cracraft@wheaties.ai.mit.edu (Stuart Cracraft) (01/22/90)
The eleventh match game between GNU Sparc and Mach 3... After assaying a Reti gambit accepted, in which Mach 3 deviates from book on move 3, GNU finds itself a pawn down and its book no help. After gaining a rook and pawn for two minor pieces, it enters a difficult middlegame, but eventually manages to break into the endgame approximately equal. Then on move 43. GNU plays the excellent g5! a strong breakthrough move using the square-of-the-pawn idea. The win/mopup after that is quick. GNU Sparc vs Mach3 (1-0) 1. g1f3,d7d5 2 . c2c4,d5c4 3 . e2e3,b7b5 4 . b1c3,c7c6 5 . f3e5,g8f6 6 . d2d4,c4d3 7 . f1d3,e7e6 8 . d1f3,d8d6 9 . d3b5,c6b5 10. e5f7,e8f7 11. f3a8,d6b6 12. a2a4,f8c5 13. a4a5,b6b7 14. a8b7,c8b7 15. e1g1,h8d8 16. f1d1,d8d1 17. c3d1,b8a6 18. d1c3,b7c6 19. c3e2,a6b4 20. a1a3,c6d5 21. e2c3,b4c2 22. c3d5,c5a3 23. d5f6,a3b2 24. c1b2,g7f6 25. g1f1,a7a6 26. f1e2,c2b4 27. b2d4,b4c6 28. d4b6,f7e7 29. f2f4,e7d6 30. e3e4,e6e5 31. f4f5,c6b8 32. b6d8,b8d7 33. g2g4,b5b4 34. h2h4,h7h6 35. e2d3,d7c5 36. d3c4,c5e4 37. d8b6,e4g3 38. h4h5,g3f1 39. b6c5,d6d7 40. c5g1,d7c6 41. c4b4,c6d5 42. b4c3,d5e4 43. g4g5,h6g5 44. h5h6,f1g3 45. g1b6,e4f5 46. h6h7,g3e4 47. c3d3,e4d6 48. h7h8,e5e4 49. d3e2,f5e6 50. h8g8,e6d7 51. g8d5,f6f5 52. b6c5,d7e7 53. d5f5,g5g4 54. f5g4,e7d8 55. c5d6,d8e8 56. g4e6,e8d8 57. e6e7,d8c8 58. e7c7 The twelfth match game. In a non-optimal line of the Gruenfeld, Mach 3 commits tactical oversights. GNU returns the favor. After the chaos settles, the counterplay and material are in GNU's favor. Mach 3 nearly succeeds in trapping Black's queen, but then falls victim to Hans Berliner's horizion effect and succumbs in a mid-board mate. This game has fireworks! Mach3 vs GNU Sparc (0-1) 1. d2d4,g8f6 2 . c2c4,g7g6 3 . b1c3,d7d5 4 . g1f3,f8g7 5 . c1f4,e8g8 6 . a1c1,d5c4 7 . e2e3,c8e6 8 . f3g5,f6d5 9 . c3d5,e6d5 10. e3e4,h7h6 11. e4d5,h6g5 12. f4e5,d8d5 13. e5g7,g8g7 14. h2h4,f8h8 15. c1c4,b8c6 16. h1h3,g5h4 17. h3c3,a8d8 18. d1c1,d5a5 19. c4c5,a5b6 20. c3b3,c6b4 21. c1c4,a7a5 22. a2a3,h4h3 23. b3f3,b4c2 24. c4c2,h3h2 25. c2c4,f7f5 26. c4e2,b6d6 27. c5d5,d6d5 28. e2e7,g7g8 29. f3g3,h8h6 30. g3g6,h6g6 31. e7h4,d5d4 32. f1c4,g8g7 33. h4d4,d8d4 34. e1e2,h2h1 35. e2e3,h1d1 36. c4e2,d4e4 37. e3f3,d1e2 (Machine settings: 60/15 time control, hashing/transposition enabled, think-on-opponent's time set.) Ben Sloss, a student at Stanford, has duplicated results previously reported here. Ben took GNU 1.55 and ran it on a Sparcstation in a 5-game match vs. Mach 3 with identical settings except time control was lengthened to one minute per move average (instead of 15 seconds). Ben's score was 4-1, which is somewhat better than the 7-3 result reported after in a previous message. With the above two games, plus Ben's report, the score for GNU 1.55 Sparc vs. Fidelity Mach 3 is 14-4. Assuming Mach 3's rating is 2215-2220, this works out to a 207-215 edge for GNU. Since most sides reduce this for machine/machine testing by a 0.75 factor, the rating, based on 18 games is between: 2215+(0.75*207) to 2220+(0.75*215) or USCF 2370 to 2381 I will leave statistical significance and confidence intervals for the mathematicians among you. Stuart