Difference between revisions of "TCEC Season 18 This is not a SuFi Bonus Game 39: LeelaCPU-Stockfish"

From TCEC wiki
Jump to: navigation, search
(Created page with "<!-- {{TCEC Game Info|season=18|div=notasufi|divname=This is not a SuFi Bonus|game=39|eco=A00|result=0-1|lichess-url=https://lichess.org/NlyX5cNC}} --> {{DEFAULTSORT:TCEC Seas...")
 
m
 
(5 intermediate revisions by the same user not shown)
Line 1: Line 1:
<!-- {{TCEC Game Info|season=18|div=notasufi|divname=This is not a SuFi Bonus|game=39|eco=A00|result=0-1|lichess-url=https://lichess.org/NlyX5cNC}} -->
+
{{TCEC Game Info|season=18|div=notasufi|divname=This is not a SuFi Bonus|game=39|eco=A00|result=0-1|lichess-url=https://lichess.org/NlyX5cNC}}
 
{{DEFAULTSORT:TCEC Season 18 This is not a SuFi Game 039}}
 
{{DEFAULTSORT:TCEC Season 18 This is not a SuFi Game 039}}
  
WORK IN PROGRESS
+
The first Grob game of the Season 18 "This is not a SuFi" bonus event. Here, Stockfish and Leela played openings from Superfinals of past seasons following the format of playing the openings from both sides. This opening is from Season 12 Superfinal between Stockfish and Komodo, games [https://www.tcec-chess.com/archive.html?season=12&div=sf&game=45 45] and [https://www.tcec-chess.com/archive.html?season=12&div=sf&game=46 46]. Both these games ended also in black wins, as well as the second Grob game in the "This is not a SuFi" bonus event <ref>[[TCEC Season 18 This is not a SuFi Bonus Game 40: Stockfish-LeelaCPU]]</ref>.
  
http://talkchess.com/forum3/viewtopic.php?f=2&t=72084&hilit=grob&start=410
+
The time control in the bonus event was 2 hours with 10 seconds increment per move. <br clear=all>
  
 +
<br clear=all>
 
= Game overview =
 
= Game overview =
 +
 +
{{Chess Pos Float|fen=1r1qkb1r/p2nnppp/Q1p5/3p4/3P2b1/2N5/PP2PPBP/R1B1K1NR|prevmove=9. Qa6}}
  
 
'''1. g4'''
 
'''1. g4'''
Line 14: Line 17:
 
For the opening move, the current SF-dev evaluation is at around -1.1 at depth 56 (330B nodes) and the Leela T42810 evaluation is at expected score of around 33%. Contrasting, the evaluation for the common first moves are at around +0.5 for Stockfish and at around 53-55% for Leela, respectively. The opening does not do any better in the opening database, black winning a bit more than half the time. Needless to say, 1. g4 is not a great opening move by any standard.
 
For the opening move, the current SF-dev evaluation is at around -1.1 at depth 56 (330B nodes) and the Leela T42810 evaluation is at expected score of around 33%. Contrasting, the evaluation for the common first moves are at around +0.5 for Stockfish and at around 53-55% for Leela, respectively. The opening does not do any better in the opening database, black winning a bit more than half the time. Needless to say, 1. g4 is not a great opening move by any standard.
  
The opening has been recently debated at the TalkChess.com forums using deep engine analysis, the question being whether it is objectively still drawn for white.<ref>'''1.g4 opening is losing?''' http://talkchess.com/forum3/viewtopic.php?f=2&t=72084</ref> The most recent studies claim that Stockfish with depth 60 analysis would not be able hold the draw against a correspondence-level preparation by black, favoring the argument that 1. g4 is losing against best play.
+
The opening has been recently debated at the TalkChess.com forums using deep engine analysis, the question being whether it is objectively still drawn for white.<ref>''1.g4 opening is losing?'' http://talkchess.com/forum3/viewtopic.php?f=2&t=72084</ref> The most recent studies claim that Stockfish with depth 60 analysis would not be able hold the draw against a correspondence-level preparation by black, favoring the argument that 1. g4 is losing against best play.
  
 
In the game, black responds with the most principled move, attacking the g4-pawn immediately.
 
