Something goes wrong with lc0 since yesterday?

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

Moderators: hgm, Rebel, chrisw

Karol Majewski
Posts: 106
Joined: Sat Jul 09, 2011 4:18 pm

Re: Something goes wrong with lc0 since yesterday?

Post by Karol Majewski »

I'm still CPU guy, so I have question regarding GPU: is Leela's search on GPU deterministic, so that you can reproduce results on the same hardware, or is it non-deterministic, just like in paralell multi-CPU search?
User avatar
George Tsavdaris
Posts: 1627
Joined: Thu Mar 09, 2006 12:35 pm

Re: Something goes wrong with lc0 since yesterday?

Post by George Tsavdaris »

Can't really believe this so far. :D
Lc0 10520 has so far a 3570 CCRL 40/4 Elo performance after 50 games. :shock: :shock:
Will continue till 100 games like with 10480.

Lc0 Test10 10520 after 50 games has 3571±89 CCRL 40/4 Elo

Code: Select all

  Program          CCRL Elo   Error(cl 95%)          Games            Score 
Lc0! Test10 10520   3571.0       ±88.5            50 (+37,=10,-3)     84.0 %

   vs.                      :  games (  +,  =, -),   (%) :    Diff,    SD, CFS (%)
   Stockfish 8              :     10 (  5,  4, 1),  70.0 :  +148.0,  45.2,   99.9
   Fire 7.1                 :     10 (  7,  2, 1),  80.0 :  +230.0,  45.2,  100.0
   Andscacs 9.3             :     10 (  9,  0, 1),  90.0 :  +364.0,  45.2,  100.0
   Gull 3                   :     10 (  7,  3, 0),  85.0 :  +378.0,  45.2,  100.0
   Texel 1.07               :     10 (  9,  1, 0),  95.0 :  +453.0,  45.2,  100.0
After his son's birth they've asked him:
"Is it a boy or girl?"
YES! He replied.....
Nay Lin Tun
Posts: 708
Joined: Mon Jan 16, 2012 6:34 am

Re: Something goes wrong with lc0 since yesterday?

Post by Nay Lin Tun »

George Tsavdaris wrote: Sat Aug 11, 2018 9:55 am Can't really believe this so far. :D
Lc0 10520 has so far a 3570 CCRL 40/4 Elo performance after 50 games. :shock: :shock:
Will continue till 100 games like with 10480.

Lc0 Test10 10520 after 50 games has 3571±89 CCRL 40/4 Elo

Code: Select all

  Program          CCRL Elo   Error(cl 95%)          Games            Score 
Lc0! Test10 10520   3571.0       ±88.5            50 (+37,=10,-3)     84.0 %

   vs.                      :  games (  +,  =, -),   (%) :    Diff,    SD, CFS (%)
   Stockfish 8              :     10 (  5,  4, 1),  70.0 :  +148.0,  45.2,   99.9
   Fire 7.1                 :     10 (  7,  2, 1),  80.0 :  +230.0,  45.2,  100.0
   Andscacs 9.3             :     10 (  9,  0, 1),  90.0 :  +364.0,  45.2,  100.0
   Gull 3                   :     10 (  7,  3, 0),  85.0 :  +378.0,  45.2,  100.0
   Texel 1.07               :     10 (  9,  1, 0),  95.0 :  +453.0,  45.2,  100.0
There is another +17 elo gain from ID 520 to ID 588. :D
User avatar
George Tsavdaris
Posts: 1627
Joined: Thu Mar 09, 2006 12:35 pm

Re: Something goes wrong with lc0 since yesterday?

Post by George Tsavdaris »

Another incredible result of Lc0 10520 testnet. (by Ankan in Leela's discord)

Leela Lc0 with 10520 net running on 1 GV100, versus Stockfish dev 18080112 running on 6 threads i7 5820k @4.2 GHz

TC was 2+2 and Arena perfect2010 opening book was used and the result was:
Leela 10520 - SF dev +8 -8 =69
0 ELo
difference with confidence interval [-32 , +32] (95% c.l).


I can give the PGN if someone wants them.
After his son's birth they've asked him:
"Is it a boy or girl?"
YES! He replied.....
Lion
Posts: 531
Joined: Fri Mar 31, 2006 1:26 pm
Location: Switzerland

Re: Something goes wrong with lc0 since yesterday?

Post by Lion »

I am interested in the pgn

thank you
User avatar
George Tsavdaris
Posts: 1627
Joined: Thu Mar 09, 2006 12:35 pm

Re: Something goes wrong with lc0 since yesterday?

