[Event "Quali97"] [Site "60/60"] [Date "1997.09.27"] [Round "1"] [Result "0-1"] [White "Dabbaba 260997"] [Black "Sibirien Chess 2.15"] 1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Bxc6 dxc6 {*} 5. Nxe5 Qd4 6. Ng4 Bxg4 7. Qxg4 Nf6 8. Qf3 Nxe4 9. c3 Qd5 10. O-O O-O-O 11. d4 Nf6 12. Qxd5 Rxd5 13. Be3 Rb5 14. b3 Rh5 15. Nd2 c5 16. dxc5 Bxc5 17. Bxc5 Rxc5 18. c4 Re8 19. Rae1 Rce5 20. Rxe5 Rxe5 21. a3 Re2 22. Nf3 Ne4 23. b4 Ra2 24. Re1 Nxf2 25. Re8+ Kd7 26. Re3 Ng4 27. Rd3+ Ke7 28. g3 f5 29. h3 Nf2 30. Re3+ Kf6 31. h4 Ne4 32. Nd4 c5 33. bxc5 Nxc5 34. Rf3 g6 35. Rc3 Ke5 36. Nf3+ Ke4 37. Ng5+ Kd4 38. Rf3 h5 39. Nf7 Nd3 40. Nh8 Ke4 41. Rf1 Ne5 42. Re1+ Kd4 43. Rd1+ Kxc4 44. Rc1+ Kb3 45. Rb1+ Rb2 46. Rxb2+ Kxb2 47. a4 Kc3 48. Kf2 b5 49. axb5 axb5 50. Nxg6 Nxg6 0-1 Two Pentium 120 Mhz was used. Dabbaba has no book and found 5.Nxe5, which is not the best. 21.a3? was the worst error. 21.Rd1 or 21.f4 were better. (While testing Dabbaba afterwards, it actually plays 21.Rd1. I have a nasty feeling, that Dabbaba plays different/worse in games compared with what it does in testpositions...) Notice: the following game was replaced by another game as I meanwhile sent a new version of Dabbaba to Torsten Schoop. ich habe mit Dabbaba die 2. Runde gespielt. Dabbaba musste gegen Schach! 1.0 (siehe CSS 6/96). Gegen diesen taktisch starken Gegner hatte Dabbaba keine Chance: [Event "Quali97"] [Site "60/60"] [Date "1997.05.10"] [Round "2"] [White "Schach! 1.0"] [Black "Dabbaba 0997"] [Result "1-0"] [ECO "B04"] [PlyCount "49"] 1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. Nf3 dxe5 5. Nxe5 g6 6. Bc4 {*} Bg7 7. Nxf7 Kxf7 8. Qf3+ Ke8 9. Bxd5 Rf8 10. Qb3 c6 11. Bf3 Qxd4 12. O-O Qd8 13. Be3 Nd7 14. Nc3 Ne5 15. Be2 b6 16. Rad1 Bd7 17. f4 Nf7 18. Rd2 Bxc3 19. Qxc3 Rc8 20. Qg7 Rh8 21. Bc4 Rf8 22. Qxh7 Nh8 23. Rfd1 Qc7 24. Rxd7 Qxd7 25. Rxd7 1-0 >ich habe die Version 0997 durch die Version 1097 ersetzt. Da die 2. Runde >noch nicht abgeschlossen war, habe ich die Parite von Dabbaba gegen Schach! >mit der neuen Version noch einmal gespielt. Diesmal entstand eine As I wrote you could have continued the tournament with the old version; but even redoing game 2 was more than I expected. But the changes in the new version are very small, so it will probably make the same bad moves in the first game against Schach! (se comments at the end). >geschlossene Stellung. Schach! spielte den fragwuerdigen Zug 22. Qxd6?!. >Schach! will diesen Zug bereits noch 4 Sekunden spielen (Score: +0.56). Er >spielt diesen Zug dann bei einer Rechentiefe von 6/25 nach ca. 1 Minute mit >einem Score von +0.30. Ich habe Schach! auf unendlich an dieser Stellung >rechnen lassen. Schach! bleibt auch nach 15 Minuten bei diesem Zug. Aus den >Entstehenden Verwicklungen kommt Dabbaba mit Springer gegen 2 Bauern gut >heraus. Dabbaba vergibt dann allerdings seinen Vorteil wieder. Am Ende >erkennt Dabbaba nicht das Remis nach der 3-Zug-Regel. As I wrote to you in september: "If Dabbaba is lucky to get into a winning position in its games, it may walk into a draw by repetition of moves, as this is not implemented yet..." But I doubt the ending position is a win for black; perhaps not even without the 2 white pawns. See this analysis by Genius2 on my old 486 50Mhz machine (3min/move): 20 Nh4-g6 0:53:19 0.36 a2-a4 b5-b4 h4-g6 f8-e8 f2-f4 20.. a5-a4 0:53:21 0.42 a5-a4 g2xc6 c8xc6 b3-d5 f8-c8 21 Qb3-d5 0:55:25 -0.60 g2xc6 c8xc6 b3-d5 f8-c8 g6-h4 21.. Rf8-e8 0:55:20 0.51 h7-f6 d5xd6 f7xg6 e1xe5 b7-d7 22 Qd5xd6 0:57:57 -0.30 g6-h4 e8-d8 d5-e4 h7-f6 e4-e3 22.. f7xg6 0:56:46 0.48 f7xg6 f5xg6 h7-f8 d6xc5 f8xg6 23 f5xg6 1:00:47 -0.39 f5xg6 h7-f8 d6-d5+g8-h8 d5xc5 23.. Nh7-f8 0:58:56 0.51 h7-f8 d6-d5+g8-h8 d5xc5 b7-a6 24 Qd6-d5+ 1:07:30 -1.06 d6xc5 f8xg6 g2-e4 g6-f4 e1-e3 It is 20.Ng6 that starts to make it bad for white - 22.Qxd6 is almost ok. Chess is often a matter of NOT going into something, that later turns out bad (the proposed 24.d6xc5 suddenly drops to -1.06) . And Schach! was just unlucky that it happened to it in this game, but also lucky that Dabbaba was not able to turn its advantage of up to 2,89 (see attached analysis by Genius 30sek/move for the whole game) into a win. Genius took 55 sec. to switch to 20.-,a4, so I assume matters are more difficult than they seems... [Event "Quali97"] [Site "60/60"] [Date "1997.10.08"] [Round "2"] [White "Schach! 1.0"] [Black "Dabbaba 1097"] [Result "1/2-1/2"] 1. e4 e5 2. Nf3 Nf6 3. d3 d6 4. g3 {*} Bg4 5. Nbd2 Be7 6. h3 Bd7 7. Bg2 O-O 8. O-O c5 9. Nc4 Nc6 10. Bg5 Be6 11. Re1 Rc8 12. c3 b5 13. Ne3 Qd7 14. g4 h6 15. Bh4 a5 16. Nf5 Bxf5 17. exf5 Nh7 18. Qb3 Bxh4 19. Nxh4 Qb7 20. Ng6 a4 21. Qd5 Rfe8 22. Qxd6 fxg6 23. fxg6 Nf8 24. Qd5+ Kh8 25. Rxe5 Rxe5 26. Qxe5 26... Nxe5 27. Bxb7 Rd8 28. d4 cxd4 29. Ba6 Nf3+ 30. Kg2 Nh4+ 31. Kh2 dxc3 32. bxc3 Rd2 33. Rf1 Rb2 34. Kg3 Nhxg6 35. f4 Nd7 36. f5 Ngf8 37. Rf2 a3 38. Bb7 Rb1 39. Re2 Rd1 40. Bc6 Rd3+ 41. Kg2 Rxc3 42. Bxb5 Nf6 43. Ba4 g6 44. Re5 Rd3 45. Bb3 N8d7 46. Ra5 gxf5 47. gxf5 Nh5 48. h4 Rd2+ 49. Kh3 Nf4+ 50. Kg4 Ng2 51. Kg3 Ne1 52. Rxa3 Kg7 53. Ra7 Kf6 54. Ra6+ Kxf5 55. Be6+ Ke5 56. Bxd7 Rxd7 57. Rxh6 Rd3+ 58. Kf2 Nc2 59. Ra6 Rd2+ 60. Kf3 Ne1+ 61. Kg4 Rd4+ 62. Kg3 Rd3+ 63. Kg4 Rd4+ 64. Kg3 Rd3+ 65. Kg4 Rd4+ 66. Kg3 in der 3. Runde gegen Nova Chess 1.2 hatte Dabbaba wieder eine Figur mehr, wie schon gegen Schach!, machte aber wieder Remis durch Zugwiederholung: [Event "Quali97"] [Site "60Z/1h"] [Date "1997.10.14"] [Date "1997.10.15"] [Round "3"] [White "Dabbaba 151097"] [Black "Nova 1.2"] [Result "1/2-1/2"] [ECO "B02"] [PlyCount "171"] 1. e4 Nf6 2. e5 Nd5 3. c4 Nb6 4. d4 d6 {*} 5. Nf3 dxe5 6. Nxe5 e6 7. c5 N6d7 8. Bf4 Nxe5 9. Bxe5 Nd7 10. Qa4 c6 11. Bg3 Be7 12. Be2 O-O 13. O-O Nf6 14. Nc3 Nd5 15. Qb3 f5 16. Be5 Nxc3 17. bxc3 f4 18. Bc4 Kh8 19. Bxe6 f3 20. Bxc8 Qxc8 21. Rfb1 Qg4 22. Bg3 b5 23. cxb6 axb6 24. Qxb6 fxg2 25. Qxc6 Rac8 26. Qb7 Bh4 27. Qb2 Rf3 28. Re1 Qh3 29. Qe2 Rf5 30. Qe6 Rcf8 31. Bxh4 h5 32. Be7 Qf3 33. Re2 Qxc3 34. Bxf8 Qxa1+ 35. Re1 Qc3 36. Qe8 Qf3 37. Be7+ Kh7 38. Bh4 Rf4 39. Qd8 Qg4 40. Bg3 Rxd4 41. Qf8 h4 42. f3 Qg5 43. f4 Qf6 44. Qxf6 gxf6 45. Bxh4 Rxf4 46. Bf2 Rg4 47. Bg3 f5 48. Re7+ Kg6 49. Kxg2 Ra4 50. Re6+ Kf7 51. Re2 Kf6 52. Be5+ Kg5 53. Kf3 Ra3+ 54. Kf2 Kg4 55. Ke1 f4 56. Bd4 Ra4 57. Rg2+ Kh3 58. Rd2 Kg4 59. Kf2 Ra5 60. Rc2 Rd5 61. Rc4 Rh5 62. Kg1 Kf3 63. a4 Ke2 64. Bf6 Rf5 65. Bh4 Kd3 66. Rb4 Ke2 67. h3 f3 68. Rb2+ Kd3 69. Bg3 Kc3 70. Rb7 Kd4 71. Kf2 Kc4 72. Re7 Kb3 73. Re4 Rf7 74. a5 Ra7 75. Re5 Rf7 76. a6 Kb4 77. Bh4 Ra7 78. Re8 Rxa6 79. Rb8+ Kc4 80. Rc8+ Kd5 81. Rd8+ Ke4 82. Re8+ Kf5 83. Rf8+ Ke4 84. Re8+ Kf5 85. Rf8+ Ke4 86. Re8+ 1/2-1/2 Instead of 77.Bh4? Dabbaba could have won with 77.Re6 followed by Bb8 and a6-a7. It was silly that Dabbaba could not win that ending. I must avoid that draw by repetition some day. I was most disappointed about 7.c4-c5? as I just had improved Dabbaba on hanging pawns in the new version (with improved Q-play in the opening) that I sent you. Dabbaba does not play 7.c4-c5 in a testposition, but as Dabbaba plays a little random in the opening, it is difficult to say anything for sure... I don't know Nova at all, but I assume it is at least as weak as Dabbaba. It lost a bishop by 29.-,Rf3-f5??, and even on my old 486 it only took Dabbaba 14,29 sec. to see that 30.Qe2-e6! wins a piece - the score raised to 4,66 pawn. It must have seen the line 30.Qe6,Rcf8 31.Bxh4,Qxh4 32.Qxf5!,Rxf5 33.Re8+,Rf8 33.Rxf8+ mate. Nova must have missed that... But Nova also played many other weak moves (fx 15.-,f7-f5?). >in der 4. Runde war Green Light Chess 2.04g zu stark fuer Dabbaba. Green >Light Chess spielt wie Dabbaba ohne Eroeffnungsbuch. Thank you for the game. I return it with a few comments below. I feared Green Light C was strong, as it won in round 3 and had 1/1 compared to Dabbabas 1/3. But it showed some weak moves in the game, so I feel that Dabbaba on a better day would have a chance against it. I have let Dabbaba make a simple trace during a game, and in a certain position in a selfplaygame (that I can redo any time) the trace and move played is different compared to the corresponding testposition: r2qk1nr/ppp2ppp/2n5/3pPb2/1b6/5NP1/PPPBPP1P/RN1QKB1R/w e5-e6? -0.00 sec. Search to depth 1 (a testposition) 0.05 sec. best move so far: d2c3 score: 108 0.16 sec. best move so far: d2b4 score: 109 0.22 sec. best move so far: h1g1 score: 133 1.48 sec. best move so far: b1c3 score: 135 2.25 sec. Search to depth 2 3.41 sec. best move so far: b1c3 score: 72 6.59 sec. best move so far: f1g2 score: 98 9.07 sec. Search to depth 3 30.99 sec. best move so far: f1g2 score: 116 60.99 sec. Search to depth 4 113.02 sec. best move so far: f1g2 score: 87 256.54 sec. Search to depth 5 256.65 sec. calculation of move ended 331.87 sec. End of Disc-trace 0.00 sec. Search to depth 1 (this is from a selfplaygame) 0.05 sec. best move so far: d2c3 score: 56 0.27 sec. best move so far: d2b4 score: 57 0.77 sec. best move so far: h1g1 score: 81 0.82 sec. best move so far: f1g2 score: 110 1.04 sec. Search to depth 2 3.52 sec. best move so far: f1g2 score: 54 10.49 sec. Search to depth 3 21.21 sec. best move so far: d2c3 score: 49 33.74 sec. best move so far: b1a3 score: 51 47.64 sec. best move so far: f1g2 score: 74 88.96 sec. best move so far: e5e6 score: 76 <----- 107.14 sec. calculation of move ended Dabbaba may give many worries, but this is probably the worst... [Event "Quali97"] [Site "60/60"] [Date "1997.11.01"] [Round "4"] [White "Dabbaba 151097"] [Black "Green Light Chess V2.04g"] [Result "0-1"] 1. e3. {a seldom opening by Dabbaba} 1.-,e5 2. d4 {Dabbaba might by its random openingplay get a strange idea here and play e3-e4...} 2.-,exd4 3. exd4 Nf6 4. Nf3 Bd6? 5. Bd2(?) O-O 6. Bd3(?) c5? {Why not -,Re8+?} 7. O-O cxd4 8. Nxd4 Be5 9. Be3 Nc6 10. c3 d5 11. Nxc6? {helps black to get rid of his isolated d-pawn} 11.-,bxc6 12. f4 Ng4 13. Bc5. {Genius2 finds Bxh7+ a little better} 13.-,Bd6 14. Bxd6 Qxd6 15. Qe2 Qc5+ 16. Kh1 Bd7 17. b4 Qe3 18. Nd2?? {Makes the Geniusscore drop a pawn. Better were 18.Qxe3} 18.-,Rae8 19. Rae1?? {Makes the Geniusscore drop another pawn down to - 2,30. Better were still 19.Qxe3. Dabbaba gets no chance now.} 19.-,Qb6! 20. Qd1 Nf2+ 21. Rxf2 Qxf2 22. Rf1 Qe3 23. Qc2 g6 24. f5 Bxf5 25. Bxf5 gxf5 26. Rxf5 Qe1+ 27. Nf1 Re2 28. Qd3 Rxa2 29. Rh5 Qe4 30. Qxe4 dxe4 31. Rg5+ Kh8 32. Re5 f5 33. Kg1 Ra3 34. Rc5 f4 35. Rxc6 Re8 36. c4 e3 * More comments on move 18 and 19 later... ------------- Dabbaba now produces a simple logfile GAMEINFO.TRC during a game. With this I have tried to find out why it plays worse in games compared to testpositions. Some fields in the program are now initialized before each move instead of only doing it in the beginning of the game, and it seems to help on the game/positions I have used to test - I just don't fully understand it... Dabbaba seems to play more 'clean' in the few testgames I have seen so far. Other changes: I have reduced the depth from 3 to 1 where Dabbaba tries to get quiet positions. This makes the iterations go faster, as they are not so deep now. If there is only one move to do, Dabbaba will now play it fast. Dabbaba should now be more aware of putting rooks on good squares. 