In the game, black responds with the most principled move, attacking the g4-pawn immediately.
Line 20: Line 23:
 
'''1... d5 2. Bg2 Bxg4 3. c4'''
 
'''1... d5 2. Bg2 Bxg4 3. c4'''
  
Fritz gambit makes the Grob
+
The Grob gambit: Fritz gambit line. Leela plays as Grob is arguably supposed to be played, creating some early chaos on the board trying to trick the opponent. But this line is not really a true gambit line, as it will be hard for black to keep the extra pawn. This said, exchanging the g-pawn to a central or queenside pawn is not exactly a great idea, should black be not fooled by the Grob tricks.
gambit more like a temporary pawn sacrifice. It will be difficult for black to
+
 
keep the pawn advantage. That said, exchanging g-pawn to a queen-side pawn is
+
'''3... c6 4. Qb3'''
not exactly great.}  3... c6 { [%csl Gb7,Gc6,Gd5]}  4. Qb3 { [%csl Gc6][%cal
+
 
Gb3b7,Gg2a8] Leela plays the Grob opening in an aggressive manner, trying
+
White plays the Grob opening in an aggressive manner, attacking the b7-pawn. This position has been seen once in a rapid game at the super-GM level.<ref>Chess.com PRO League Group Stage, Round 10: ''Nakamura-Mamedyarov'' ½-½, March 2019 (rapid 15m+2spm). https://www.chessgames.com/perl/chessgame?gid=1963060</ref> In that game, black played a bit more conservatively protecting the pawn with the queen but also relinquishing a bit of the advantage by not keeping the center: 4... Qd7 5. cxd5 Nf6 6. Nc3 cxd5 7. Nxd5 Nc6 and the damage to white's position was less than in this game.
posing some immediate problems for black for the pawn.}  4... e6 { With e6,
+
 
black keeps the white fianchetto bishop out of play.}  ( { In
+
In the game, black played a sharper move, maintaining the center while giving up the b7-pawn. The resulting semi-Slav central pawn structure severely restricts white's light-square bishop, questioning the merits of white's opening moves.
Nakamura-Mamedyarov, https://www.chessgames.com/perl/chessgame?gid=1963060 (
+
 
TC: 15m+2spm ) : } 4... Qd7 5. cxd5 Nf6 6. Nc3 cxd5 7. Nxd5 Nc6 )( { The
+
'''4... e6 5. Qxb7 Nd7 6. Nc3'''
engines' second choice } 4... Nf6 5. Qxb7 Nbd7 )5. Qxb7 Nd7 6. Nc3 ( { The c6
+
 
pawn was not freeing, and taking it can be dangerous, as black would get more
+
The c6 pawn was poisoned and taking it would have been dangerous, as black would get some dangerous initiative by better development: 6. Qxc6? Rc8 7. Qa6 Rxc4.
initiative } 6. Qxc6 $2 Rc8 7. Qa6 Rxc4 8. Na3 ( 8. Nc3 Qc7 { The white queen
+
 
would start to feel stranded, there is the threat of e4 for winning the bishop
+
'''6... Ne7 7. cxd5 exd5 8. d4 Rb8 9. Qa6''' ''(diagram)''
on c1 with no counterplay to speak of.}  )8... Rc8 )6... Ne7 7. cxd5 exd5 8. d4
+
 
Rb8 9. Qa6 ( { Not a free pawn. Black either gets the d4-pawn in exchange, or
+
The theme of the poisoned pawn repeats: Taking a7 would come with a hefty price, as black would get the d4-pawn in exchange. For example: 9. Qxa7? Nf5 10. Nf3 Ra8 11. Qb7 Bxf3 12. exf3 Nxd4. Here white is probably objectively losing against black's initiative.<br clear=all>
worse } 9. Qxa7 $2 Nf5 10. Nf3 ( 10. e3 Nh4 11. Kf1 )10... Ra8 11. Qb7 Bxf3 12.
+
 
exf3 ( 12. Bxf3 $2 Nxd4 { fork threat that cannot be really parried}  13. Bf4 (
+
{{Chess Pos Float|fen=2q2rk1/p3npp1/2rb3p/2Ppn2P/N7/1P1Q1N2/P3PP2/R1B1K2R|prevmove=20... Ne5}}
13. O-O) 13... Nc5 14. Qc7 Qxc7 15. Bxc7 Nc2+ )12... Nxd4 )9... Rb6 10. Qd3 Ng6
+
 
