Tuesday, June 30, 2020

Revolution in the millennium part 2

The possibilities are nowadays almost unlimited to improve at chess. With a few clicks of the mouse we can access countless games played at a high standard. A student can learn a lot by studying the good examples but does it also make sense to look at wrong examples? There is no agreement about that between trainers.

Even analyzing our own games already generated some discussions recently see my articles "Which games should I analyze? " part 1part 2 and part 3. Some young masters only do a blundercheck (15 minutes maximally) anymore of their games and prefer to spend more time at studying games of stronger players. Also for my classes I doubt it is a good idea to analyze games of my students. Sometimes I decide to include one of their games to discuss a specific new theme but generally I try to avoid it also because some of my students like to mock about someone other's mistakes.

Nevertheless amateurs make a lot of mistakes so mistakes are often much more important than finding complicated concepts or discovering new ideas. It is for a reason why we often say that the player whom made the penultimate mistake, wins the game. Therefore ignoring mistakes, doesn't seem to be for me the right choice. Some authors already understood this perfectly see e.g. swindels part 2. I also like that in the book "On the Origin of Good Moves" not only is covered what was known in the early years of the chess-history but also what the best players were still doing wrong.

Especially the first official world-champion Wilhelm Steinitz is targeted in the book. Tactically  but also strategically and positionally he gets harsh comments. Particularly Wilhem's theory that the king can take care of itself, is refuted convincingly by showing several failures of Wilhem. I think nowadays any experienced club-player would consider such risky strategy as nonsense but Wilhelm didn't hesitate to snatch a pawn even if this would mean that his king would have to stay for a longtime at the center. It is not a coincidence that a line of the kingsgambit with Ke2 got Wilhelm's name.

We should also not forget that Wilhem won many games with this risky strategy in his best years. Naturally it is not his fault that his opponents weren't strong enough to punish it. It is fully understandable to stick to something which worked before. I believe it is still possible to score points with such strategy even today against weaker players. However against a master it is a different game. I experienced that last in Cappelle La Grande. My opponent the strong French IM Chistophe Sochacki didn't know the opening and I thought to profit from it by winning a pawn but quickly regretted my decision.
[Event "Open Cappelle La Grande 2de ronde"] [Site "?"] [Date "2020"] [Round "?"] [White "Sochacki, C"] [Black "Brabo"] [Result "1-0"] [SetUp "1"] [FEN "r1bqkb1r/pp4pp/2pp4/5n2/8/2NQ4/PPP2PPP/R1B1KB1R w KQkq - 0 10"] [ECO "A04"] [WhiteElo "2475"] [BlackElo "2251"] [PlyCount "41"] [EventDate "2020"] 10. h4?! {(My opponent was already out of book since move 5 so he made up this interesting practical novelty at the board. Critical is Bd2 with very complex play.)} (10. Bd2! Be7 11. O-O-O O-O 12. f4 d5 13. g4 Nd6 14. Qh3 Bh4) 10... Qe7+? {(This is far too optimistic. I win a pawn in the next moves but I quickly regretted my decision. The most accurate is d5 here and black has slightly the better chances.)} 11. Be2 Nxh4? {(Consequently played but black is lost after this move. I never get my development anymore finalized after this move.)} (11... d5! 12. Bg5 Qf7 13. O-O-O Be7 14. Kb1 h6 15. Bc1 O-O!? 16. g4 Nd6) 12. Bd2 Bf5 13. Qd4 Ng6 14. O-O-O Qf6 15. Qa4 Qf7 16. Rhe1 Kd7 17. g4 Be6 18. f4 Bd5 19. Nxd5 Qxd5 20. f5 Ne5 21. Kb1 Qg2 22. Bf4 Kc8 23. Qb3 a5 {(Guess with which piece white is planning to give mate and then look at how the rest of the game enfolds.)} 24. a3 Kc7 25. Bxe5 dxe5 26. Rd7+ Kxd7 27. Qxb7+ Ke8 {(I will be mated after Kd6 but this is not really better.)} (27... Kd6 28. Rd1+ Kc5 29. b4+ axb4 30. axb4# {(Likely my opponent had foreseen this scenario when playing 24.a3.)}) 28. Qxa8+ Kf7 29. Bc4+ Ke7 30. g5 {(With this precise move it is mate in 7 so I resigned. One wrong decision and I was busted. It is a pity that my opponent left the tournament a couple of rounds later as that damaged a lot my final standing in the tournament.)} 1-0
I was annihilated in the game. White kept my king in the center on pain of huge material losses and demonstrated with the refined 24.a3 how hopeless my position already was. In other words masters know nowadays how to exploit a king stranded in the center. At least till shortly as recently we hear regularly another sound. With this we arrive to the essence of this article.

