4 53. a8=Q Rxa8 54. Rxa8 h4 55. Rh8 h3 56. Rxh3 1-0 (actually
>went to mate }
>
>So, we wanted facts, here are facts. Why didn't we talk about this earlier ?
>Well, we (mostly Ernest in fact) tried to have a better contact with the DB Team
>and get access to an internet site where we could play DBjr more thoroughly.
>Unfortunately they do not seem interested in that. So all we have are these 5
>games. It is not much, but it is better than nothing. I don't draw any
>conclusion from these games, except that now, I am not really convinced that
>DBjr would kill every micro program at each game (and it was 10 minutes blitz,
>so not a very easy setting for the micro).
>
>But of course, maybe DBjr was doing all sorts of things at the other end of the
>phone line (playing Q3Test ? :)) and did not had all its computing power
>available...Well, all I can say is that I don't want to begin a new war between
>the pro-DBs and the anti-DBs. It was not a scientific experiment, but an
>opportunity to play a few games against something we don't really know about.
>Even if it proves nothing, it was fun to mate DBjr ('cause of course it did not
>resign).
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.