Super bullet chess: LC0 network 31510 vs Fire 7.1

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

Hai
Posts: 598
Joined: Sun Aug 04, 2013 1:19 pm

Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by Hai »

Super bullet chess: LC0 network 31510 vs Fire 7.1

40 games
ponder on
LC0 with super weak gpu 1 kN/s
Fire 7.1 with 4 cores @ 4.2 ghz
Time control: 1 min per game + 1 second per move.

Result:
+11 =15 -14
LC0 18.5 points
Fire 21.5 points
= LC0 is -26 elo weaker.
Hai
Posts: 598
Joined: Sun Aug 04, 2013 1:19 pm

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by Hai »

LC0 is a genius.
At the beginning of a game you should take 2 pawns and the 2 queens of the board and start an attack with your king in the middle of the board.

1.f4:
[pgn][Event "\[126\] LC0 (ID 31510) vs Fire 7.1 (3350 e"] [Site "Lokaler Computer"] [Date "2018.11.24"] [Round "23"] [White "Lc0 v0.19.0"] [Black "Fire 7.1 x64 popcnt"] [Result "1-0"] [ECO "A02"] [WhiteElo "3400"] [BlackElo "3350"] [PlyCount "317"] [EventDate "2018.??.??"] 1. f4 Nc6 2. e4 d5 3. e5 d4 4. Nf3 e6 5. a3 Bc5 6. b4 Bb6 7. c4 dxc3 8. dxc3 Qxd1+ 9. Kxd1 Nh6 10. Bd3 Ng4 11. Ke2 Nf2 12. c4 a5 13. c5 Nxd3 14. Kxd3 Ba7 15. b5 Ne7 16. Kc4 Nd5 17. Bd2 O-O 18. a4 f6 19. Nc3 Bd7 20. Rhe1 c6 21. b6 Bb8 22. Ne4 fxe5 23. fxe5 Rxf3 24. gxf3 Bxe5 25. f4 Bxa1 26. Rxa1 Ne7 27. Nd6 Bc8 28. Kd4 Kf8 29. Rg1 Nd5 30. h4 Rb8 31. Ke5 Ra8 32. Ne4 Ne7 33. h5 Nd5 34. Bc3 h6 35. Ba1 Rb8 36. Bb2 Ra8 37. Ba1 Rb8 38. Rg3 Ra8 39. Bc3 Rb8 40. Rg1 Ra8 41. Bb2 Kg8 42. Kd6 e5 43. Bxe5 Bf5 44. Nf6+ Nxf6 45. Bxf6 g6 46. hxg6 Rc8 47. Rh1 Bxg6 48. Rxh6 Kf7 49. Bc3 Be4 50. Kd7 Bf5+ 51. Kd6 Rd8+ 52. Kc7 Rd7+ 53. Kb8 Ke8 54. Bxa5 Bc2 55. Bc3 Ke7 56. a5 Bf5 57. Be5 Bd3 58. Kc8 Bf5 59. Bc7 Rd6+ 60. Kxb7 Rxh6 61. a6 Rh8 62. Be5 Rh7 63. Ka8 Ke6 64. b7 Kd5 65. b8=Q Kc4 66. Qb6 Rd7 67. Qxc6 Kd3 68. Kb8 Rd8+ 69. Kc7 Rc8+ 70. Kb6 Rxc6+ 71. Kxc6 Be6 72. a7 Ba2 73. a8=Q Be6 74. Kd6 Bg4 75. c6 Bf5 76. c7 Ke3 77. Ke7 Bg4 78. c8=Q Bxc8 79. Qxc8 Ke4 80. Kd8 Kd3 81. Bb8 Ke3 82. Ke7 Kd4 83. Kf8 Kd3 84. Kg8 Ke2 85. Kh8 Kf3 86. Qh3+ Ke4 87. Qh5 Kd3 88. Qh2 Kc3 89. Qh5 Kd3 90. f5 Kc3 91. Qh2 Kc4 92. Qh5 Kb4 93. f6 Kc4 94. f7 Kb3 95. Qh7 Kc3 96. Qg8 Kb2 97. Qg7+ Kb3 98. Qg5 Kc2 99. Qg2+ Kb3 100. Qb7+ Kc3 101. Ba7 Kc4 102. Qc8+ Kb4 103. Qb7+ Kc4 104. Qh1 Kc3 105. Qh7 Kd2 106. Kg7 Kc1 107. Qg8 Kb1 108. Bc5 Ka1 109. Qh7 Kb2 110. Qh4 Kb3 111. Qh7 Kc4 112. Qh3 Kxc5 113. Qh5+ Kd4 114. Qh7 Kc4 115. Qh4+ Kb3 116. Qh8 Kc3 117. Qh5 Kd2 118. Qh4 Kc1 119. Qh7 Kd2 120. Qh3 Kc2 121. Qh5 Kb2 122. Qh7 Kc3 123. Qh4 Kb2 124. Qh8 Kc3 125. Qh1 Kd3 126. Qb7 Kd2 127. Qh1 Kc3 128. Qh8 Kb3 129. Qh6 Kb4 130. Qh3 Kc5 131. Qh8 Kb5 132. Qh4 Kc6 133. Qh7 Kb7 134. Qh8 Kb6 135. Qh5 Ka6 136. Qh7 Kb7 137. Qb1+ Kc6 138. Qh1+ Kb5 139. Qh8 Kc4 140. Qg8 Kd3 141. Kf8 Ke4 142. Qg5 Kd4 143. Qg2 Kc5 144. Qg6 Kd5 145. Qa6 Ke4 146. Qh6 Kd3 147. Qh2 Kc3 148. Qh5 Kd2 149. Qh2+ Kd3 150. Qh5 Ke3 151. Qh4 Ke2 152. Qh7 Kf2 153. Qc2+ Kf1 154. Ke7 Ke1 155. Kd6 Kf1 156. Ke5 Ke1 157. Kf4 Kf1 158. Kf3 Ke1 159. Qe2# 1-0[/pgn]
Hai
Posts: 598
Joined: Sun Aug 04, 2013 1:19 pm