In the book "On the Origin of Good Moves" we get the theory of evolution presented. The improvement of chess has been a very slow process during the history. However in that case we are only talking about the human player. In the last decades there was a parallel evolution happening of engines which not only ran much more via jumps but also sometimes 100 times faster. Especially the introduction of neural networks opened a new world of chess which we never thought of as humans. Suddenly there was an engine playing hundred of points stronger positionally than any other one with knowledge built by playing games against itself in just a couple of hours.

A lot has been written already about it but I think one of these revolutionary changes has been largely neglected by most authors. Leela evaluates the safety of the king very differently compared to traditional engines like Stockfish. Last couple of months we didn't see much well played chess by humans but engines didn't suffer of the corona-virus and kept spoiling us with high-quality played games. The TCEC super-final of season 18 is still ongoing but in April we had already the fantastic TCEC super-final of season 17 which for the first time was even live commented by (top-) grandmasters at chess24.com. It is hard to choose between the plethora of games but below one is a great example of how Leela doesn't mind to omit castling.
[Event "TCEC Season 17 - Superfinal"] [Site "https://tcec-chess.com"] [Date "2020.04.17"] [Round "71.1"] [White "Stockfish 20200407DC"] [Black "LCZero v0.24-sv-t60-3010"] [Result "1/2-1/2"] [ECO "B00"] [WhiteElo "3851"] [BlackElo "3840"] [PlyCount "397"] 1. e4 Nc6 2. d4 d5 3. e5 Bf5 4. c3 e6 5. Be2 {(Leela chose Bd3 in the return-game.)} 5... f6 6. f4 g5 {(Activity supersedes safety of the king.)} 7. Bh5+ Kd7 8. fxg5 fxe5 9. Nf3 h6 10. g4 Be4 11. Nbd2 Bxf3 12. Nxf3 e4 13. Nh4 hxg5 14. Ng6 Rh7 15. O-O Bd6 16. Qe2 a6 17. Kg2 Rg7 18. Be3 Nge7 19. Nf8+ Qxf8 {(Another typical feature of Leela is that she estimates the queen less powerful than more traditional engines.)} 20. Rxf8 Rxf8 21. Rf1 Bf4 22. h4 Rh8 23. hxg5 Bxg5 24. Bxg5 Rxg5 25. Kf2 Nf5 26. Ke1 Nce7 27. Qf2 Rhxh5 28. gxh5 Rxh5 {(Black has only 2 knights and 2 pawns for the queen but all the pieces are harmoniously cooperating.)} 29. Kd1 Rh4 30. a3 Ng6 31. Kc1 e3 32. Qg1 Nf4 33. Kc2 e2 34. Re1 Rg4 35. Qf2 a5 36. b3 b6 37. Qf3 Rh4 38. a4 Kc8 39. Rxe2 {(Stockfisch detects the danger and pulls on time the emergency-break.)