11. h3 { Typical for Leela, fighting for counterplay11... Be6 12. Nf3 Bd6
+
'''9... Rb6 10. Qd3 Ng6 11. h3'''
13. h4 h6 $146 { https://lichess.org/NlyX5cNC#25
+
 
Black prevents white from
+
As is typical for Leela, white fights for some counterplay.
playing Ng5 immediately.}  ( { In Ohtake-Parsons (Correspondence game from
+
 
2014): ½-½ } 13... O-O 14. b3 Qf6 15. Kf1 h6 16. h5 Bf5 17. Qd1 Nf4 )14. Bh3 {
+
'''11... Be6 12. Nf3 Bd6 13. h4 h6'''(N)<ref>Lichess masters database. https://lichess.org/NlyX5cNC#25</ref>
White forces the exchange of the bad bishop and tries to mount some counterplay
+
 
on the open king-side. But it is a bit double-edged, as it also exchanges a
+
The last remaining game in the opening database was a correspondence game from 2014.<ref>''Ohtake-Parsons'' ½-½, 2014. https://lichess.org/9qWOVKzC</ref> The game continued with 13... O-O 14. b3 Qf6 15. Kf1 h6 16. h5 Bf5 17. Qd1 Nf4 with white being in time to mount defenses. In the game, white was able to exchange enough pieces to hold against black's pressure.
defender and puts a stop on hopes of king safety.}  14... Bxh3 15. Rxh3 O-O {
+
 
[%cal Gc1d2,Ga1g1]}  16. h5 Ne7 17. Rh1 c5 { Logical for black to start opening
+
Here black plays in a bit more direct manner.
the position to further take advantage of better development. This also creates
+
 
the threat of c4, which white immediately discourages.}  18. b3 Qc8 19. Na4 Rc6
+
'''14. Bh3'''
20. dxc5 Ne5 { The intermezzo allows black to develop the bishop to a better
+
 
square with tempo. The tempo is used to maneuver Rc6-e6 for putting pressure on
+
White forces the exchange of the bad bishop and tries to mount some counterplay on the open kingside files. But this plan is a bit double-edged, as it also exchanges a defender, and consequently, the white king will have difficulties in finding safe squares.
the e-file.}  ( 20... Nxc5 21. Nxc5 Rxc5 )21. Nxe5 Bxe5 22. Rb1 Re6 23. Bd2 Re8
+
 
24. Rc1 Nc6 { Now that the rooks are doubled on the semi-open e-file, white's
+
'''14... Bxh3 15. Rxh3 O-O 16. h5 Ne7 17. Rh1 c5'''
king safety is being seriously questioned. Black has multiple additional
+
 
resources to add to attack, as the light pieces are centralized and the
+
It is logical for black to start opening the position to take advantage of better development and white's king safety issues. 17... c5 also creates the immediate threat of c4, which white promptly discourages.
untouchable d5-pawn controls important squares, and is ready to advance. White
+
 
king sidesteps the rook battery.}  25. Kf1 ( { Taking the d5-pawn would be a
+
'''18. b3 Qc8 19. Na4 Rc6 20. dxc5 Ne5''' ''(diagram)''
huge mistake. For example: } 25. Qxd5 Rd8 26. Qg2 Qd7 27. Rd1 Bf4 28. e3 Qd3
+
 
29. Nb2 Qc2 30. Nc4 Nd4 { White's position is beginning to crumble}  )25... Bd4
+
An important intermezzo that allows black to develop the remaining bishop to a better square with tempo. This tempo is used to maneuver the rook on c6 to e6 with a plan of doubling the rooks on the e-file.<br clear=all>
26. Nc3 Rf6 27. Rh2 Re5 28. Nb5 Qg4 { [%cal Ga2a6,Gb3b6][%csl Ga2,Gb3,Ge2,Gf2]}
+
 
29. Nxd4 { White tries to relieve some of the pressure, but there white's
+
{{Chess Pos Float|fen=6k1/p4pp1/2n2r1p/1NPpr2P/3b2q1/1P1Q4/P2BPP1R/2R2K2|prevmove=28... Qg4}}
defenses are getting overloaded.}  ( { The threat: } 29. a4 Rxh5 30. Rxh5 Rxf2+
+
 