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by Hai »

How to win with 1.h4 and kingside attack

[pgn][Event "\[126\] LC0 (ID 31510) vs Fire 7.1 (3350 e"] [Site "Lokaler Computer"] [Date "2018.11.24"] [Round "31"] [White "Lc0 v0.19.0"] [Black "Fire 7.1 x64 popcnt"] [Result "1-0"] [WhiteElo "3400"] [BlackElo "3350"] [PlyCount "359"] [EventDate "2018.??.??"] 1. h4 e5 2. c4 Nf6 3. d3 d5 4. cxd5 Qxd5 5. Nc3 Qa5 6. Nf3 Nc6 7. Bd2 Qb6 8. Na4 Qb5 9. e3 Qd5 10. a3 Bg4 11. Be2 Be7 12. b4 a6 13. Rc1 Rd8 14. O-O O-O 15. Bc3 b5 16. Nb2 h6 17. Qc2 a5 18. bxa5 Bxa3 19. Ra1 Bd6 20. Rfc1 Qe6 21. Nd1 Ra8 22. Be1 Ra6 23. Nc3 Rb8 24. Ne4 Nxe4 25. dxe4 Na7 26. Rd1 c5 27. Rd5 c4 28. Rad1 Bc7 29. Qb2 f6 30. h5 Qf7 31. Nd4 c3 32. Qc2 Bxe2 33. Nxe2 b4 34. Ng3 Kh8 35. Nf5 Qe6 36. Rd7 Nb5 37. Re7 Qc4 38. Rdd7 Rg8 39. Nxh6 Rd6 40. Rxc7 b3 41. Qxc3 Qxc3 42. Bxc3 Rd1+ 43. Kh2 Nxc7 44. Nxg8 Nb5 45. Bb2 Rd2 46. Rb7 Rxb2 47. Nxf6 gxf6 48. Rxb5 Kg7 49. Kg3 Kh6 50. Rb6 Kg5 51. h6 Kxh6 52. Rxf6+ Kh7 53. Rb6 Ra2 54. Rxb3 Rxa5 55. f4 Ra4 56. Kf3 Ra6 57. Rb5 exf4 58. exf4 Ra1 59. Kg4 Ra2 60. g3 Ra4 61. e5 Kg6 62. Rb6+ Kf7 63. Kf5 Ke8 64. Kf6 Ra8 65. e6 Kf8 66. f5 Ke8 67. g4 Kd8 68. Kf7 Ra7+ 69. Kf8 Ra8 70. e7+ Kd7+ 71. Kf7 Rh8 72. f6 Rh7+ 73. Kg6 Rh8 74. Kg7 Ra8 75. Kf7 Rh8 76. g5 Ra8 77. Rb7+ Kc6 78. e8=Q+ Kxb7 79. Qxa8+ Kxa8 80. Kg6 Ka7 81. Kh5 Kb6 82. Kh4 Kb5 83. Kh3 Kb4 84. Kh2 Kc4 85. Kg2 Kc5 86. Kg1 Kd6 87. Kf1 Ke6 88. Ke1 Kf7 89. Ke2 Ke6 90. Kd2 Kf7 91. Kd1 Kg6 92. Ke2 Kf7 93. Ke1 Ke6 94. Kf2 Kd6 95. Ke2 Ke6 96. Kd3 Kf7 97. Kd2 Kg6 98. Kc2 Kf7 99. Kd3 Ke6 100. Kc2 Kd6 101. Kc1 Kd7 102. Kd2 Ke6 103. Kc3 Kf7 104. Kc2 Kg6 105. Kc1 Kh7 106. Kb2 Kg6 107. Kb1 Kf7 108. Ka2 Ke6 109. Kb2 Kf7 110. Kb1 Ke6 111. Ka2 Kf7 112. Kb3 Kg6 113. Kb4 Kf7 114. Kc3 Ke6 115. Kb3 Kf7 116. Kc4 Kg6 117. Kd5 Kf7 118. Kd4 Kg6 119. Ke5 Kf7 120. Kf5 Kg8 121. Kf4 Kh7 122. Kf3 Kh8 123. g6 Kg8 124. Kf2 Kf8 125. Ke2 Kg8 126. Kd2 Kf8 127. Kd1 Ke8 128. Ke2 Kf8 129. Kd2 Ke8 130. Kd1 Kd7 131. f7 Ke7 132. Ke2 Kf8 133. Kd2 Kg7 134. Kd1 Kf8 135. Ke2 Ke7 136. Kd2 Kf8 137. Kc2 Kg7 138. Kb2 Kf8 139. Kb1 Kg7 140. Kc2 Kf8 141. Kd3 Ke7 142. Kd4 Kf8 143. Ke3 Ke7 144. Kf4 Kf8 145. Kf3 Kg7 146. Kg4 Kf8 147. Kg3 Kg7 148. Kg2 Kf8 149. Kf2 Kg7 150. Kg1 Kf8 151. Kg2 Ke7 152. Kh1 Kf8 153. Kg1 Kg7 154. Kf2 Kf8 155. Kf3 Kg7 156. Ke4 Kf8 157. Kd4 Kg7 158. Kd5 Kf8 159. Kc4 Kg7 160. Kb4 Kf8 161. Kb5 Kg7 162. Kc6 Kf8 163. Kc5 Ke7 164. Kb5 Kf8 165. Kb4 Kg7 166. Ka3 Kf8 167. Kb3 Kg7 168. Kc4 Kf8 169. Kd5 Ke7 170. Ke5 Kf8 171. Ke6 Kg7 172. Ke7 Kxg6 173. f8=Q Kg5 174. Ke6 Kg4 175. Ke5 Kg3 176. Ke4 Kg2 177. Ke3 Kh2 178. Kf2 Kh3 179. Qb4 Kh2 180. Qh4# 1-0[/pgn]
Uri Blass
Posts: 10267
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by Uri Blass »