Post by George Tsavdaris »

Lion wrote: Sat Aug 11, 2018 5:33 pm I am interested in the pgn

thank you
https://drive.google.com/open?id=1bOZGV ... 9YxBNDbeSu
After his son's birth they've asked him:
"Is it a boy or girl?"
YES! He replied.....
Lion
Posts: 531
Joined: Fri Mar 31, 2006 1:26 pm
Location: Switzerland

Re: Something goes wrong with lc0 since yesterday?

Post by Lion »

Thanks!

Just dl them and will look at them asap

rgds
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: Something goes wrong with lc0 since yesterday?

Post by Laskos »

George Tsavdaris wrote: Sat Aug 11, 2018 2:54 pm Another incredible result of Lc0 10520 testnet. (by Ankan in Leela's discord)

Leela Lc0 with 10520 net running on 1 GV100, versus Stockfish dev 18080112 running on 6 threads i7 5820k @4.2 GHz

TC was 2+2 and Arena perfect2010 opening book was used and the result was:
Leela 10520 - SF dev +8 -8 =69
0 ELo
difference with confidence interval [-32 , +32] (95% c.l).


I can give the PGN if someone wants them.
I have on my hardware: Lc0 ID10610 on GTX 1060 6GB GPU, SF8 on one i7 3.8 GHz core.

Time control 2' + 2'' (equivalent to CCRL 40/4' CPU hardware-time wise)

Score of lc0_v16 10610 vs SF 8: 9 - 10 - 21 [0.487]
Elo difference: -8.69 +/- 75.25

40 of 40 games finished.

Performance of 3420 CCRL 40/4' Elo points. At longer time control or stronger hardware it surely increases, as Lc0 scales better
The thing is the result would have been much better for Lc0 if not for stupid endgame blunders (if not bugs). Here I present 2 of them in those 40 games:


One won game by Lc0 which it managed to draw in what seems to be even a bug:

