Lc0 - SF10 fun games

[MikeGL]

This is what happens when Lc0 steers the game into a tactical meelee against Stockfish.
Nice game between these two.

[pgn] [Event "CCC10 Quarterfinals (10|3)"] [Site "?"] [Date "2019.09.03"] [Round "1"] [White "Lc0"] [Black "Stockfish"] [Result "0-1"] [ECO "A00"] [GameDuration "00:13:29"] [GameEndTime "2019-09-03T23:42:02.382 PDT"] [GameStartTime "2019-09-03T23:28:33.297 PDT"] [Opening "Grob"] [PlyCount "72"] [TimeControl "600+3"] [Variation "Fritz gambit"] 1. g4 d5 2. Bg2 Bxg4 3. c4 c6 4. Qb3 e6 5. Qxb7 Nd7 6. Nc3 Ne7 7. cxd5 exd5 8. d3 Rb8 9. Qxa7 Nf5 10. h3 Bc5 11. Qa4 Rb4 12. Qa6 Bxf2+ 13. Kd1 Bh5 14. Qxc6 O-O 15. Qxd5 Ne3+ 16. Bxe3 Bxe3 17. Rb1 Qb8 18. Kc2 Nf6 19. Qa5 Rc8 20. Nf3 Rb5 21. Qa4 Nd5 22. Nd2 Bxe2 23. Nc4 Nb4+ 24. Kb3 Bxd3 25. Qxb5 Bxc4+ 26. Qxc4 Na6+ 27. Qb5 Rxc3+ 28. Ka4 Nc5+ 29. Ka5 Qc7+ 30. Kb4 Nd3+ 31. Ka4 Qa7+ 32. Qa5 Qd7+ 33. Qb5 Rc4+ 34. b4 Nc5+ 35. Ka5 Qa7+ 36. Qa6 Qxa6# 0-1 [/pgn]

conclusioin: quick mate

[mwyoung]

This is what happens when Lc0 steers the game into a tactical meelee against Stockfish.
Nice game between these two.

[pgn] [Event "CCC10 Quarterfinals (10|3)"] [Site "?"] [Date "2019.09.03"] [Round "1"] [White "Lc0"] [Black "Stockfish"] [Result "0-1"] [ECO "A00"] [GameDuration "00:13:29"] [GameEndTime "2019-09-03T23:42:02.382 PDT"] [GameStartTime "2019-09-03T23:28:33.297 PDT"] [Opening "Grob"] [PlyCount "72"] [TimeControl "600+3"] [Variation "Fritz gambit"] 1. g4 d5 2. Bg2 Bxg4 3. c4 c6 4. Qb3 e6 5. Qxb7 Nd7 6. Nc3 Ne7 7. cxd5 exd5 8. d3 Rb8 9. Qxa7 Nf5 10. h3 Bc5 11. Qa4 Rb4 12. Qa6 Bxf2+ 13. Kd1 Bh5 14. Qxc6 O-O 15. Qxd5 Ne3+ 16. Bxe3 Bxe3 17. Rb1 Qb8 18. Kc2 Nf6 19. Qa5 Rc8 20. Nf3 Rb5 21. Qa4 Nd5 22. Nd2 Bxe2 23. Nc4 Nb4+ 24. Kb3 Bxd3 25. Qxb5 Bxc4+ 26. Qxc4 Na6+ 27. Qb5 Rxc3+ 28. Ka4 Nc5+ 29. Ka5 Qc7+ 30. Kb4 Nd3+ 31. Ka4 Qa7+ 32. Qa5 Qd7+ 33. Qb5 Rc4+ 34. b4 Nc5+ 35. Ka5 Qa7+ 36. Qa6 Qxa6# 0-1 [/pgn]

conclusioin: quick mate

The correct conclusion is this is what happens when Lc0 is forced to play a bullshit opening.

[MikeGL]

This is what happens when Lc0 steers the game into a tactical meelee against Stockfish.
Nice game between these two.