Hai wrote: Sat Nov 24, 2018 1:52 pm LC0 is a genius.
At the beginning of a game you should take 2 pawns and the 2 queens of the board and start an attack with your king in the middle of the board.

1.f4:
[pgn][Event "\[126\] LC0 (ID 31510) vs Fire 7.1 (3350 e"] [Site "Lokaler Computer"] [Date "2018.11.24"] [Round "23"] [White "Lc0 v0.19.0"] [Black "Fire 7.1 x64 popcnt"] [Result "1-0"] [ECO "A02"] [WhiteElo "3400"] [BlackElo "3350"] [PlyCount "317"] [EventDate "2018.??.??"] 1. f4 Nc6 2. e4 d5 3. e5 d4 4. Nf3 e6 5. a3 Bc5 6. b4 Bb6 7. c4 dxc3 8. dxc3 Qxd1+ 9. Kxd1 Nh6 10. Bd3 Ng4 11. Ke2 Nf2 12. c4 a5 13. c5 Nxd3 14. Kxd3 Ba7 15. b5 Ne7 16. Kc4 Nd5 17. Bd2 O-O 18. a4 f6 19. Nc3 Bd7 20. Rhe1 c6 21. b6 Bb8 22. Ne4 fxe5 23. fxe5 Rxf3 24. gxf3 Bxe5 25. f4 Bxa1 26. Rxa1 Ne7 27. Nd6 Bc8 28. Kd4 Kf8 29. Rg1 Nd5 30. h4 Rb8 31. Ke5 Ra8 32. Ne4 Ne7 33. h5 Nd5 34. Bc3 h6 35. Ba1 Rb8 36. Bb2 Ra8 37. Ba1 Rb8 38. Rg3 Ra8 39. Bc3 Rb8 40. Rg1 Ra8 41. Bb2 Kg8 42. Kd6 e5 43. Bxe5 Bf5 44. Nf6+ Nxf6 45. Bxf6 g6 46. hxg6 Rc8 47. Rh1 Bxg6 48. Rxh6 Kf7 49. Bc3 Be4 50. Kd7 Bf5+ 51. Kd6 Rd8+ 52. Kc7 Rd7+ 53. Kb8 Ke8 54. Bxa5 Bc2 55. Bc3 Ke7 56. a5 Bf5 57. Be5 Bd3 58. Kc8 Bf5 59. Bc7 Rd6+ 60. Kxb7 Rxh6 61. a6 Rh8 62. Be5 Rh7 63. Ka8 Ke6 64. b7 Kd5 65. b8=Q Kc4 66. Qb6 Rd7 67. Qxc6 Kd3 68. Kb8 Rd8+ 69. Kc7 Rc8+ 70. Kb6 Rxc6+ 71. Kxc6 Be6 72. a7 Ba2 73. a8=Q Be6 74. Kd6 Bg4 75. c6 Bf5 76. c7 Ke3 77. Ke7 Bg4 78. c8=Q Bxc8 79. Qxc8 Ke4 80. Kd8 Kd3 81. Bb8 Ke3 82. Ke7 Kd4 83. Kf8 Kd3 84. Kg8 Ke2 85. Kh8 Kf3 86. Qh3+ Ke4 87. Qh5 Kd3 88. Qh2 Kc3 89. Qh5 Kd3 90. f5 Kc3 91. Qh2 Kc4 92. Qh5 Kb4 93. f6 Kc4 94. f7 Kb3 95. Qh7 Kc3 96. Qg8 Kb2 97. Qg7+ Kb3 98. Qg5 Kc2 99. Qg2+ Kb3 100. Qb7+ Kc3 101. Ba7 Kc4 102. Qc8+ Kb4 103. Qb7+ Kc4 104. Qh1 Kc3 105. Qh7 Kd2 106. Kg7 Kc1 107. Qg8 Kb1 108. Bc5 Ka1 109. Qh7 Kb2 110. Qh4 Kb3 111. Qh7 Kc4 112. Qh3 Kxc5 113. Qh5+ Kd4 114. Qh7 Kc4 115. Qh4+ Kb3 116. Qh8 Kc3 117. Qh5 Kd2 