[pgn][Event "My Tournament"] [Site "?"] [Date "2018.08.12"] [Round "39"] [White "lc0_v16 10610"] [Black "SF 8"] [Result "1/2-1/2"] [FEN "rnbqk2r/ppppppbp/5np1/8/2P5/5NP1/PP1PPP1P/RNBQKB1R w KQkq - 0 1"] [GameDuration "00:12:49"] [GameEndTime "2018-08-12T01:04:24.153 GTB Daylight Time"] [GameStartTime "2018-08-12T00:51:34.698 GTB Daylight Time"] [PlyCount "271"] [SetUp "1"] [TimeControl "120+2"] 1. Bg2 {+0.23/5 7.0s} O-O {-0.27/23 6.7s} 2. O-O {+0.24/5 7.7s} c5 {-0.18/23 5.3s} 3. d4 {+0.34/6 6.2s} cxd4 {-0.07/23 4.9s} 4. Nxd4 {+0.37/8 3.4s} d5 {-0.20/21 0.67s} 5. cxd5 {+0.40/6 3.2s} Nxd5 {-0.14/24 5.5s} 6. Nc3 {+0.43/5 2.7s} Nxc3 {-0.24/21 0.65s} 7. bxc3 {+0.46/5 2.8s} Qc7 {-0.26/23 10s} 8. Be3 {+0.49/4 6.8s} Rd8 {-0.16/22 6.5s} 9. Qb3 {+0.67/4 6.5s} Nc6 {-0.33/20 1.2s} 10. Rfd1 {+0.60/4 5.2s} Bg4 {-0.24/22 3.4s} 11. Qc4 {+0.55/4 6.2s} Bd7 {-0.20/24 6.2s} 12. Nxc6 {+0.84/3 7.5s} bxc6 {-0.17/24 15s} 13. a4 {+1.00/3 6.1s} Rdc8 {-0.24/23 7.4s} 14. a5 {+1.48/4 5.1s} Be6 {-0.22/24 4.1s} 15. Qc5 {+1.36/4 3.1s} Be5 {-0.20/26 2.5s} 16. a6 {+1.30/3 2.8s} Bd6 {-0.12/25 0.70s} 17. Qd4 {+1.26/4 2.4s} c5 {-0.25/28 4.3s} 18. Qd3 {+1.25/4 4.4s} Rab8 {-0.22/25 3.8s} 19. Bb7 {+1.23/4 2.7s} Rd8 {-0.23/23 1.2s} 20. Bh6 {+1.22/3 3.4s} Be5 {-0.07/23 3.7s} 21. Qe3 {+1.16/4 3.0s} Bh3 {-0.10/24 4.7s} 22. Rdb1 {+1.25/3 8.0s} Qd6 {-0.20/25 5.6s} 23. f3 {+1.49/4 4.4s} Qc7 {-0.09/23 3.6s} 24. Kf2 {+1.55/3 4.6s} Bd6 {-0.22/24 5.1s} 25. g4 {+1.71/3 4.4s} Rd7 {-0.83/25 8.1s} 26. Be4 {+2.34/3 4.5s} Rxb1 {-1.37/25 13s} 27. Rxb1 {+2.27/4 4.0s} Rd8 {-1.22/25 2.4s} 28. Bd5 {+2.70/4 3.1s} Bxh2 {-2.04/22 6.7s} 29. f4 {+2.87/3 3.5s} Bxg4 {-2.76/27 9.7s} 30. c4 {+4.35/3 3.8s} Qd6 {-3.60/24 15s} 31. Kg2 {+5.11/4 3.1s} Bc8 {-3.93/27 9.1s} 32. Kxh2 {+10.29/3 4.7s} Bxa6 {-4.82/24 3.2s} 33. Qc3 {+13.85/4 4.0s} Qf6 {-3.83/25 1.1s} 34. Qxf6 {+13.96/7 2.7s} exf6 {-3.93/26 0.47s} 35. e4 {+13.65/5 5.0s} Bc8 {-3.84/27 1.3s} 36. Rb8 {+14.12/3 4.9s} a5 {-3.98/27 0.40s} 37. Ra8 {+15.39/4 4.7s} a4 {-4.51/32 7.1s} 38. Rxa4 {+15.49/4 5.3s} Be6 {-4.72/27 2.6s} 39. Ra7 {+14.22/4 5.9s} Rc8 {-4.72/27 1.4s} 40. Re7 {+16.75/4 6.3s} Rb8 {-4.72/26 1.4s} 41. Kg3 {+18.31/4 6.1s} Rb3+ {-8.91/26 4.2s} 42. Kf2 {+20.62/5 5.1s} Rb2+ {-M26/35 2.1s} 43. Ke3 {+21.95/6 5.8s} Rb3+ {-M24/36 0.34s} 44. Kd2 {+21.63/6 3.9s} Rb2+ {-M20/42 1.7s} 45. Kc3 {+25.47/6 2.1s} Rb8 {-M18/44 0.32s} 46. Bxe6 {+17.81/5 2.2s} fxe6 {-4.91/22 1.0s} 47. e5 {+13.73/4 2.0s} fxe5 {-5.46/25 3.1s} 48. fxe5 {+13.36/4 1.3s} Ra8 {-6.09/26 5.0s} 49. Rxe6 {+15.25/5 2.4s} Kf7 {-6.54/26 2.5s} 50. Rc6 {+13.66/4 2.3s} Ke7 {-5.51/21 1.6s} 51. Kd3 {+16.45/4 2.1s} Rd8+ {-6.10/23 2.4s} 52. Ke4 {+18.15/5 1.9s} Rd4+ {-6.75/23 2.0s} 53. Kf3 {+14.00/6 1.7s} Kd7 {-6.94/24 2.0s} 54. Rxc5 {+15.08/4 2.3s} Ke6 {-7.43/23 2.0s} 55. Be3 {+14.05/4 2.1s} Rh4 {-8.04/24 2.0s} 56. Bf4 {+12.51/4 1.9s} Kf5 {-5.81/22 1.4s} 57. Bg3 {+13.65/4 1.4s} Rd4 {-7.50/24 2.6s} 58. Rc6 {+14.19/4 2.5s} Rd3+ {-8.68/25 2.0s} 59. Kf2 {+11.15/5 2.1s} Kg4 {-8.94/25 2.0s} 60. e6 {+13.85/5 1.1s} Rf3+ {-6.94/22 1.0s} 61. Ke2 {+14.61/5 1.5s} Rf8 {-8.08/23 0.41s} 62. Bd6 {+17.69/3 3.1s} Re8 {-11.02/26 3.3s} 63. e7 {+18.73/5 1.8s} Kf5 {-11.43/23 0.48s} 64. Ra6 {+19.96/4 2.5s} h5 {-11.98/24 4.5s} 