[pgn] [Event "CCC10 Quarterfinals (10|3)"] [Site "?"] [Date "2019.09.03"] [Round "1"] [White "Lc0"] [Black "Stockfish"] [Result "0-1"] [ECO "A00"] [GameDuration "00:13:29"] [GameEndTime "2019-09-03T23:42:02.382 PDT"] [GameStartTime "2019-09-03T23:28:33.297 PDT"] [Opening "Grob"] [PlyCount "72"] [TimeControl "600+3"] [Variation "Fritz gambit"] 1. g4 d5 2. Bg2 Bxg4 3. c4 c6 4. Qb3 e6 5. Qxb7 Nd7 6. Nc3 Ne7 7. cxd5 exd5 8. d3 Rb8 9. Qxa7 Nf5 10. h3 Bc5 11. Qa4 Rb4 12. Qa6 Bxf2+ 13. Kd1 Bh5 14. Qxc6 O-O 15. Qxd5 Ne3+ 16. Bxe3 Bxe3 17. Rb1 Qb8 18. Kc2 Nf6 19. Qa5 Rc8 20. Nf3 Rb5 21. Qa4 Nd5 22. Nd2 Bxe2 23. Nc4 Nb4+ 24. Kb3 Bxd3 25. Qxb5 Bxc4+ 26. Qxc4 Na6+ 27. Qb5 Rxc3+ 28. Ka4 Nc5+ 29. Ka5 Qc7+ 30. Kb4 Nd3+ 31. Ka4 Qa7+ 32. Qa5 Qd7+ 33. Qb5 Rc4+ 34. b4 Nc5+ 35. Ka5 Qa7+ 36. Qa6 Qxa6# 0-1 [/pgn]

conclusioin: quick mate

The correct conclusion is this is what happens when Lc0 is forced to play a bullshit opening.

You're funny.
Reminds me of that quote, "only real geniuses, like Fischer, can afford grabbing poisoned pawns".
Here, Lc0 grabbed the poisoned QNP and, as expected, lagged in piece development.

edit: sorry, not just QNP, Lc0 also grabbed a7 and c6 gambit pawns intentionally sac'd by SF to speed up blacks develpment.

[mwyoung]

This is what happens when Lc0 steers the game into a tactical meelee against Stockfish.
Nice game between these two.

[pgn] [Event "CCC10 Quarterfinals (10|3)"] [Site "?"] [Date "2019.09.03"] [Round "1"] [White "Lc0"] [Black "Stockfish"] [Result "0-1"] [ECO "A00"] [GameDuration "00:13:29"] [GameEndTime "2019-09-03T23:42:02.382 PDT"] [GameStartTime "2019-09-03T23:28:33.297 PDT"] [Opening "Grob"] [PlyCount "72"] [TimeControl "600+3"] [Variation "Fritz gambit"] 1. g4 d5 2. Bg2 Bxg4 3. c4 c6 4. Qb3 e6 5. Qxb7 Nd7 6. Nc3 Ne7 7. cxd5 exd5 8. d3 Rb8 9. Qxa7 Nf5 10. h3 Bc5 11. Qa4 Rb4 12. Qa6 Bxf2+ 13. Kd1 Bh5 14. Qxc6 O-O 15. Qxd5 Ne3+ 16. Bxe3 Bxe3 17. Rb1 Qb8 18. Kc2 Nf6 19. Qa5 Rc8 20. Nf3 Rb5 21. Qa4 Nd5 22. Nd2 Bxe2 23. Nc4 Nb4+ 24. Kb3 Bxd3 25. Qxb5 Bxc4+ 26. Qxc4 Na6+ 27. Qb5 Rxc3+ 28. Ka4 Nc5+ 29. Ka5 Qc7+ 30. Kb4 Nd3+ 31. Ka4 Qa7+ 32. Qa5 Qd7+ 33. Qb5 Rc4+ 34. b4 Nc5+ 35. Ka5 Qa7+ 36. Qa6 Qxa6# 0-1 [/pgn]

conclusioin: quick mate

The correct conclusion is this is what happens when Lc0 is forced to play a bullshit opening.