118. Qh4 Kc1 119. Qh7 Kd2 120. Qh3 Kc2 121. Qh5 Kb2 122. Qh7 Kc3 123. Qh4 Kb2 124. Qh8 Kc3 125. Qh1 Kd3 126. Qb7 Kd2 127. Qh1 Kc3 128. Qh8 Kb3 129. Qh6 Kb4 130. Qh3 Kc5 131. Qh8 Kb5 132. Qh4 Kc6 133. Qh7 Kb7 134. Qh8 Kb6 135. Qh5 Ka6 136. Qh7 Kb7 137. Qb1+ Kc6 138. Qh1+ Kb5 139. Qh8 Kc4 140. Qg8 Kd3 141. Kf8 Ke4 142. Qg5 Kd4 143. Qg2 Kc5 144. Qg6 Kd5 145. Qa6 Ke4 146. Qh6 Kd3 147. Qh2 Kc3 148. Qh5 Kd2 149. Qh2+ Kd3 150. Qh5 Ke3 151. Qh4 Ke2 152. Qh7 Kf2 153. Qc2+ Kf1 154. Ke7 Ke1 155. Kd6 Kf1 156. Ke5 Ke1 157. Kf4 Kf1 158. Kf3 Ke1 159. Qe2# 1-0[/pgn]
I do not consider LC0 to be a genius
She clearly showed that she is a stupid engine in this game.
Other chess programs or humans who are not beginners do not need 80 moves to mate with KQBP vs K and can win faster.
LC0 is lucky that she has an increment of 1 second otherwise she probably could often lose on time.
User avatar
CMCanavessi
Posts: 1142
Joined: Thu Dec 28, 2017 4:06 pm
Location: Argentina

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by CMCanavessi »

And how many of those "chess programs or humans who are not beginners" can beat Fire 7.1?
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
F. Bluemers
Posts: 868
Joined: Thu Mar 09, 2006 11:21 pm
Location: Nederland

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by F. Bluemers »

CMCanavessi wrote: Sat Nov 24, 2018 9:09 pm And how many of those "chess programs or humans who are not beginners" can beat Fire 7.1?
I think this is a bit unfair towards Uri. He sometimes seems to be trolling a bit , but he is spot on here.
The play in the endings is just extremely lousy.It can't hurt to point this out.
In the extreme case in otb meetings ab-searchers will just click a box when playing lc0 and have an easy win even though rating lists
would predict otherwise.
Best
Fonzy
User avatar
CMCanavessi
Posts: 1142
Joined: Thu Dec 28, 2017 4:06 pm
Location: Argentina