65. Ra1 {+20.63/4 2.0s} Ke6 {-18.43/24 2.3s} 66. c5 {+18.47/3 1.7s} h4 {-26.71/23 2.0s} 67. Kf3 {+20.92/3 2.3s} Kf5 {-53.93/23 2.0s} 68. c6 {+20.87/4 2.0s} Kg5 {-21.30/21 2.0s} 69. c7 {+22.58/4 2.0s} h3 {-M30/22 2.0s} 70. Ke4 {+23.41/4 2.1s} h2 {-46.94/20 2.0s} 71. Bxh2 {+25.36/4 1.9s} Rxe7+ {-114.68/21 2.0s} 72. Kd5 {+20.45/5 2.0s} Re8 {-M24/31 2.0s} 73. Bd6 {+24.34/3 2.0s} Rc8 {-M22/31 0.27s} 74. Ke6 {+22.64/4 2.0s} Kg4 {-M28/28 1.6s} 75. Kd7 {+20.09/4 1.9s} Rg8 {-M26/31 1.5s} 76. Ke6 {+22.74/3 2.0s} Rc8 {0.00/73 0.91s} 77. Be5 {+20.89/4 2.0s} Kf3 {-M34/31 3.0s} 78. Kd5 {+20.61/3 2.1s} Re8 {-M48/24 2.2s} 79. Bb2 {+23.34/3 2.0s} Rg8 {-M52/23 2.3s} 80. Bf6 {+23.97/4 2.0s} Rc8 {-M34/30 2.0s} 81. Be5 {+17.78/4 2.0s} Re8 {0.00/75 0.90s} 82. Bf6 {+22.72/3 2.0s} Rc8 {0.00/83 1.6s} 83. Rc1 {+16.56/3 2.0s} g5 {-107.81/29 5.1s} 84. Bxg5 {+19.35/3 2.0s} Kg4 {-M44/36 2.5s} 85. Be3 {+21.59/3 2.0s} Kf3 {-M46/33 1.1s} 86. Bd4 {+20.80/4 2.0s} Kf4 {-M38/32 3.0s} 87. Bb6 {+19.72/4 2.0s} Kf5 {-M32/31 1.3s} 88. Bf2 {+21.71/4 2.0s} Kf4 {-M30/33 1.3s} 89. Rc3 {+20.79/4 2.0s} Kg5 {-M30/36 1.5s} 90. Ke6 {+21.28/4 2.0s} Kf4 {-M28/36 0.61s} 91. Be1 {+21.24/4 2.0s} Re8+ {-M26/33 1.6s} 92. Kd5 {+23.68/5 2.0s} Rc8 {-M28/38 1.4s} 93. Bg3+ {+19.38/4 2.0s} Kf5 {-M28/37 1.7s} 94. Rc1 {+21.38/5 2.0s} Kg6 {-M26/36 1.6s} 95. Ke6 {+24.67/4 2.0s} Kh5 {-M24/37 2.0s} 96. Kf5 {+33.20/4 1.8s} Rf8+ {-M24/36 1.7s} 97. Ke4 {+24.44/5 2.2s} Rc8 {-M20/46 1.7s} 98. Kf5 {+30.71/5 1.7s} Rf8+ {0.00/79 1.1s} 99. Ke5 {+21.29/5 2.2s} Rc8 {-M26/42 3.9s} 100. Bf4 {+22.05/4 2.0s} Re8+ {-M24/42 2.2s} 101. Kf5 {+35.49/5 1.7s} Rf8+ {-M22/44 0.39s} 102. Ke4 {+24.76/5 2.2s} Rc8 {-M18/49 1.9s} 103. Kf5 {+29.22/5 1.9s} Rf8+ {0.00/56 2.7s} 104. Ke6 {+22.86/5 2.1s} Kg4 {-M24/41 3.7s} 105. Bd6 {+24.26/4 2.0s} Re8+ {-M20/34 1.7s} 106. Kd5 {+22.73/5 1.9s} Kf5 {-M18/41 1.7s} 107. Bg3 {+24.58/4 2.1s} Rc8 {-M22/43 1.6s} 108. Bh2 {+21.95/4 2.0s} Kf6 {-M22/39 4.0s} 109. Kd6 {+21.15/4 1.9s} Kf5 {-M22/36 1.4s} 110. Kd5 {+21.64/5 2.0s} Kf6 {0.00/74 1.0s} 111. Bf4 {+21.05/4 2.0s} Ke7 {-M22/42 3.6s} 112. Kc6 {+18.98/4 2.0s} Ke6 {-M26/35 1.5s} 113. Kb7 {+20.03/4 2.0s} Rf8 {-M24/34 1.7s} 114. Bh6 {+22.86/4 2.0s} Rg8 {-M28/34 1.8s} 115. Bd2 {+22.40/3 2.0s} Rg7 {-M30/34 1.6s} 116. Kc6 {+18.72/4 2.0s} Rg8 {-M26/35 2.4s} 117. Bc3 {+24.95/4 2.0s} Ra8 {-M28/35 1.5s} 118. Bb2 {+24.83/4 2.0s} Rg8 {-M30/31 1.5s} 119. Bd4 {+24.28/4 2.0s} Rc8 {-M24/39 1.9s} 120. Bb6 {+20.17/3 2.0s} Rh8 {-M28/32 1.7s} 121. Bd4 {+24.11/4 2.0s} Rc8 {0.00/76 1.1s} 122. Re1+ {+18.36/3 2.0s} Kf5 {-M22/47 3.4s} 123. Kd6 {+19.12/3 2.0s} Rg8 {-M24/37 1.9s} 124. Kd5 {+26.57/4 2.0s} Rc8 {-M24/38 1.6s} 125. Bb6 {+18.25/3 1.9s} Kf4 {-M26/30 2.4s} 126. Ba5 {+20.82/4 2.0s} Rh8 {-M34/24 1.5s} 127. Bc3 {+25.06/4 1.9s} Rc8 {-M24/34 2.4s} 128. Re7 {+19.32/4 2.0s} Kf5 {-M28/31 1.8s} 129. Be5 {+21.00/4 2.0s} Ra8 {-M24/33 2.6s} 130. Bd4 {+24.94/4 2.1s} Kg4 {-M28/34 1.4s} 131. Ke4 {+28.50/4 1.9s} Rc8 {-0.08/25 1.6s} 132. Bf6 {+7.53/4 2.0s} Kh3 {-0.08/29 1.6s} 133. Re8 {+0.40/4 1.9s} Rxe8+ {-0.08/29 0.39s} 134. Kd5 {+0.14/4 2.0s} Rc8 {-0.08/32 1.8s} 135. Kc6 {+0.03/4 1.9s} Rxc7+ {-0.08/27 1.8s} 136. Kxc7 {0.00/1001 1.1s, Draw by insufficient mating material} 1/2-1/2[/pgn]


[d]6r1/2PK4/3B2p1/8/6k1/8/8/R7 w - - 8 76
White Lc0 ID10610 to move.
Draw!


Another, a completely drawn late endgame, which Lc0 manages to lose, playing really miserably.

[pgn][Event "My Tournament"] [Site "?"] [Date "2018.08.12"] [Round "40"] [White "SF 8"] [Black "lc0_v16 10610"] [Result "1-0"] [FEN "rnbqk2r/ppppppbp/5np1/8/2P5/5NP1/PP1PPP1P/RNBQKB1R w KQkq - 0 1"] [GameDuration "00:12:42"] [GameEndTime "2018-08-12T01:17:06.535 GTB Daylight Time"] [GameStartTime "2018-08-12T01:04:24.169 GTB Daylight Time"] [PlyCount "273"] [SetUp "1"] [TimeControl "120+2"] 1. d4 {+0.09/25 19s} O-O {-0.15/6 7.6s} 2. Bg2 {+0.22/23 5.2s} c6 {-0.17/8 4.6s} 3. O-O {+0.06/24 9.2s} d5 {-0.16/7 4.5s} 4. cxd5 {+0.14/22 1.3s} cxd5 {-0.16/6 3.3s} 5. Nc3 {+0.10/24 6.2s} Nc6 {-0.18/6 4.4s} 6. Ne5 {+0.29/22 4.5s} Bf5 {-0.24/5 6.0s} 7. Be3 {+0.15/23 14s} Qb6 {-0.21/4 7.0s} 8. Qd2 {+0.18/23 4.0s} Rfd8 {+0.04/5 4.8s} 9. Nxc6 {+0.11/21 1.0s} Qxc6 {+0.02/5 7.6s} 10. f3 {+0.16/23 2.9s} h5 {+0.08/4 7.7s} 11. h3 {+0.16/22 3.8s} Rac8 {-0.01/4 6.6s} 12. g4 {+0.16/23 4.5s} hxg4 {-0.01/5 6.1s} 13. hxg4 {+0.28/23 3.1s} Bd7 {-0.01/5 4.8s} 14. Bg5 {+0.21/23 8.0s} b5 {+0.07/4 6.0s} 15. a3 {+0.19/23 4.3s} Qb6 {+0.04/5 6.3s} 16. e3 {+0.26/21 0.86s} a5 {-0.01/3 7.2s} 17. Kf2 {+0.37/22 2.6s} b4 {+0.07/5 6.2s} 18. axb4 {+0.33/23 11s} axb4 {0.00/5 3.2s} 19. Na2 {+0.50/22 1.9s} Rb8 {-0.10/3 6.0s} 20. Nc1 {+0.47/22 4.0s} Bb5 {+0.15/4 4.7s} 21. Rh1 {+0.20/22 9.2s} Bc4 {-0.03/4 6.7s} 22. b3 {+0.57/21 3.3s} Rbc8 {-0.12/4 4.4s} 23. bxc4 {+0.53/22 6.6s} dxc4 {0.00/5 3.4s} 24. Rb1 {+0.68/22 1.9s} b3 {0.00/4 3.0s} 25. Qb2 {+0.77/21 0.68s} Rd5 {+0.10/3 5.9s} 26. Bxf6 {0.00/24 15s} Bxf6 {+0.04/4 2.3s} 27. Ne2 {0.00/25 2.4s} e5 {+0.37/4 3.6s} 28. g5 {0.00/25 0.52s} Bg7 {+0.53/4 3.7s} 29. Nc3 {0.00/27 7.5s} Rdd8 {+0.50/3 3.5s} 30. d5 {0.00/23 0.45s} e4 {+0.53/4 1.8s} 31. Qd2 {0.00/24 3.2s} exf3 {+0.82/4 3.3s} 32. Bxf3 {-0.13/22 0.67s} Qb4 {+1.05/3 2.4s} 33. Rhc1 {-0.10/22 1.9s} Be5 {+1.08/3 2.3s} 34. Kf1 {-0.21/24 3.5s} Kg7 {+1.28/3 3.9s} 35. Bg2 {-0.05/23 2.4s} Rd6 {+1.27/4 6.8s} 36. Bh3 {0.00/22 3.5s} Rc7 {+1.13/3 5.3s} 37. Bg2 {-0.28/25 12s} Rb6 {+1.47/3 3.8s} 38. Qe1 {-0.44/22 4.5s} Rc8 {+1.72/4 3.3s} 39. Bh3 {-0.25/21 2.9s} Ra8 {+1.83/5 5.8s} 40. Bg2 {-0.91/22 3.9s} Ra3 {+1.63/3 3.1s} 41. Nd1 {-1.22/23 2.0s} Qe7 {+1.83/3 4.6s} 42. Qh4 {+0.20/21 1.4s} c3 {+1.90/4 1.7s} 43. Rxc3 {+0.08/21 0.36s} Bxc3 {+0.96/4 2.5s} 44. Nxc3 {0.00/25 2.0s} Qxe3 {+0.53/3 1.7s} 45. Re1 {0.00/25 0.36s} Qd3+ {+0.02/4 1.3s} 46. Kf2 {0.00/27 2.2s} Rb8 {+1.22/3 3.4s} 47. Rh1 {+0.25/21 1.2s} Qc2+ {-0.68/3 2.3s} 48. Ne2 {+0.23/22 0.61s} Qc5+ {+0.04/4 1.9s} 49. Nd4 {+0.23/24 1.9s} Ra2+ {-0.16/3 1.4s} 50. Kg3 {+0.20/23 1.2s} Qc7+ {-0.10/3 2.0s} 51. Kg4 {+0.19/25 3.9s} Rxg2+ {-0.10/3 1.2s} 52. Kf3 {+0.19/27 1.2s} Rg3+ {-0.04/3 1.2s} 53. Kf2 {+0.20/27 2.4s} Qe5 {0.00/3 1.7s} 54. Qh6+ {+0.20/27 2.9s} Kg8 {+0.01/7 0s} 55. Qh7+ {+0.18/26 2.1s} Kf8 {-0.13/5 0s} 56. Qh8+ {+0.19/27 0.54s} Qxh8 {-0.11/6 3.6s} 57. Rxh8+ {+0.18/29 3.3s} Ke7 {-0.10/5 1.2s} 58. d6+ {+0.18/25 1.7s} Kxd6 {+0.07/5 4.2s} 59. Rxb8 {+0.18/25 0.76s} Rxg5 {+0.08/4 2.1s} 60. Rxb3 {+0.18/25 2.3s} Rd5 {+0.07/3 6.1s} 61. Ke3 {+0.18/25 1.8s} Ke5 {+0.09/4 2.6s} 62. Nf3+ {+0.19/22 1.8s} Kf6 {0.00/5 2.2s} 63. Ke4 {+0.20/24 2.1s} Ra5 {+0.02/4 2.0s} 64. Rb6+ {+0.18/24 1.8s} Kg7 {0.00/5 1.8s} 65. Ne5 {+0.18/22 0.41s} Ra7 {+0.02/4 2.2s} 66. Kf4 {+0.17/24 2.6s} Re7 {+0.01/4 2.0s} 67. Ra6 {+0.18/24 1.8s} Rb7 {-0.04/4 1.9s} 68. Nd3 {+0.18/26 2.2s} Rb3 {+0.10/3 2.1s} 69. Nc5 {+0.18/24 2.0s} Rb5 {+0.06/3 1.9s} 70. Ne4 {+0.18/24 0.56s} Rb4 {+0.05/4 2.0s} 71. Ke3 {+0.17/24 2.1s} Rb7 {+0.01/4 2.1s} 72. Kd3 {+0.18/25 1.8s} Re7 {+0.07/3 1.9s} 73. Kd4 {+0.18/25 0.48s} Kh6 {+0.05/4 2.1s} 74. Kd5 {+0.17/25 3.6s} f5 {+0.05/4 2.0s} 75. Nd6 {+0.11/23 2.8s} Kh5 {+0.06/3 1.8s} 76. Nc4 {+0.17/20 1.5s} g5 {-0.01/4 2.1s} 77. Ra8 {+0.08/26 2.8s} Re2 {+0.05/3 2.1s} 78. Ne5 {0.00/25 0.94s} g4 {+0.14/4 1.6s} 79. Re8 {0.00/28 2.4s} f4 {+0.24/3 2.2s} 80. Rh8+ {+0.08/25 1.5s} Kg5 {+0.11/4 0.072s} 81. Rg8+ {+0.08/26 0.46s} Kf5 {+0.21/4 2.8s} 82. Nxg4 {+0.08/29 2.5s} f3 {+0.28/3 1.6s} 83. Nh6+ {+0.08/28 0.46s} Kf4 {+0.13/5 2.1s} 84. Rf8+ {+0.06/23 2.1s} Kg3 {+0.02/4 2.2s} 85. Nf5+ {+0.07/24 0.93s} Kg2 {-0.01/4 2.6s} 86. Nh4+ {+0.68/21 1.4s} Kg3 {-0.09/5 2.4s} 87. Nxf3 {+0.69/28 2.2s} Rf2 {-0.18/5 2.2s} 88. Ke4 {+0.69/30 2.4s} Re2+ {-0.26/5 