You're funny.
Reminds me of that quote, "only real geniuses, like Fischer, can afford grabbing poisoned pawns".
Here, Lc0 grabbed the poisoned QNP and, as expected, lagged in piece development.

edit: sorry, not just QNP, Lc0 also grabbed a7 and c6 gambit pawns intentionally sac'd by SF to speed up blacks develpment.

That is because whites position is already busted before the pawn grab.

[MikeGL]

How SF managed to trap its own dragon B without even realizing it.
This is a curious game where SF didn't realize its Bh8 is dead, useless piece. In informator lingo it
would be called (xBh8), and its eval score didn't even show that he is already a piece down.

[pgn] [Event "CCC10 Semifinals (10|3)"] [Site "?"] [Date "2019.09.13"] [Round "1"] [White "Leelenstein"] [Black "Stockfish"] [Result "1-0"] [ECO "A45"] [GameDuration "00:29:35"] [GameEndTime "2019-09-13T06:44:07.583 PDT"] [GameStartTime "2019-09-13T06:14:32.279 PDT"] [Opening "Trompovsky attack (Ruth, Opovcensky Opening)"] [PlyCount "233"] [TimeControl "600+3"] 1. d4 Nf6 2. Bg5 Ne4 3. Bf4 c5 4. f3 Qa5+ 5. c3 Nf6 6. d5 Qb6 7. Bc1 d6 8. e4 g6 9. a4 Bg7 10. Ne2 O-O 11. Na3 Qd8 12. Nc4 Nbd7 13. g4 Nb6 14. Ng3 Nxc4 15. Bxc4 e6 16. g5 Ne8 17. dxe6 fxe6 18. Ba2 Nc7 19. f4 d5 20. e5 b6 21. h4 d4 22. h5 dxc3 23. bxc3 Qxd1+ 24. Kxd1 Nd5 25. h6 Bh8 26. Ne2 Ba6 27. a5 Rad8 28. Ke1 b5 29. Rh3 c4 30. Bb1 Rf7 31. Be4 b4 32. Bxd5 exd5 33. Ra4 b3 34. Rb4 Rb7 35. Rxb7 Bxb7 36. Kd2 Rb8 37. Bb2 Bc6 38. Nd4 Bd7 39. Rh1 Rf8 40. Ke3 a6 41. Ba3 Rd8 42. Kd2 Re8 43. Kc1 Kf7 44. Kb2 Rc8 45. Re1 Be6 46. Bd6 Ra8 47. Nc6 Rc8 48. Nb4 d4 49. Re4 d3 50. Rd4 Ra8 51. Bc5 Re8 52. Bb6 Bg4 53. Nd5 Bxe5 54. fxe5 Rxe5 55. Ne3 Be6 56. Nxc4 Rxg5 57. Nd2 Rh5 58. Rxd3 g5 59. c4 Rxh6 60. Kxb3 Rh3 61. Rxh3 Bxh3 62. Kb4 Bc8 63. Kc5 Ke6 64. Kc6 Ke5 65. Kc7 Be6 66. Kb7 Kd6 67. Be3 g4 68. Kxa6 Kc6 69. Ka7 Bc8 70. a6 g3 71. Nf3 g2 72. Bg1 Kc7 73. Nd4 h5 74. Bh2+ Kd7 75. Kb6 g1=R 76. Bxg1 Bxa6 77. Kxa6 Kd6 78. c5+ Kxc5 79. Nf5+ Kd5 80. Ng3 h4 81. Nf5 h3 82. Ng3 Ke5 83. Be3 Kd6 84. Nh1 Ke5 85. Nf2 h2 86. Kb5 Kf5 87. Bd2 h1=Q 88. Nxh1 Ke4 89. Ba5 Kf3 90. Bc7 Kg2 91. Ng3 Kf3 92. Kc5 Ke3 93. Kc4 Kf3 94. Kd4 Kf2 95. Ke4 Kg2 96. Kf4 Kf2 97. Bb6+ Kg2 98. Kg4 Kh2 99. Kf3 Kh3 100. Bg1 Kh4 101. Ne4 Kh5 102. Kf4 Kg6 103. Ng5 Kg7 104. Bc5 Kf6 105. Ba3 Kg6 106. Be7 Kh5 107. Ne4 Kh6 108. Kf5 Kg7 109. Bf6+ Kh6 110. Bg5+ Kg7 111. Nd6 Kh7 112. Kf6 Kh8 113. Kg6 Kg8 114. Be7 Kh8 115. Nf7+ Kg8 116. Nh6+ Kh8 117. Bf6# 1-0 [/pgn]