} 39... Nxe2 40. Qxe2 Re4 41. Qh5 Kb7 42. Qg6 Nd6 43. Kd2 Nf5 44. Qf7 Nd6 45. Qd7 Kb8 46. Qd8+ Kb7 47. Qd7 e5 48. Qe6 exd4 49. Qxd5+ Kb8 50. cxd4 Re8 51. Kc2 Rd8 52. Kd3 Nb7 53. Qb5 Rd6 54. Qe8+ Nd8 55. Kc4 Ka7 56. b4 axb4 57. d5 Nb7 58. Qe5 Nc5 59. a5 Na6 60. Qd4 Rg6 61. axb6+ Rxb6 62. d6 cxd6 {(This is a tablebase-draw but the engines get only access to the tablebases of maximum 6 pieces so the game still continues.)} 63. Qg7+ Kb8 64. Qf8+ Kb7 65. Qe7+ Kc6 66. Qe4+ Kc7 67. Qe7+ Kc6 68. Qe4+ Kc7 69. Qh7+ Kb8 70. Qg8+ Kb7 71. Qd5+ Ka7 72. Qf7+ Ka8 73. Kb3 Nc5+ 74. Kb2 Kb8 75. Kb1 b3 76. Qe7 Rc6 77. Qd8+ Kb7 78. Kb2 Rb6 79. Kb1 Ka6 80. Qc8+ Kb5 81. Qe8+ Kb4 82. Qe1+ Ka4 83. Qd2 Ka3 84. Qc1+ Ka4 85. Qd1 Kb4 86. Qc1 Na4 87. Qd2+ Kb5 88. Qd3+ Kc5 89. Qe3+ Kc4 90. Qc1+ Kb5 91. Qd2 Nc5 92. Qe2+ Kb4 93. Qd1 Ka3 94. Qc1+ Kb4 95. Qd1 Na4 96. Qd4+ Ka3 97. Qa1+ Kb4 98. Qd4+ Ka5 99. Qd5+ Nc5 100. Qd2+ Kb5 101. Qe2+ Kc6 102. Qe8+ Kb7 103. Qd8 Ka6 104. Qc8+ Kb5 105. Qf5 Rb7 106. Qf1+ Kc6 107. Qf8 d5 108. Qf6+ Kb5 109. Qf1+ Kb4 110. Qe1+ Ka4 111. Qc1 Rb5 112. Qd1 Ne4 113. Kb2 Ka5 114. Qa1+ Kb6 115. Qa3 Nc5 116. Qa8 Kc7 117. Qe8 Rb7 118. Qe7+ Kc6 119. Qe8+ Rd7 120. Kc3 Kb7 121. Qe1 Kc6 122. Qe8 Kc7 123. Qe1 Kd6 124. Qg3+ Kc6 125. Qe5 Rd8 126. Qf6+ Rd6 127. Qf8 Kb5 128. Qf1+ Kb6 129. Qf4 Kc6 130. Qf8 Kb5 131. Qf1+ Kb6 132. Kb4 Rc6 133. Qb5+ Kc7 134. Qa5+ Kb7 135. Kc3 d4+ 136. Kb2 d3 137. Kc3 Kc8 138. Qb5 Kc7 139. Kd2 Kd6 140. Kc3 Kc7 141. Kb2 Kd6 142. Qb8+ Rc7 143. Qb6+ Kd7 144. Qh6 Kc8 145. Qg5 Kb7 146. Qd5+ Kb6 147. Qd8 Kb7 148. Qd6 Rc8 149. Ka3 Rc6 150. Qd5 Kb6 151. Qd8+ Rc7 152. Qd6+ Kb7 153. Kb2 Rc8 154. Ka3 Rc6 155. Qe5 Ka7 156. Qe7+ Ka6 157. Qf8 Ka7 158. Qe7+ Ka6 159. Qf8 Kb7 160. Qf2 Ka7 161. Kb2 Ka6 162. Qf8 Kb6 163. Qd8+ Ka6 164. Qa8+ Kb6 165. Kc3 Rc7 166. Qb8+ Kc6 167. Qf8 Kb7 168. Qf5 Kb6 169. Qf2 Kb5 170. Qh4 Rc8 171. Qb4+ Kc6 172. Kc4 Kd6 173. Kd4 Ke6 174. Qb6+ Kd7 175. Qb5+ Kd6 176. Qb6+ Kd7 177. Kd5 Rc7 178. Qa5 Kd8 179. Qa8+ Rc8 180. Qa1 d2 181. Qd4 Rc7 182. Qh8+ Ke7 183. Qg7+ Ke8 184. Qe5+ Kd7 185. Qf5+ Kd8 186. Qf8+ Kd7 187. Qf5+ Ke7 188. Qg5+ Ke8 189. Qg6+ Kd8 190. Qg8+ Kd7 191. Qg4+ Ke8 192. Qg6+ Rf7 193. Qg8+ Ke7 194. Qg5+ Rf6 195. Qg7+ Rf7 196. Qe5+ Kd8 197. Qb8+ Kd7 198. Qd6+ Ke8 199. Kxc5 1/2-1/2