2.2s} 89. Kd4 {+0.69/32 2.3s} Ra2 {-0.21/4 2.1s} 90. Ne5 {+0.69/31 0.81s} Ra7 {-0.22/4 2.0s} 91. Nc4 {+0.69/30 3.0s} Ra4 {-0.19/4 2.0s} 92. Kd5 {+0.69/29 0.62s} Ra7 {-0.22/4 2.0s} 93. Ke4 {+0.69/32 2.1s} Ra4 {-0.19/5 1.8s} 94. Kd4 {+0.68/33 2.3s} Ra7 {-0.22/4 2.2s} 95. Kd5 {+0.68/32 2.2s} Ra4 {-0.20/4 2.0s} 96. Ne5 {+0.68/30 3.2s} Ra6 {-0.22/3 2.0s} 97. Nd3 {+0.68/33 2.1s} Ra3 {-0.20/3 2.0s} 98. Nc5 {+0.68/34 2.3s} Ra5 {-0.20/4 2.0s} 99. Kc4 {+0.68/33 0.66s} Kg4 {-0.18/4 2.0s} 100. Kb4 {+0.68/33 2.4s} Ra1 {-0.20/5 2.0s} 101. Ne4 {+0.68/34 2.4s} Ra7 {-0.20/4 1.9s} 102. Nc3 {+0.68/34 2.8s} Ra1 {-0.21/4 2.0s} 103. Kb5 {+0.68/33 0.71s} Ra7 {-0.21/4 2.0s} 104. Nd5 {+0.68/33 2.6s} Ra3 {-0.19/4 2.0s} 105. Kc5 {+0.68/32 2.4s} Ra7 {-0.19/4 2.0s} 106. Ne3+ {+0.68/31 9.1s} Kg5 {-0.22/5 1.9s} 107. Kd6 {+0.68/34 2.7s} Ra3 {-0.16/4 2.0s} 108. Nd5 {+0.68/35 2.7s} Ra7 {-0.21/4 2.0s} 109. Ke5 {+0.68/35 1.6s} Ra1 {-0.23/4 2.0s} 110. Rg8+ {+0.70/27 1.6s} Kh4 {-0.34/4 2.0s} 111. Nf6 {+0.70/32 1.9s} Kh3 {-0.33/4 2.0s} 112. Ne4 {+0.70/30 2.1s} Kh4 {-0.32/4 2.1s} 113. Kf4 {+0.70/31 2.2s} Rf1+ {-0.41/4 1.8s} 114. Ke3 {+0.70/32 0.34s} Ra1 {-0.36/3 2.1s} 115. Nf2 {+0.70/34 1.8s} Kh5 {-0.30/4 1.8s} 116. Kf4 {+0.70/34 2.0s} Rf1 {-0.27/4 2.2s} 117. Rg2 {+0.70/33 0.38s} Kh6 {-0.39/4 2.0s} 118. Kf5 {+0.70/32 2.4s} Kh7 {-0.46/4 1.8s} 119. Ke5 {+0.70/32 5.1s} Ra1 {-0.39/5 2.2s} 120. Nd3 {+0.70/31 5.1s} Ra3 {-0.32/4 2.0s} 121. Nf4 {+0.70/27 1.6s} Ra1 {-0.46/4 1.9s} 122. Ke4 {+0.70/27 2.3s} Ra5 {-0.38/4 2.1s} 123. Rg1 {+0.70/26 2.2s} Ra3 {-0.40/3 2.0s} 124. Ke5 {+0.69/23 2.5s} Ra6 {-0.47/4 2.0s} 125. Kf5 {+0.68/24 2.0s} Ra2 {-0.53/4 1.9s} 126. Kf6 {+0.59/22 2.0s} Rf2 {-0.56/4 1.9s} 127. Rg7+ {+M31/42 0.72s} Kh6 {-0.92/5 1.4s} 128. Rg4 {+M29/46 0.30s} Rh2 {-1.31/4 1.2s} 129. Rg1 {+M17/57 1.2s} Kh7 {-3.01/5 1.7s} 130. Kf7 {+M15/61 0.31s} Kh8 {-5.33/3 1.9s} 131. Rg8+ {+M13/70 1.6s} Kh7 {-1.28/5 0.021s} 132. Rg7+ {+M11/77 0.38s} Kh6 {-12.22/5 3.4s} 133. Rg6+ {+M9/93 2.1s} Kh7 {-6.83/4 0s} 134. Nd5 {+M7/126 0.41s} Kh8 {-21.08/5 5.5s} 135. Rg8+ {+M5/127 0.36s} Kh7 {-10.43/1004 0s} 136. Nf6+ {+M3/127 0.002s} Kh6 {-21.25/1002 0s} 137. Rg6# {+M1/127 0.001s, White mates} 1-0[/pgn]