That looks like an avoid move (Bh8) puzzle, because there surely is an alternative to avoid playing Bh8.
Maybe Rf7 instead of the N hop, so that Bf8 can be played (avoiding Bh8). Haven't analyzed, but looks like SF was blind with that long term immobility of Bh8.

[Jouni]

Here is really impressive game where SFdev beats mighty Houdini after 1.g4 opening

[pgn] [White "Stockfish 260819 64 BMI2"] [Black "Houdini 6 x64-pext"] [Result "1-0"] [ECO "A00"] [Annotator "-0.81;-0.90"] [PlyCount "135"] [TimeControl "60/300:0/0:0/0"] 1. g4 {Both last book move} d5 {-0.90/20 13} 2. e3 {-0.81/23 5 (Bg2)} e5 {-0.91/20 7} 3. d4 {-0.81/ 23 6 (Nc3)} exd4 {-1.00/21 4 (Nc6)} 4. exd4 {-0.84/19 2} Qe7+ {-0.92/21 5 (Nf6) } 5. Be3 {-0.89/22 3} Qe4 {-0.84/22 7} 6. Nf3 {-0.85/21 3} Bxg4 {-0.73/23 6} 7. Nbd2 {-0.76/22 2} Qe6 {-0.81/21 2} 8. c4 {-0.88/26 21 (h3)} Nf6 {-0.80/20 7} 9. Qa4+ {-0.81/23 7} Qc6 {-0.80/21 5} 10. Qc2 {-0.70/23 4} Bxf3 {-0.80/19 1} 11. Nxf3 {-0.69/25 3} Bb4+ {-0.89/21 5} 12. Nd2 {-0.95/27 8} dxc4 {-0.94/23 17} 13. Rg1 {-0.97/26 2} O-O {-0.98/19 1} 14. O-O-O {-0.97/25 2 (Bxc4)} c3 {-0.90/23 6} 15. bxc3 {-1.05/27 4} Qxc3 {-1.07/22 6} 16. Qxc3 {-0.90/26 3} Bxc3 {-1.10/21 0} 17. Bh6 {-0.84/28 7 (Nc4)} Ne8 {-1.13/21 6} 18. Ne4 {-1.07/27 6} Bb4 {-1.02/21 8} 19. Nc5 {-0.80/25 3} b6 {-0.93/22 15 (a5)} 20. Nd3 {-0.91/24 5} Ba3+ {-0.93/ 22 13 (Bd6)} 21. Kb1 {-0.63/24 3} Kh8 {-0.92/24 6} 22. Bg2 {-0.45/27 6 (Bf4)} c6 {-0.91/23 7} 23. Bf4 {-0.69/29 9 (Be3)} Bd6 {-0.95/24 5} 24. Ne5 {-0.63/27 3 (Bg3)} f6 {-0.57/23 15} 25. Nd3 {-0.46/27 3} Na6 {-0.88/22 13 (a5)} 26. Bxc6 {-0.44/25 5} Rd8 {-0.90/21 4} 27. h4 {-0.34/25 9 (d5)} Nec7 {-0.78/24 8} 28. d5 {-0.45/25 6} Rf7 {-0.75/24 6 (Nb8)} 29. h5 {-0.25/27 11 (Rge1)} Kg8 {-0.68/24 5 } 30. h6 {-0.29/24 2 (Rge1)} g6 {-0.71/21 5} 31. Rge1 {-0.13/26 4} Rff8 {-0.74/ 20 3 (Kf8)} 32. a4 {-0.15/23 3 (Kb2)} Rf7 {-0.39/21 7} 33. Kb2 {-0.32/27 8} Nb8 {-0.52/21 3} 34. Bb7 {0.00/27 3 (Bxd6)} Bf8 {-0.91/21 2} 35. Rc1 {0.00/27 4 (d6)} Nxd5 {-0.38/25 13 (Nba6)} 36. Bxd5 {0.00/29 3} Rxd5 {-0.19/25 5} 37. Bxb8 {0.00/32 3} Rd8 {-0.10/24 3} 38. Bg3 {0.00/31 5} Rxd3 {-0.12/23 3} 39. Rc8 {0. 00/30 4} f5 {-0.01/24 4 (Rd5)} 40. Ree8 {0.29/29 3 (Bb8)} f4 {0.05/24 4 (g5)} 41. Bh4 {0.00/36 8} Rd2+ {0.04/26 2} 42. Ka1 {0.00/39 4 (Kb3)} Rdd7 {0.05/26 3 (Rd1+)} 43. f3 {0.51/34 3 (Kb2)} Rb7 {0.08/26 2 (Rd1+)} 44. Kb2 {1.00/33 5} Rbd7 {0.08/28 2} 45. Ra8 {0.82/34 6 (Ka1)} Rc7 {0.08/27 3} 46. Bd8 {0.77/36 8} Rb7 {0.08/24 0} 47. Kb1 {0.61/39 11 (Bh4)} Rbd7 {0.08/27 2} 48. Kc2 {0.60/38 5 (Kb2)} Rb7 {0.08/29 2} 49. Bh4 {0.66/39 3 (Kb1)} Rbc7+ {0.08/28 2} 50. Kb2 {0. 60/41 8 (Kb3)} Rcd7 {0.08/29 2} 51. Kb1 {0.60/42 2 (Rac8)} Rd1+ {0.08/28 2 (Rb7)} 52. Kc2 {1.15/31 2} Rdd7 {0.08/29 0} 53. Be1 {0.96/32 7 (Kb2)} Rc7+ {0. 08/24 2} 54. Bc3 {0.38/32 10} g5 {0.08/25 2} 55. Kd3 {0.38/30 2} Rce7 {0.08/25 1} 56. Rxe7 {0.77/33 7} Rxe7 {0.06/23 0} 57. Bg7 {1.04/32 8 (Bb4)} Rf7 {0.04/ 26 2} 58. Ke2 {0.58/31 4 (Bxf8)} a6 {0.35/26 6} 59. Kf1 {0.53/30 1} b5 {0.51/ 19 1 (a5)} 60. a5 {2.04/27 2 (axb5)} b4 {0.65/27 7} 61. Kg2 {3.78/32 6} g4 {1. 59/25 3} 62. fxg4 {5.78/32 6} f3+ {1.68/30 14} 63. Kf2 {6.44/32 5 (Kf1)} b3 { 0.53/27 8} 64. g5 {7.20/33 8 (Rb8)} Rf5 {0.47/28 11 (b2)} 65. g6 {9.48/30 6} hxg6 {2.99/27 11} 66. Bxf8 {9.51/30 6} g5 {2.95/26 0} 67. Bg7+ {11.40/29 6} Kh7 {5.35/26 15} 68. Ra7 {11.57/27 6 adjud.} 1-0 [/pgn]

This doesn't change the fact, that g4 is losing move.

[Kotlov]

This doesn't change the fact, that g4 is losing move.

all 20 starting moves is draw.
e4 == g4
draw == draw

[OneTrickPony]

1.g4 may actually be losing though
As to Leelenstein game: one recurring them in NN vs alpha-beta engines games is the NN engine pushing h4-h5-h6. Maybe SF could be improved if the value of the maneuver was somehow included in search or eval.