Besides I can recommend to people willing to see more and learn more of such games, to create their own matches with Leela at home. Last year I already did that a few times see my article testing chess-engines and also during the corona-crisis I again repeated this enjoyable activity. Also in my own organized rapidmatches Leela proofed that it isn't afraid of keeping the king in the center. I've again selected one game which demonstrates this surprising feature very well.
[Event "LAPTOP-CD6JDPMJ, Rapid 15m+10s"] [Site "LAPTOP-CD6JDPMJ"] [Date "2020.03.13"] [Round "26"] [White "Stockfish 11 64"] [Black "Lc0 v0.24.0+git.2449d9e"] [Result "1/2-1/2"] [ECO "B32"] [PlyCount "304"] 1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 e5 5. Nb5 d6 6. c4 Be7 7. N1c3 a6 8. Na3 Be6 9. Nc2 Bg5 10. Be2 Bxc1 11. Rxc1 {(Leela chose Qxc1 in the first leg.)} 11... Qg5 12. O-O Rd8 13. Nd5 Nf6 14. Nc7+ Kd7 15. Nxe6 fxe6 16. b4 Nxe4 {(Wilhelm Steinitz would've been proud about Leela. Black won a pawn and the king stays in the center.)} 17. Qd3 Nf6 18. Bf3 Qf5 19. Bxc6+ bxc6 20. Qa3 Ra8 21. Ne3 Qe4 22. Rcd1 Ke7 23. Rfe1 Qg6 24. b5 c5 25. Qa5 Qe8 26. Qd2 Qb8 27. a4 Rd8 28. Qe2 Kf7 29. Ng4 Nxg4 30. Qxg4 axb5 31. axb5 d5 32. h3 e4 33. cxd5 exd5 34. Qf5+ Kg8 35. Rxd5 Qxb5 36. Rxc5 Qb6 37. Rxe4 {(Black returned 2 pawns so the king was able to evacuate out of the center. The endgame is easy to defend for Leela.)} 37... Rf8 38. Qe5 Rae8 39. Qd5+ Kh8 40. Rxe8 Rxe8 41. h4 Qd8 42. Qc4 h6 43. h5 Qb8 44. g4 Rf8 45. Qe2 Qa8 46. Re5 Qf3 47. Qxf3 Rxf3 48. Kg2 Rf4 49. f3 Ra4 50. Re8+ Kh7 51. Re7 Ra2+ 52. Kh3 Ra3 53. Kg3 Kg8 54. Rd7 Kh7 55. Rb7 Kh8 56. Re7 Kh7 57. Rb7 Kh8 58. Kf4 Kh7 59. Ke4 Ra4+ 60. Kf5 Ra3 61. Ke4 Ra4+ 62. Ke3 Ra3+ 63. Kf2 Ra2+ 64. Kg3 Ra3 65. Rc7 Kh8 66. Kf2 Kh7 67. Kg3 Kh8 68. Re7 Kh7 69. Kg2 Kh8 70. Kf2 Kh7 71. Kg3 Kh8 72. Re8+ Kh7 73. Rd8 Rb3 74. Rd7 Ra3 75. Kf4 Ra1 76. Kf5 Ra3 77. f4 Ra1 78. Rd5 Rb1 79. Ra5 Rb2 80. Rd5 Rb1 81. Rd6 Ra1 82. Rd3 Ra5+ 83. Ke4 Ra1 84. Rd4 Rg1 85. Kf5 Ra1 86. Rb4 Ra2 87. Rb5 Ra1 88. Rc5 Rb1 89. Rd5 Ra1 90. Rd8 Rb1 91. Rd7 Rf1 92. Rb7 Ra1 93. Rb5 Ra2 94. Ke6 Rg2 95. Kf5 Rf2 96. Rb8 Ra2 97. Rb5 Ra1 98. Rc5 Ra2 99. Re5 Ra1 100. Re2 Rb1 101. Rd2 Ra1 102. Ke4 Rg1 103. Kf5 Ra1 104. Rd8 Rb1 105. Rd6 Ra1 106. Rd4 Rb1 107. Ra4 Rb2 108. Ra3 Rb1 109. Ke4 Rg1 110. Kf5 Rb1 111. Ra2 Rc1 112. Ke4 Rg1 113. Kf3 Rf1+ 114. Kg3 Rg1+ 115. Kh3 Rf1 116. f5 Rf3+ 117. Kg2 Rf4 118. Kg3 Rb4 119. Re2 Rb3+ 120. Kf4 Rb4+ 121. Kg3 Rb3+ 122. Kh4 Kg8 123. Re4 Ra3 124. Rb4 Kf7 125. Rb6 Ra1 126. Rb4 Kg8 127. Kg3 Kf7 128. Rb7+ Kf6 129. Rb6+ Kf7 130. Re6 Ra3+ 131. Kh4 Rf3 132. Rb6 Ra3 133. Rb8 Ra1 134. Rc8 Rb1 135. Rc4 Ra1 136. Kg3 Ra6 137. Kf4 Ra1 138. Rc7+ Kf6 139. Rc6+ Kf7 140. Re6 Ra4+ 141. Kg3 Ra3+ 142. Kh4 Rf3 143. Re4 Ra3 144. Rb4 Ra1 145. Kg3 Kg8 146. Rd4 Kf7 147. Kf4 Kf6 148. Rd6+ Kf7 149. Rb6 Ra4+ 150. Kg3 Ra3+ 151. Kf2 Ra4 152. Rb7+ Kf6 1/2-1/2
It didn't take long before humans started to pick up this new insights from Leela and implement it in their own practice. I clearly notice this in my own opening-choices in which I much more go for maximum activity of the pieces even at the expense of the safety of my own king. A nice example of this is the opening which I played against the Belgian expert Tijs Cocquyt in Cappelle La Grande.  Optically it looks very dodgy for black as even some grandmasters already chose to play the position with white. However the engines show at the other hand that black is doing more than fine.
[Event "Open Cappelle La Grande 4de ronde"] [Site "?"] [Date "2020"] [Round "?"] [White "Cocquyt, T"] [Black "Brabo"] [Result "*"] [ECO "A80"] [WhiteElo "2115"] [BlackElo "2251"] [PlyCount "18"] 1. d4 f5 2. Bg5 h6 3. Bh4 g5 4. e4 Bg7 {(As remarked earlier already by the Belgian FM Arno Bomans, I do indeed follow closely all the recent developments of the engines. Therefore my opening-choices are also largely connected to what Stockfish and Leela recommend. Activity is the key here even if this means that I need to sacrifice some of the safety of my king.)} 5. Bg3 d6 6. exf5 Bxf5 7. Qh5+?! {(This line has been played by already 2 grandmasters: Anthony Wirig and Alexander Dgebuadze. Scoring 1,5/2 they can't complain as they were too optimistic about whites chances.)} 7... Kd7 8. Qf3 e6 9. Bd3 (9. Nc3 Nc6 10. O-O-O Nxd4 11. Qxb7 Qb8 12. Qa6 Qb6 13. Qa4+ {(Till here we followed the game between Anthony Wirig and Alf Andersen played in 2008.)} 13... Kd8 {(Stockfish and Leela consider the activity of the black pieces more important than the stranded king in the center.)}) (9. Na3 Nc6 10. c3 Nge7 11. Qd1 {(This line was played in the game between Alexander Dgebuadze and Stefan Beukema.)} 11... h5 {(Leela doesn't lose time and shows that white also has problems with the safety of his king.)}) 9... Ne7 {(I played in my game the weaker Bxd3 still maintaining a small edge.)} *
Another example of which I was impressed was the march of the black king played by the Belgian FM Hendrik Ponnet in our most recent mutual game. In a board full of pieces he decides to use his own king in the frontline so his other pieces can take up other duties. Later I read a report about the chess-weekend for adults only, earlier this year organized by Schaakinitiatief Vlaanderen in which was stated that Hendrik gave a presentation about the currently existing tools online to play and learn chess. So I suspect Hendrik also has experience with Leela or other neural networks.
[Event "Interclub Deurne - KGSRL"] [Site "?"] [Date "2020"] [Round "?"] [White "Brabo"] [Black "Ponnet, H"] [Result "1/2-1/2"] [SetUp "1"] [FEN "3r1rk1/1b2p1bp/p3n1p1/P1n2p2/1NN1p3/4B3/1PP2PPP/3RRBK1 b - - 0 20"] [ECO "B06"] [WhiteElo "2296"] [BlackElo "2288"] [PlyCount "87"] 20... Kf7 {(I didn't expect this move at all. Later I read that Hendrik gave a presentation about online software for chess in the training-weekend for adults organized by Schaakinitiatief Vlaanderen so I assume Hendrik has also some experience already with Leela.)} 21. c3 g5 22. Nb6 f4 23. Bc1 Nc7 24. Bc4+ e6 25. Nc2 Nb5 26. Nb4 Be5 27. Nc2 Kf6 28. Bf1 Kf5 {(Wow, the king supports the pawns in the front-line so other pieces can take up other duties. I am steamrolled completely. Leela evaluates this position already as very good for black and is clearly more optimistic than Stockfish.)} 29. Nc4 Bc7 30. Nb4 Nb3 31. Nc2 e5 32. Nb4 Nxa5 33. Nxa5 Bxa5 34. Nc2 Bb6 35. Bc4 Nc7 36. b3 Nd5 37. Bb2 e3 38. fxe3 Nxe3 39. Nxe3+ Bxe3+ 40. Kf1 Bc5 41. b4 Bb6 42. b5 axb5 43. Bxb5 e4 44. Ba3 Rf7 45. c4 f3 46. c5 e3 47. gxf3 Rxd1 48. Rxd1 Bxf3 49. Be2 Bxe2+ 50. Kxe2 Bc7 51. Kxe3 Bxh2 52. Kf3 g4+ 53. Kg2 Be5 54. Bb4 Ke6 55. Be1 Ra7 56. Rd2 Ra3 57. Bf2 Rc3 58. Re2 Kf5 59. Ra2 h5 60. Re2 Bf4 61. Ra2 g3 62. Be1 Rxc5 63. Bxg3 Bxg3 1/2-1/2
This doesn't mean that Wilhelm Steinitz was right however. There is a big difference between the type of positions which Wilhelm obtained and the one Leela likes to play with the king in the center. Activity is the key here. Wilhelm captured material but very often got himself into terribly passive positions in which he could only hope that his opponent wouldn't have the tactical skills to punish it. That is for sure not the case in the positions chosen by Leela as she always can create very active counterplay. A last important remark is that we shouldn't forget that we humans don't have the calculating power of engines. Not rarely a position is theoretically fine but we as humans still go wrong as we are not able to find the best moves with the king in the center.


No comments:

Post a Comment

Note: Only a member of this blog may post a comment.