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by CMCanavessi »

F. Bluemers wrote: Sun Nov 25, 2018 1:06 am
CMCanavessi wrote: Sat Nov 24, 2018 9:09 pm And how many of those "chess programs or humans who are not beginners" can beat Fire 7.1?
I think this is a bit unfair towards Uri. He sometimes seems to be trolling a bit , but he is spot on here.
The play in the endings is just extremely lousy.It can't hurt to point this out.
In the extreme case in otb meetings ab-searchers will just click a box when playing lc0 and have an easy win even though rating lists
would predict otherwise.
Best
Fonzy
We already know that, currently, the endgame is the weakest part in lc0... yet, Leela still won the game. And besides, he said (quote) "She clearly showed that she is a stupid engine in this game". A stupid engine that beat the 4th or 5th strongest engine in the world?
"Game" is not the same as "endgame".
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
F. Bluemers
Posts: 868
Joined: Thu Mar 09, 2006 11:21 pm
Location: Nederland

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by F. Bluemers »

CMCanavessi wrote: Sun Nov 25, 2018 1:25 am
F. Bluemers wrote: Sun Nov 25, 2018 1:06 am
CMCanavessi wrote: Sat Nov 24, 2018 9:09 pm And how many of those "chess programs or humans who are not beginners" can beat Fire 7.1?
I think this is a bit unfair towards Uri. He sometimes seems to be trolling a bit , but he is spot on here.
The play in the endings is just extremely lousy.It can't hurt to point this out.
In the extreme case in otb meetings ab-searchers will just click a box when playing lc0 and have an easy win even though rating lists
would predict otherwise.
Best
Fonzy
We already know that, currently, the endgame is the weakest part in lc0... yet, Leela still won the game. And besides, he said (quote) "She clearly showed that she is a stupid engine in this game". A stupid engine that beat the 4th or 5th strongest engine in the world?
"Game" is not the same as "endgame".
Are you taking Uri's critisim personal?
FWIW these endgames look like soccer games where all members of the opponent team got a red card and the ball is played all over the field except were it should land,eventually winning is not an excuse …
Why not try to fix it?
Best
Fonzy
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by carldaman »

I raised a similar point.
Yes, it can achieve a winning endgame in grandmasterly fashion, but just when things should be super easy they become almost impossibly hard, taking the longest possible time to win such a simple ending. Could be a faulty TB implementation? And why are the depths getting lower and not higher when there are only 3 (three) pieces left on the board?

What do the experts think?
Then nabildanial replied with:
It is an artifact or a side-effect to the current test30 "TB rescoring" implementation. She can convert those endgames quite well, at least better than previous test runs at the same LR, but it takes a little bit of pieces reshuffling and just before the games approaching the 50-move draw she'll start converting those endgames.
http://talkchess.com/forum3/viewtopic.php?f=2&t=68963

I do hope they make it better because the way it is now is like a bad joke (after coming this far, to the developers' credit).
User avatar
CMCanavessi
Posts: 1142
Joined: Thu Dec 28, 2017 4:06 pm
Location: Argentina

Re: Super bullet chess: LC0 network 31510 vs Fire 7.1

Post by CMCanavessi »

F. Bluemers wrote: Sun Nov 25, 2018 2:07 am
CMCanavessi wrote: Sun Nov 25, 2018 1:25 am
F. Bluemers wrote: Sun Nov 25, 2018 1:06 am
CMCanavessi wrote: Sat Nov 24, 2018 9:09 pm And how many of those "chess programs or humans who are not beginners" can beat Fire 7.1?
I think this is a bit unfair towards Uri. He sometimes seems to be trolling a bit , but he is spot on here.
The play in the endings is just extremely lousy.It can't hurt to point this out.
In the extreme case in otb meetings ab-searchers will just click a box when playing lc0 and have an easy win even though rating lists
would predict otherwise.
Best
Fonzy
We already know that, currently, the endgame is the weakest part in lc0... yet, Leela still won the game. And besides, he said (quote) "She clearly showed that she is a stupid engine in this game". A stupid engine that beat the 4th or 5th strongest engine in the world?
"Game" is not the same as "endgame".
Are you taking Uri's critisim personal?
FWIW these endgames look like soccer games where all members of the opponent team got a red card and the ball is played all over the field except were it should land,eventually winning is not an excuse …
Why not try to fix it?
Best
Fonzy
Not personal at all, just pointing at an obvious trolling.
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls