Strange loss by LC0 at CCCC

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

MikeGL
Posts: 1010
Joined: Thu Sep 01, 2011 2:49 pm

Strange loss by LC0 at CCCC

Post by MikeGL »

Ok, the currently running CCCC is exciting to watch and the current standings shows Lc0 as third,
So far, only 3 games between Lc0-SF9 at CCCC with 2 draws and 1 win in favor of SF. But the win of SF on this game is really strange as the K of
white was active and K of black was trapped at K side.


Current Standings (games still ongoing)

1
Stockfish (3443)
94.5/104

2
Houdini (3408)
90.5/105

3
Lc0 (3300)
81.5/105

4
Ethereal (3057)
80.5/105

5
Fire (3298)
80.0/105

6
Komodo (3409)
79.0/105

7
Laser (3147)
68.0/105

8
Andscacs (3251)
67.5/104

9
Shredder (3292)
66.0/104

10
Xiphos (3116)
64.5/104

11
Booot (3228)
64.0/105

12
Chiron (3201)
63.0/105

13
Pedone (3098)
54.0/104

14
Komodo MC (3184)
53.0/105

15
Fritz (3204)
52.5/105

16
Arasan (3126)
50.5/104

17
Gull (3182)
50.0/105

18
Texel (3149)
49.0/104

19
Nemorino (3102)
49.0/105

20
Vajolet (3048)
47.5/104

21
Fizbo (3286)
41.0/105

22
Equinox (3177)
40.5/105

23
Wasp (3046)
40.5/105

24
Critter (3161)
40.5/105

25
Nirvana (3193)
40.0/104

26
Hannibal (3156)
38.5/105

27
Senpai (3127)
37.0/104

28
Ivanhoe (3114)
35.5/105

29
Protector (3140)
32.0/104

30
Bobcat (3027)
30.0/105

31
Black Mamba (3124)
26.0/105

32
Crafty (3013)
17.0/105

33
Alfilx64 (3139)
4.0/105

Both engines were showing a drawn endgame and suddenly dropped which gave black(SF) the point.

[pgn] [Event "CCC 2: Blitz Battle (5|2)"] [Site "Chess.com"] [Date "2018.10.12"] [Round "?"] [White "Lc0 0.18.1"] [Black "Stockfish 18092721"] [Result "0-1"] [WhiteElo "2000"] [BlackElo "3287"] [WhiteType "program"] [BlackType "program"] [ECO "E19"] [Opening "Queen's Indian"] [Variation "old main line, 9.Qxc3"] [TimeControl "300+2"] [Time "20:59:41"] [Termination "normal"] [PlyCount "224"] 1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. g3 Bb7 5. Bg2 Be7 6. O-O O-O 7. Nc3 Ne4 8. Qc2 Nxc3 9. Qxc3 c5 10. Rd1 d6 11. Qd3 Nc6 12. d5 exd5 13. cxd5 Nb4 14. Qb1 Bf6 15. e4 c4 16. Nd4 Rc8 17. Be3 Qd7 18. h4 Ba6 19. a4 Rc7 20. Kh2 h6 21. Ra3 Nd3 22. Qc2 Nb4 23. Qd2 Nd3 24. Rc3 Nc5 25. Nf5 Re8 26. f3 Bxc3 27. bxc3 Re5 28. Bxh6 Rxf5 29. exf5 Qxf5 30. Bf4 Nd3 31. Bxd6 Rc8 32. Qg5 Qxg5 33. hxg5 Rd8 34. Ba3 Rxd5 35. f4 Rd8 36. f5 Bc8 37. g4 g6 38. Rf1 Re8 39. Bd6 Kg7 40. f6+ Kg8 41. Be7 Bxg4 42. Kg3 Be6 43. Ra1 Rc8 44. Bb7 Rb8 45. Bc6 Kh7 46. Rh1+ Kg8 47. Bb5 Nb2 48. Rb1 Nd3 49. Rh1 Rc8 50. Ba6 Rb8 51. Rb1 Bd7 52. a5 Bf5 53. Rb5 Bd7 54. Rd5 Be6 55. Rd4 bxa5 56. Bxc4 Bxc4 57. Rxc4 Rb5 58. Kh4 Rb8 59. Bd6 Rd8 60. Be7 Rb8 61. Bd6 Re8 62. Rc7 a6 63. Kg4 a4 64. Be7 Ne5+ 65. Kg3 Nd3 66. Kg4 a5 67. Kf3 Kh7 68. Bd6 Kg8 69. Be7 a3 70. Bxa3 Ne1+ 71. Kf4 Nd3+ 72. Kf3 Ne1+ 73. Kg4 Nd3 74. Bd6 Ra8 75. Rc4 Nb2 76. Rd4 Rc8 77. Be7 a4 78. Bb4 Rc4 79. Kf3 Rxd4 80. cxd4 Nc4 81. Bc3 a3 82. Ke4 a2 83. Kd5 Nd2 84. Bb2 Kh7 85. Kc5 Ne4+ 86. Kc6 Nxg5 87. d5 Nh3 88. Kb5 Nf4 89. d6 g5 90. Kb4 g4 91. Ka3 g3 92. Kxa2 Ne6 93. Bc1 Kg6 94. Be3 Kxf6 95. Kb3 g2 96. Kc4 Nf4 97. Kb5 Ke6 98. Kc6 Ng6 99. Kc7 Ne5 100. Bd4 f5 101. Bc5 f4 102. Bd4 f3 103. Bc5 Kf5 104. Bg1 Kg4 105. Bd4 Kh3 106. Kb7 Nf7 107. d7 Kh2 108. Bc3 g1=Q 109. Kc8 Qc5+ 110. Kb7 Nd6+ 111. Ka6 Qc6+ 112. Ka5 Qb5# 0-1 [/pgn]

[Event "CCC 2: Blitz Battle (5|2)"]
[Site "Chess.com"]
[Date "2018.10.12"]
[Round "?"]
[White "Lc0 0.18.1"]
[Black "Stockfish 18092721"]
[Result "0-1"]
[WhiteElo "2000"]
[BlackElo "3287"]
[WhiteType "program"]
[BlackType "program"]
[ECO "E19"]
[Opening "Queen's Indian"]
[Variation "old main line, 9.Qxc3"]
[TimeControl "300+2"]
[Time "20:59:41"]
[Termination "normal"]
[PlyCount "224"]

1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. g3 Bb7 5. Bg2 Be7 6. O-O O-O 7. Nc3 Ne4 8. Qc2
Nxc3 9. Qxc3 c5 10. Rd1 d6 11. Qd3 Nc6 12. d5 exd5 13. cxd5 Nb4 14. Qb1 Bf6 15.
e4 c4 16. Nd4 Rc8 17. Be3 Qd7 18. h4 Ba6 19. a4 Rc7 20. Kh2 h6 21. Ra3 Nd3 22.
Qc2 Nb4 23. Qd2 Nd3 24. Rc3 Nc5 25. Nf5 Re8 26. f3 Bxc3 27. bxc3 Re5 28. Bxh6
Rxf5 29. exf5 Qxf5 30. Bf4 Nd3 31. Bxd6 Rc8 32. Qg5 Qxg5 33. hxg5 Rd8 34. Ba3
Rxd5 35. f4 Rd8 36. f5 Bc8 37. g4 g6 38. Rf1 Re8 39. Bd6 Kg7 40. f6+ Kg8 41. Be7
Bxg4 42. Kg3 Be6 43. Ra1 Rc8 44. Bb7 Rb8 45. Bc6 Kh7 46. Rh1+ Kg8 47. Bb5 Nb2
48. Rb1 Nd3 49. Rh1 Rc8 50. Ba6 Rb8 51. Rb1 Bd7 52. a5 Bf5 53. Rb5 Bd7 54. Rd5
Be6 55. Rd4 bxa5 56. Bxc4 Bxc4 57. Rxc4 Rb5 58. Kh4 Rb8 59. Bd6 Rd8 60. Be7 Rb8
61. Bd6 Re8 62. Rc7 a6 63. Kg4 a4 64. Be7 Ne5+ 65. Kg3 Nd3 66. Kg4 a5 67. Kf3
Kh7 68. Bd6 Kg8 69. Be7 a3 70. Bxa3 Ne1+ 71. Kf4 Nd3+ 72. Kf3 Ne1+ 73. Kg4 Nd3
74. Bd6 Ra8 75. Rc4 Nb2 76. Rd4 Rc8 77. Be7 a4 78. Bb4 Rc4 79. Kf3 Rxd4 80. cxd4
Nc4 81. Bc3 a3 82. Ke4 a2 83. Kd5 Nd2 84. Bb2 Kh7 85. Kc5 Ne4+ 86. Kc6 Nxg5 87.
d5 Nh3 88. Kb5 Nf4 89. d6 g5 90. Kb4 g4 91. Ka3 g3 92. Kxa2 Ne6 93. Bc1 Kg6 94.
Be3 Kxf6 95. Kb3 g2 96. Kc4 Nf4 97. Kb5 Ke6 98. Kc6 Ng6 99. Kc7 Ne5 100. Bd4 f5
101. Bc5 f4 102. Bd4 f3 103. Bc5 Kf5 104. Bg1 Kg4 105. Bd4 Kh3 106. Kb7 Nf7 107.
d7 Kh2 108. Bc3 g1=Q 109. Kc8 Qc5+ 110. Kb7 Nd6+ 111. Ka6 Qc6+ 112. Ka5 Qb5# 0-1

At first sight, I think the below position is impossible for white to lose because of its active K!
While blacks K was stuck at Kg8.

Really strange how white lost this position bit by bit.

[d]1r4k1/p3Bp2/5Pp1/p5P1/2R4K/2Pn4/8/8 w - - 7 61

Game credit: Chess.com for this interesting CCCC 2018 tourn.
I told my wife that a husband is like a fine wine; he gets better with age. The next day, she locked me in the cellar.
Jouni
Posts: 3283
Joined: Wed Mar 08, 2006 8:15 pm

Re: Strange loss by LC0 at CCCC

Post by Jouni »

I think LC0 move 83.Kd5?? was a simple tactical miss losing immediately after -Nd2! LCO misses 5 move tactics :o .
Jouni
Nay Lin Tun
Posts: 708
Joined: Mon Jan 16, 2012 6:34 am

Re: Strange loss by LC0 at CCCC

Post by Nay Lin Tun »

That neural network ID has wrong policy head, that means suicidal. Leela didnt see 83.Kd5 ?? was losing until 1.6 million positions have been searched by MCTS!

Although 1.6 million positions may take 1 sec for AB engines, it will take about 1000 secs for Leela.

(Leela's strength is positional evaluation, 1 positional evaluaiton is approximately equivalent to 1000 positional evaluaiton of Stockfish.) :D
jp
Posts: 1470
Joined: Mon Apr 23, 2018 7:54 am

Re: Strange loss by LC0 at CCCC

Post by jp »

Which network ID is being used for the current chess.com CCCC? I can't see this info on chess.com.
Which ID was used for the TCEC Cup?
Nay Lin Tun
Posts: 708
Joined: Mon Jan 16, 2012 6:34 am

Re: Strange loss by LC0 at CCCC

Post by Nay Lin Tun »

jp wrote: Sat Oct 27, 2018 6:56 pm Which network ID is being used for the current chess.com CCCC? I can't see this info on chess.com.
Which ID was used for the TCEC Cup?
Both (engine version 18.1, network ID 11248)