2 errors found and corrected. I have had a look on hanging pawns. They get so much score, that Dabbaba nearly offers a pawn (1.d4, e5?) to give the opponent a hanging pawn. After 2.dxe5 the pawn on e5 is so far ahead, that it is considered 'hanging'. I hope everything will be fine in the version you get! Compared to a month ago, Dabbaba solves a few more of the testpositions and now only uses only 80% of the time - probably due to the reduced quiet depth. ------------- Hi Torsten! You should have received Dabbaba by now. I think it only SEEMS that the bad moves cannot be reproduced. The good moves are seldom tried to be reproduced, so you seldom discover it cannot be done... Dabbaba has no hashtables or anything calculated from the previous move(s), so its moves should could be reproduced... I have just received PLY today. Thoralf Karlsson mentions a blixttournament you have done with 7 programs. I have just written to him about something else, and I have also offered him information about your current tournament. But perhaps you want to give him that yourself? Greetings Jens ps. I have just learned Dabbaba that a king that is alone should be forced to the corner, if you have no pawns. So now it can mate with KBB<>K in 10s/m; before that it could not with 3m/m. KBN<>K is still to difficult even with 3m/m... ------------ [Event "Quali97"] [Site "60/60"] [Date "1997.11.06"] [Round "5"] [White "Dragon 1.0"] [Black "Dabbaba 051197"] [Result "1-0"] 1.e4 c5 2.Nf3 Nf6 3.e5 {*} Ng4 4.h3 Nh6 5.d4 d4 6.Bh6 h6 7.Qd4 Nc6 8.Qe3 d5 9.Bb5 Bd7 10.Nc3 a6 11.Bc6 Bc6 12.O-O Bg7 13.Rfe1 Qd7 14.Rad1 O-O-O 15.Qc5 Kb8 16.Nd5 Bd5 17.c4 Qc6 18.Rd5 Qc5 19.Rc5 e6 20.Ra5 Rd3 21.c5 Rg8 22.b4 Bh8 23.Kh2 Rc8 24.Re4 Rcd8 25.Rh4 Bg7 26.c6 Bf8 27.Rf4 f6 28.b7 Kb7 29.f6 Bd6 30.Ne5 Bc7 31.f7 Ra3 32.f8Q Rf8 33.Rf8 Ra5 34.a5 Be5+ 35.Kg1 Bc3 36.Rf7+ Kc6 37.Rh7 Ba5 38.Rh6 Kd6 39.h4 Bc3 40.h5 Kd5 41.Rg6 Bd2 42.h6 a5 43.h7 Bc3 44.Rg5+ Kc6 45.Ra5 Once again Dabbaba was lucky to survive a bad opening (Genius2 only gives white 0.12 on move 8). The critical part of the game: 14.-,0-0-0? allows Nc3xd5! 15.Qc5? Dragon misses it... 15.-,Kb8? with e7-e6 Dabbaba could still prevent Nc3xd5! 16.Nc3xd5! Now Dragon got it! 16.-,Bxd5 Stupid Dabbaba still thinks it is 1,79 pawn ahead... 17.c4,Qc6?! See below. With e7-e6 or Genius' Bxf3!? black still had a chance... You were right; the logfile is interesting although it still is quite simple. Look here at Dabbaba's move 17: 0.00 sec. Search to depth 1 0.11 sec. best move so far: d5c6 score: 291 0.16 sec. best move so far: d5e6 score: 242 0.33 sec. best move so far: d5f3 score: 195 0.38 sec. best move so far: e7e6 score: 86 0.60 sec. best move so far: d7c6 score: -180 0.66 sec. Search to depth 2 0.93 sec. best move so far: d7c6 score: 117 2.14 sec. Search to depth 3 2.86 sec. calculation of move ended It moved just after 3 seconds - did you notice that? I assume that after you typed Dragons move c2c4 by accident touched the keyboard, which makes Dabbaba move immediately (but never before its minimumsearch of 2 iterations). I have tried the position on my own PC: 1k1r3r/1p1qppbp/p6p/2QbP3/2P5/5N1P/PP3PP1/3RR1K1/b Dab moved in 3 sec. Dragon 0.00 sec. Search to depth 1 0.11 sec. best move so far: d5c6 score: 291 0.16 sec. best move so far: d5e6 score: 242 0.49 sec. best move so far: d5f3 score: 195 0.55 sec. best move so far: e7e6 score: 86 1.10 sec. best move so far: d7c6 score: -180 1.26 sec. Search to depth 2 1.76 sec. best move so far: d7c6 score: 117 5.71 sec. Search to depth 3 7.31 sec. best move so far: d7c6 score: 99 15.44 sec. best move so far: e7e6 score: 86 16.26 sec. Search to depth 4 80.33 sec. best move so far: e7e6 score: 79 102.25 sec. Search to depth 5 102.42 sec. calculation of move ended It soon (even on my slow PC) prefers e7e6, which makes a recapture on d5 possible and also avoids exchange of queens compared to the played Qd7-c6. But just leave it as it is. You were "too fair" in the first round giving Dabbaba a new chance against Schach!, Dabbaba stood nextbest when it happened (Genius2 says -0.69) and Dabbaba won't win the tournament anyway! It seems that I now, with the modifications after game 4, can reproduce Dabbaba's moves in the game. Here is Dabbabas last move as a test and from the game: 8/7P/4p3/p2k2R1/8/2b5/P4PP1/6K1/b Dab's last move against Dragon - repro... -0.00 sec. Search to depth 1 0.11 sec. best move so far: e6e5 score: 1184 0.16 sec. best move so far: d5c6 score: 490 0.16 sec. best move so far: d5d6 score: 489 0.22 sec. best move so far: d5e4 score: 488 0.27 sec. Search to depth 2 0.55 sec. best move so far: d5e4 score: 495 0.60 sec. Search to depth 3 1.76 sec. best move so far: d5e4 score: 597 2.58 sec. Search to depth 4 7.42 sec. best move so far: d5e4 score: 606 8.46 sec. Search to depth 5 28.96 sec. best move so far: d5e4 score: 669 36.48 sec. best move so far: d5d6 score: 648 55.27 sec. best move so far: d5c6 score: 609 64.56 sec. Search to depth 6 133.19 sec. best move so far: d5c6 score: 616 155.55 sec. Search to depth 7 155.71 sec. calculation of move ended 0.00 sec. Search to depth 1 0.11 sec. best move so far: e6e5 score: 1184 0.16 sec. best move so far: d5c6 score: 490 0.22 sec. best move so far: d5d6 score: 489 0.27 sec. best move so far: d5e4 score: 488 0.33 sec. Search to depth 2 0.49 sec. best move so far: d5e4 score: 495 0.55 sec. Search to depth 3 0.99 sec. best move so far: d5e4 score: 597 1.26 sec. Search to depth 4 2.80 sec. best move so far: d5e4 score: 606 3.13 sec. Search to depth 5 9.23 sec. best move so far: d5e4 score: 669 11.48 sec. best move so far: d5d6 score: 648 16.98 sec. best move so far: d5c6 score: 609 19.78 sec. Search to depth 6 20.49 sec. calculation of move ended Completely identical except that your PC is 3,3 times faster than mine. I assume Dabbaba shows a speedfactor around 33 when you start Dabbaba, and it searches ca. 7000 n/s on your PC. Not 20000 n/s as Dragon does... By the way; I think cutoffs/second is the most important for a chessprogram... I have tried Dabbabas 2 bad moves against Green Light in round 4 as a test: r4rk1/p2b1ppp/2p5/3p4/1P3Pn1/2PBq3/P3Q1PP/RN3R1K/w Nb1-d2? Green Light 0.00 sec. Search to depth 0 0.05 sec. Search to depth 0 0.11 sec. Search to depth 0 0.16 sec. Search to depth 1 0.22 sec. best move so far: e2d2 score: -13 0.22 sec. best move so far: e2e3 score: 13 0.49 sec. best move so far: c3c4 score: 28 0.77 sec. Search to depth 2 1.65 sec. best move so far: c3c4 score: -86 1.81 sec. best move so far: e2e3 score: 6 2.69 sec. best move so far: e2c2 score: 7 2.97 sec. Search to depth 3 4.07 sec. best move so far: e2c2 score: 32 7.20 sec. Search to depth 4 21.43 sec. best move so far: e2c2 score: -14 22.53 sec. best move so far: e2e3 score: -2 36.32 sec. Search to depth 5 45.38 sec. best move so far: e2e3 score: 15 124.62 sec. best move so far: e2c2 score: 31 157.25 sec. Search to depth 6 157.42 sec. calculation of move ended It never considers the played Nb1-d2? 4rrk1/p2b1ppp/2p5/3p4/1P3Pn1/2PBq3/P2NQ1PP/R4R1K/w Ra1-e1? Green Light -0.00 sec. Search to depth 0 0.11 sec. Search to depth 0 0.16 sec. Search to depth 0 0.22 sec. Search to depth 1 0.38 sec. best move so far: e2e3 score: 1 0.60 sec. best move so far: a1e1 score: 17 0.77 sec. Search to depth 2 1.54 sec. best move so far: a1e1 score: -12 2.25 sec. best move so far: e2e3 score: -3 3.35 sec. Search to depth 3 4.12 sec. best move so far: e2e3 score: -54 9.18 sec. best move so far: a1e1 score: 15 16.04 sec. Search to depth 4 62.86 sec. best move so far: a1e1 score: -107 66.87 sec. best move so far: e2e3 score: -71 92.97 sec. Search to depth 5 106.26 sec. best move so far: e2e3 score: -55 212.86 sec. Search to depth 6 213.02 sec. calculation of move ended Here it considers the played Ra1-e1?, but it seems that with a little more time it had found the right Qe2xe3, that is a pawn better(Genius2)... Back to the game against Dragon. Here is the start of Dabbaba's calculation of 40.-,Kd6-d5: -0.00 sec. Search to depth 1 0.11 sec. best move so far: c3b4 score: 351 0.16 sec. best move so far: c3d4 score: 349 0.22 sec. best move so far: d6e5 score: 335 <----- 0.22 sec. Search to depth 2 0.33 sec. best move so far: a6a5 score: 434 0.38 sec. best move so far: d6e5 score: 420 <----- 0.44 sec. best move so far: c3g7 score: 404 0.49 sec. best move so far: c3d2 score: 402 0.49 sec. Search to depth 3 It is inefficient, that iteration 2 doesn't start with the best move from the previous iteration. Some day I must find out why it does so... I have seen 4 such cases in the game against Dragon (move 15, 38, 39, 40). It was also interesting to see how much time Dabbaba used on each move. Good idea with that logfile! Enough for today! Greetings Jens ---------------- Hi Jens, ich habe versucht die Partie gegen Dragon zu reproduzieren, aber Dabbaba hat gleich im 2. Zug d5 anstelle von Nf6 gespielt. Deswegen habe ich die Partie einfach weiter gespielt. Die Partie war diesmal ausgeglichener, aber Dabbaba hat im Turmenspiel ein schweren Fehler gemacht (47. ... Ra2??), danach war die Partie schnell verloren. [Event "Quali97"] [Site "60Z/1h"] [Date "1997.11.10"] [Round "5"] [White "Dragon 1.0"] [Black "Dabbaba 051197"] [Result "1-0"] 1.e4 c5 2.Nf3 {*} d5 3.d5 Qd5 4.Nc3 Qd8 5.Ne5 Nf6 6.Bb5+ Bd7 7.Qe2 e6 8.O-O Bd6 9.f4 O-O 10.Re1 Nc6 11.Bc6 Bc6 12.Nc6 c6 13.g3 Qb6 14.Ne4 c4+ 15.Kg2 Ne4 16.Qe4 Qc5 17.a4 e5 18.e5 Rae8 19.Qf5 Qd5+ 20.Kh3 Be5 21.c3 f6 22.Qf1 f5 23.Qg2 f4 24.Qd5+ d5 25.d4 d3 26.f4 Bf6 27.Rd1 d4 28.d4 Bd4 29.Rd3 Re4 30.Kg4 Rfe8 31.Raa3 Bc5 32.Ra2 Re2 33.h3 Kh8 34.Rd7 a5 35.Bd2 Rg2+ 36.Kf3 Rg1 37.Rd5 Rf1+ 38.Kg4 Rg1+ 39.Kf5 g6+ 40.Kf6 Rf8+ 41.Ke6 Bb4 42.Bb4 b4 43.a5 Rg2 44.a6 b3 45.Ra4 Rb2 46.a7 Re2+ 47.Re5 Ra2 48.Ra2 a2 49.Ra5 Re8+ 50.Kd7 Ra8 51.Kc6 Rc8+ 52.Kb7 Rf8 53.a8Q * Ich habe die Log-Datei wieder angehaengt. In der 6. Runde muss Dabbaba gegen PAX spielen. PAX spielt wirklich sehr schwach (0/5). Gute Chancen fuer Dabbaba. Ciao Torsten I regard the last game Dragon-Dabbaba as the official game in the tournament, as it was without an operator accident. Dragon got its deserved victory against Dabbaba, although Dabbaba had the best position in the middlegame; 1.33 at move 22. Dabbaba could probably play better openings by setting the random factor to zero ('r'+return switches random off/on), but I think a program should always play a little different in the opening. --------- Torsten Schoop in Germany runs a tournament with 23 computers and himself participating. He uses 2 Pentium 120 Mhz with 16 Mb ram. There are 9 rounds played at 60 moves / hour. My program Dabbaba is one of the computerprograms. After 6 rounds it has 3 losses, 2 draws and 1 win. I have improved it 3 times during the tournament. In the first four rounds it had a nasty bug, that made it play bad moves that it did not play after the game in the corresponding testpositions. That probably turned two draws into losses; Dabbaba was 0,21 and 0,39 behind when the errors occured in the two games. In the two draws it had an advantage of 3,03 pawns in one of the games, but did not manage to win, and in the other draw it had an advantage of 6,42(!) pawns, but went into a repeat 3 times, as it does not know this rule yet. In round 5 it had an advantage of 1,33 pawns, but lost it anyway... All the above evaluations are made by Genius2. The reason for the bad results despite the good positions is probably also caused by Dabbabas lack of hashtables, that gives it a bad endgame. Dabbaba also has no openingbook and has played some weird openings, but no opponent has taken real advantage of that yet. Finally in round 6 it got a nice win: [Event "Quali97"] [Site "60/60"] [Date "1997.08.11"] [Round "6"] [White "Dabbaba 051197"] [Black "PAX 0.0001L "] [Result "1-0"] 1. e4 c5 2. Nf3 d6 3. d4 cxd4 {*} 4. Bb5+ Bd7 5. Nxd4 Nf6 6. Nd2 e6 7. O-O d5 PAX is a weak opponent, as it has lost its first 5 games, so Dabbaba plays on a cheap background here. But Dabbaba plays very well in this game; in fact all the rest of its moves is identical to the moves Genius2 proposes in an analysis of the game! 8. exd5 e5? 9. Re1 Qc7 10. N2f3 e4? 11. Bg5 a6(?) 12. Bxf6 gxf6 13. Rxe4+ Kd8 (worse were 13.-,Be7 14.Nf5) 14. Bxd7 Kxd7? (Genius2: 14.-,Nxd7 15.Qe2,Bd6 16.Re1,Ne5) 15. c3 (with the nasty threat 16.Qa4+,b5 17.Nxb5) 15. - Qc4? (Genius2: 15.-,Kd8 15.Qe2,Bd6 16.Re1,Qd7) The log shows, that Dabbaba after 49 seconds understood it could improve its score from 4 pawns to 7 pawns by the move 16. Ne5+! fxe5 17. Qg4+ Kd6? Makes it easier for white, but whatever black does, he loses the queen as the knight moves with a check and the rook can capture the queen: 17.-,f5 18.Qxf5+,Ke8/Kd6 19.Qxe5+,Kd7 20.Qe8+ followed by check with the knight 19.-,Kf7 20.Qe6+,Kg7 followed by 21.Rg4+ mate. 17.-,Ke8 18.Nc6!! with the threats Rxc4/Qc8+ mate. (18.Rxe5 is also good) 18. Nf5+ Kc7 19. Rxc4+ Kd8? It seems that PAX wants to end the game... 20. Qg5+ Ke8? 21. Rc8+ {PAX resigns because of 22.Qd8+ mate} I have just corrected another bug in Dabbaba and hope for a good result in the last 3 rounds! Hi Jens, hier ist die Partie der 7. Runde: [Event "Quali97"] [Site "60Z/1h"] [Date "1997.11.27"] [Round "7"] [White "Dabbaba 171197"] [Black "Gromit 1.2"] [Result "0-1"] 1. Pd2-d4 Pg7-g6 2. Pe2-e4 Bf8-g7 3. Ng1-f3 Pd7-d6 4. Bf1-d3 Ng8-f6 5. O-O O-O 6. Bc1-e3 Nb8-c6 7. Nb1-d2 Nf6-g4 8. Qd1-e2 Ng4xe3 9. Qe2xe3 Pe7-e5 10. Pc2-c3 Pe5xd4 11. Pc3xd4 Qd8-f6 12. Nd2-b3 Bc8-g4 13. Pd4-d5 Nc6-b4 14. Nf3-d4 Pa7-a5 15. Bd3-e2 Bg4xe2 16. Qe3xe2 Pa5-a4 17. Qe2-c4 Pa4xb3 18. Nd4xb3 Nb4xa2 19. Qc4xc7 Qf6xb2 20. Ra1-b1 Qb2-a3 21. Qc7-b6 Na2-c3 22. Rb1-e1 Qa3-b2 23. Re1-a1 Bg7-d4 24. Qb6xb7 Ra8xa1 25. Rf1xa1 Nc3xe4 26. Ra1-f1 Bd4xf2+ 27. Kg1-h1 Qb2-e2 28. Qb7xf7+ Kg8xf7 29. Ph2-h4 Qe2xf1+ 30. Kh1-h2 Bf2-g3+ 31. Kh2-h3 Qf1-f5+ * Dabbaba got too many pieces hanging around. I guess that 15.Be2?? was played because Dabbaba expected to be able to continue with 18.Qxb4 - in the selektive part of the search it will not detect, that black can play Qxd4 as the simple and quick information available for square d4 does not tell, that the square is also attacked by Bg7 behind Qf6. This must be corrected some day. Die Log-Datei habe ich wieder angehaengt. Zu Deiner letzten Mail: Die Schachprogramme reagieren sehr unterschiedlich auf den K6-200. Sehr wenig profitiert z. B. CSTal (nur 1.8x schneller); schon besser Genius 5 (ca. 2.3x so schnell) und Fritz 5 (ca. 2.5x schneller); bisher am meisten profitierte Rebel 8 (ca. 2.8x so schnell). Bei Dabbaba waeren er dann 3.0x so schnell. Was wird bei Deinem Speed-Test gemessen? Ist der Dabbaba in der eigentlichen Schachsuche auch 3x so schnell? Nein, es sind keine weiteren "scandinavian programmers" an meinem Turnier beteiligt. Die Partie gegen PAX hat Dabbaba wirklich gut gespielt. Ich habe allen Programmierern mit denen ich waehrend des Turniers Kontakt habe den Auto232-Treiber fuer Windows geschickt und ihnen gesagt, dass Du jemanden zum Gedankenaustausch und testen suchst. Ciao Torsten Hi Jens, ich habe Dabbaba mit 70 min./Partie spielen lassen. Dabbaba hat gut gespielt und verdient gewonnen: [Event "Quali97"] [Site "60/60"] [Date "1997.12.11"] [Round "8"] [White "Dabbaba 031297"] [Black "Diep 1.41"] [Result "1-0"] 1.Nf3 Nc6 2.d4 d5 3.e3 Bg4 4.Bb5 Qd6 5.Bd2 O-O-O 6.c4 Qe6 7.d5 Qd5 8.Nc3 Qh5 9.Bc6 c6 10.Rc1 Nh6 11.h3 Bf3 12.Qf3 Qf3 13.f3 Nf5 14.Ne4 Nh4 15.Ng5 Rd5 16.Nf7 Nf3+ 17.Ke2 Rg8 18.Kf3 Rf5+ 19.Ke2 Rf7 20.Rc6 Rf6 21.Rc2 e6 22.Rhc1 Bd6 23.Ba5 Rgf8 24.f4 R8f7 25.Rc6 Rf5 26.Bd2 Kd7 27.e4 Rf4 28.Bf4 Bf4 29.R1c2 Bg3 30.R6c3 Rf2+ 31.Ke3 Rc2 32.Rc2 e5 33.Kf3 Bf4 34.Rc5 Kd6 35.Ra5 Bh2 36.Ra6+ Kd7 37.b3 d4 38.Ra7 Kc6 39.Ra8 Kb6 40.Rh8 h6 41.Rg8 Be5 42.Re8 Bd6 43.Re6 Kb5 44.a4+ Kb4 45.e5 Bf8 46.Re8 Bc5 47.Rc8 Kb3 48.Rc7 Bb4 49.Rg7 d3 50.Rh7 Kc2 51.Rd7 Kc3 52.e6 Bc5 53.e7 Be7 54.Re7 Kb4 55.Re4+ Kc3 56.a5 Kc2 57.Rd4 d2 58.Ke2 d1Q+ 59.Rd1 Kc3 60.a6 Kb3 61.a7 Kc2 62.Rd3 h5 63.a8Q Kb2 64.Qb8+ Kc2 65.Qb3+ Kc1 66.Rc3+ 1-0 Diep war nach 1. Nf3 aus dem Buch und mußte 1. Nc6 bereits berechnen. Ich habe die Log-Datei wieder angehängt. Ciao Torsten Hi Torsten! Thanks for the game. Dieps play was not impressive. The queen was developed much too early. It should have played 26.-,Kb7 instead of 26.-,Kd7. I have just read CSS and saw that Diep got 6 out of 11 at the mikro-WM held recently in Paris. But perhaps it is not the same version that you have... CSS also reported that Diep had an openinglibrary, but only with 1.000 moves, so no wonder it had no reply to 1.Nf3. Dabbaba played fairly well; I'm especially satisfied that it managed to put pressure on the c-file with its rooks. And it avoided draw by repeating moves. I assume it had almost used all of its time at move 60...? Greetings Jens Hi Jens, << I have just read CSS and saw that Diep got 6 out of 11 at the mikro-WM held recently in Paris. But perhaps it is not the same version that you have...>> Ja, natürlich. Die neue Version von Diep wird ja kommerziell vertrieben und solche Programme spielen in meinem "Qualifikations-Turnier" nicht mit. << Dabbaba played fairly well; I'm especially satisfied that it managed to put pressure on the c-file with its rooks.>> Ja, das hat mir auch gut gefallen. Dabbaba hat die Schwäche (Bauer auf c6) erkannt und gut dagegen gespielt. << And it avoided draw by repeating moves.>> Hast Du dafür etwas an den Algorithmen geändert? << I assume it had almost used all of its time at move 60...? >> Ja, es es waren ca. 60 Minuten. Bis bald Torsten [Event "Quali97"] [Site "60/60"] [Date "1998.01.06"] [Round "9"] [White "SSE Chess II 1.43"] [Black "Dabbaba 031297"] [Result "1-0"] 1.e4 Nf6 2.e5 Nd5 3.d4 d6 4.Nf3 {*} e5 5.e5 e6 6.Nbd2 Bd7 7.c4 Nb6 8.Bd3 Be7 9.O-O Na6 10.Be4 Nc5 11.Qc2 h6 12.Rd1 Rc8 13.Nb3 Ne4 14.Qe4 c6 15.Qg4 Kf8 16.Na5 Rc7 17.Be3 Na4 18.Rab1 Kg8 19.Nb3 Qf8 20.Rd3 h5 21.Qf4 c5 22.Na5 b6 23.Nb3 Bc6 24.Nbd2 Rd7 25.Rd7 Bd7 26.Qe4 a6 27.Qd3 Bc6 28.b3 Qb8 29.a4 Ba4 30.Qa3 Bc2 31.Rb2 Bd1 32.Qa6 Bf3 33.Nf3 Qd8 34.Bd2 Qd3 35.Rb6 Qd8 36.Qa7 Qf8 37.Rb8 It was a disappointing game by Dabbaba in the last round. I had hoped it was improved so much during the tournament so it would not play that bad. At least move 12, 15, 26, 29, 35 and 36 are positions to later improvement, but also move 4, 5, 8, 14, 16 and 18 and perhaps 19 and 20 could probably all be better. If Dabbaba had an opening book I would make it so it never answered 1.e4 by -,Nf6. This opening gives white much and easy development, and after a few weak moves black has a difficult position. (White could also have played better on move 6, 8 and 16.) ------------------ Dabbaba played the 2 last rounds without random openingplay.