Author: Ed Schröder
Date: 04:48:34 09/05/99
Yesterdays game was a disaster. From the very first move the program was handicapped by a not optimal running PC. The first sign was a not working LIVE support. Instead of showing the Rebel-Hoffman game the LIVE support showed an old logfile that was used to test the LIVE support. When Rebel was out of book on move (after 15..Rb7) the machine immediately crashed. The PC in this game crashed in total 10 times resulting in an enormous time loss on the clock and the loss of the permanent brain 10 times too. The most worse error in the game was on move 40. Rebel played 40. Rg4 showing a score of +2.30 in favor of Rebel. This is of course total nonsense as GM Hoffman had a clear advantage in that stage of the game. I then changed to the version that played GM Sorín and GM Rohde that had run rock solid. Of course it was already to late as Rebel already was in trouble. Surprisingly this version played the game till the end without any crash which made me conclude Rebel was to blame for the crashes. This was a premature conclusion. Testing the machine overnight proved that the Sorín / Rohde version also crashed. This morning I started to run a special Rebel version that checks the reliability of a PC. It's done via the "Analyze EPD" function which (also) displays the number of nodes searched in the EPD logfile. An EPD file was made containing the 40.Rg4 position 500 times, the PLY=10 instruction of "Analyze EPD" ensures that all of the position must have the same number of positions searched. The test is running now on 8 PC's, 4 x Pentium-II 266, 2 x Pentium-II 450, 1 x AMD-333 and the KryoTech AMD-600 machine. Sofar (a few hours) the test shows that all PC's are okay except the KryoTech. The outcome of the 40.Rg4 position should be: 40.b3 (and not Rg4) and 34.606.181 nodes searched. On all machines this is the case expect the Kryotech machine which a) displays different numbers of nodes from time to time or b) worse simply crashes. The sad news about this is that the Rebel-Hoffman game isn't reliable to count it for the original goal of the GM challenge which is to discover the current state of art of the micro's (Rebel in this case) against grandmasters in their strongest area, playing on tournament time control. Yesterdays game was simply a lost chance and month. Of course the game result (0-1) remains this includes the financial consequences as well. I will run the test for 24 hours as after the nightmare of yesterday I tend to mistrust trust every PC now. Next month Rebel will run on another PC during the GM challenge as I just must have a proper working PC on which I can rely or otherwise this whole GM challenge makes no sense. Below is the game in PGN. Ed Schroder ------------------------------------------------ [Event "Monthly GM Challenge"] [Site "Internet Game 40/2:00"] [Date "1999.09.04"] [Round "?"] [White "Rebel Century"] [Black "GM A. Hoffman"] [Result "0-1"] [WhiteElo "?"] [BlackElo "?"] [ECO "A58"] 1. d4 Nf6 2. c4 c5 3. d5 b5 4. cxb5 a6 5. bxa6 g6 6. Nc3 Bxa6 7. Nf3 d6 8. e4 Bxf1 9. Kxf1 Bg7 10. g3 O-O 11. Kg2 Nbd7 12. Re1 Qb6 13. h3 Rfb8 14. Re2 Ne8 15. Bf4 Rb7 16. Qc1 Nc7 17. Bh6 Bh8 18. a4 Qa6 19. Ra2 Rab8 20. Re1 Rb4 21. Qc2 Ne5 22. Nxe5 Bxe5 23. f4 Bd4 24. Bg5 f6 25. Bh4 Qc4 26. Re2 Rb3 27. Nd1 Rxg3+ 28. Bxg3 Qxa2 29. Bf2 Bxf2 30. Rxf2 Rb4 31. Nc3 Qc4 32. f5 Qd4 33. fxg6 hxg6 34. Qe2 Kf7 35. Kh1 Rb8 36. Rg2 Rh8 37. Qg4 g5 38. Qf3 Qe5 39. Rg1 Na6 40. Rg4 c4 41. Kg2 Nc5 42. Qf5 Qd4 43. Kh2 Qe3 44. Rg2 Nd3 45. Nd1 Qc1 46. Nc3 Ne5 47. Ne2 Qxb2 48. a5 Qa3 49. Qe6+ Kf8 50. Nd4 Qc3 51. Qc8+ Kg7 52. Nf5+ Kh7 53. Qb7 Nf3+ 54. Kg3 Nh4+ 55. Kh2 Qe5+ 56. Rg3 Nxf5 57. exf5 Kg7 58. Qa7 c3 59. Qe3 Rc8 60. Qxe5 dxe5 61. Rg1 Ra8 62. Ra1 c2 63. Rc1 Rxa5 64. Rxc2 Rxd5 0-1
This page took 0.01 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.