[d]8/4rpk1/R5p1/8/3KN3/8/8/8 b - - 26 73
Black Lc0 ID10610 to move.
Black loses!
Sven
Posts: 4052
Joined: Thu May 15, 2008 9:57 pm
Location: Berlin, Germany
Full name: Sven Schüle

Re: Something goes wrong with lc0 since yesterday?

Post by Sven »

Laskos wrote: Sun Aug 12, 2018 12:40 am One won game by Lc0 which it managed to draw in what seems to be even a bug:
[...]

[d]6r1/2PK4/3B2p1/8/6k1/8/8/R7 w - - 8 76
White Lc0 ID10610 to move.
Draw!
Agreed, this must be a bug, why doesn't LC0 promote the pawn?
Laskos wrote: Sun Aug 12, 2018 12:40 am Another, a completely drawn late endgame, which Lc0 manages to lose, playing really miserably.
[...]

[d]8/4rpk1/R5p1/8/3KN3/8/8/8 b - - 26 73
Black Lc0 ID10610 to move.
Black loses!
Here I think this can be hard to find without EGTB. Strictly spoken the next 52 moves after the diagram were not played "miserably" since the position was still a TB draw until 126...Rf2? which leads to a mate in 16. Of course losing both pawns was absolutely not necessary but not losing them would not have changed the optimal outcome of the position, so losing them was not "incorrect".
Sven Schüle (engine author: Jumbo, KnockOut, Surprise)
tmokonen
Posts: 1296
Joined: Sun Mar 12, 2006 6:46 pm
Location: Kelowna
Full name: Tony Mokonen

Re: Something goes wrong with lc0 since yesterday?

Post by tmokonen »

So... the main server nets are gone, never to return?