31. Ke1 Qg1# )29... Nxd4 30. Rg2 Qh4 31. Qg3 { White enforces the queens off
+
'''21. Nxe5 Bxe5 22. Rb1 Re6 23. Bd2 Re8 24. Rc1 Nc6'''
the board with the immediate Qxg7# threat. However, black wins material in the
+
 
tactics afterwards}  ( { The threat by black } 31. Re1 Rxf2+ 32. Rxf2 Qh1#
+
Now that the rooks are doubled on the semi-open e-file, white's king safety is being seriously questioned. Black has multiple additional resources to add in the attack, as the light pieces are nicely centralized. The pawn on d5 is currently untouchable. It controls some important central squares and is ready to advance. The white king sidesteps the rook battery.
)31... Qxg3 32. Rxg3 Nxe2 { Wins a pawn and forks the rooks}  33. Rgc3 Nxc3 34.
+
 
Rxc3 Rc6 { [%csl Gh5][%cal Gd5d3]}  35. b4 Rxh5 36. Ke2 { [%cal Gh5h4,Gh4c4]}
+
Note that taking the d5-pawn would be a big mistake. For example: 25. Qxd5? Rd8 26. Qg2 Qd7 27. Rd1 Bf4 28. e3 Qd3 29. Nb2 Qc2 30. Nc4 Nd4. White's position would collapse soon.
36... f6 { Makes room for Kf7-e6 and supports g5 later on, and goes for a sharp
+
 
pawn race}  ( { Here black had multiple ways to convert the game. Perhaps the
+
'''25. Kf1 Bd4 26. Nc3 Rf6 27. Rh2 Re5 28. Nb5 Qg4''' ''(diagram)'' '''29. Nxd4'''
easiest would have been to stop all white queenside counterplay and then start
+
 
pushing with the kingside pawns: } 36... d4 37. Rc4 Rh3 38. Rxd4 Ra3 39. Rd8+
+
White tries to relieve some of the pressure, but the white defenses are already getting overloaded.<br clear=all>
Kh7 40. Be3 a6 { As now the white qside pawns are blockaded, at least for the
+
 
time being, black would have plenty of time to start pushing the king-side
+
{{Chess Pos Float|fen=6k1/p4pp1/2r4p/2Pp3r/1P6/2R5/P2BKP2/8|prevmove=36. Ke2}}
pawns.}  )37. Kd3 Rc8 38. Be3 Rh4 39. f4 { Natural move to prevent the black
+
 
rook attacking b4. A similarly natural response would be to either play g5,
+
'''29... Nxd4 30. Rg2 Qh4 31. Qg3'''
intending to create connected passers, or Rh1 to start attacking the white
+
 
pawns from behind and to make room for the h-pawn push. However, black goes for
+
White enforces the queens off the board with the immediate 32. Qxg7# threat. However, black wins material in the tactics afterwards.
a faster conversion with the rush of getting the h-pawn promoted.}  39... h5 (
+
 
39... g5 40. fxg5 hxg5 )40. c6 Rh1 41. b5 h4 42. Bxa7 h3 43. b6 h2 44. b7 Re8
+
The threat prompting the queen exchange was 31... Rxf2+ 32. Rxf2 Qh1#
45. Kc2 { [%cal Ge8e2,Gh1c1]}  ( { Even if the pawns on 7th look menacing, and
+
 
even winning, white is simply not in time } 45. c7 Rhe1 46. b8=Q h1=Q 47. Kd4
+
'''31... Qxg3 32. Rxg3 Nxe2'''
Qe4+ 48. Kc5 Rb1 49. a3 Qe1 50. Rb3 Qe7+ 51. Kc6 Qe6+ 52. Kc5 Rc1+ 53. Kb5 Qd7+
+
 
54. Kb6 Rc6+ 55. Kb7 Rxb8+ 56. Bxb8 d4 57. f5 Rc1 { White can't both stop the
+
Black wins a pawn and forks the rooks.
d-pawn advancing and Qc6+ mating}  58. a4 Qc6+ 59. Ka7 Qxa4+ 60. Kb7 Qc6+ 61.
+
 
Kc8 Re1 )45... Rc1+ { Cute little tactic to exchange to rook for a queen.}  46.
+
'''33. Rgc3 Nxc3 34. Rxc3 Rc6 35. b4 Rxh5 36. Ke2''' ''(diagram)'' '''f6'''
Kxc1 h1=Q+ 47. Kb2 d4 48. Bxd4 Qd5 { The dangerously looking bc pawns are now
+
 
dynamically blockaded. Any attempt to pust them would lose them. Black uses the
+
Black makes room for Kf7-e6 and supports g5 with the h-pawn push. This also goes for a sharp pawn race.
better mobilitity of the heavy pieces to overrun white and with
+
 
material-winning threats.}  49. Be3 Kf7 50. a4 ( 50. c7 Qxb7+ )50... Rh8 51.
+
Here black had multiple ways to convert the game. Perhaps the simplest would have been to put a halt to all white queenside counterplay and only then begin pushing with the kingside pawns: 36... d4 37. Rc4 Rh3 38. Rxd4 Ra3 39. Rd8+ Kh7 40. Be3 a6. As now the white queenside pawns are blockaded for the time being, black would have plenty of time to start pushing the kingside pawns.<br clear=all>
Rc2 Qe4 { Forking the bishop and the a4-pawn.}  52. Bd2 Qxa4 53. Rc3 ( 53. c7
+
 
Qb5+ )53... Qd4 54. Be3 ( 54. Kc2 Rh2 55. b8=Q Qxd2+ 56. Kb3 Qb2+ 57. Kc4 Qxb8
+
{{Chess Pos Float|fen=4r1k1/1P4p1/2P2p2/3q4/3B1P2/2R5/PK6/8|prevmove=48... Qd5}}
)54... Rh2+ 55. Kb3 Qd5+ 56. Rc4 Qd3+ ( 56... Qb5+ 57. Rb4 Rb2+ ( 57... Qd3+
+
 
58. Ka4 Ra2# )58. Kxb2 Qxb4+ 59. Ka2 Qa4+ ( 59... Ke6 60. Ba7 Qa4+ 61. Kb2 Qxa7
+
'''37. Kd3 Rc8 38. Be3 Rh4 39. f4'''
)60. Kb2 )57. Kb4 ( 57. Rc3 Qb5+ 58. Ka3 Rb2 { M4}  )57... Rb2+ 58. Kc5 Ke6 {
+
 
Threat: Qd6#}  59. Rb4 Qd5+ 60. Kb6 Rxb4+ 61. Ka7 Ra4+ 62. Kb6 Qa5#  0-1
+
Natural move to prevent the black rook attacking b4. A similarly natural response would be to either play g5, intending to create connected passers, or Rh1 to start attacking the white pawns from behind and to make room for the h-pawn push. However, black goes for a faster conversion with the rush of getting the h-pawn promoted.
 +
 
 +
'''39... h5 40. c6 Rh1 41. b5 h4 42. Bxa7 h3 43. b6 h2 44. b7 Re8 45. Kc2'''
 +
 
 +
Even if the pawns on c6 and b7 look menacing or even winning, white is simply not in time to decide the game with the pawns queening. An example line: 45. c7 Rhe1 46. b8=Q h1=Q 47. Kd4 Qe4+ 48. Kc5 Rb1 49. a3 Qe1 50. Rb3 Qe7+ 51. Kc6 Qe6+ 52. Kc5 Rc1+ 53. Kb5 Qd7+ 54. Kb6 Rc6+ 55. Kb7 Rxb8+ 56. Bxb8 d4 57. f5 Rc1. White cannot both stop the d-pawn advancing and Qc6+ mating.
 +
 
 +
'''45... Rc1+'''
 +
 
 +
A cute little tactic to exchange the rook for a queen.
 +
 
 +
'''46. Kxc1 h1=Q+ 47. Kb2 d4 48. Bxd4 Qd5''' ''(diagram)''
 +
 
 +
Black gave up the d-pawn to dynamically blockade the b/c pawns. Any attempt to push either of them would lose them. The rest is now technique, as black uses the better mobility of the heavy pieces to overrun white with material-winning threats.<br clear=all>
 +
 
 +
{{Chess Pos Float|fen=8/1P4p1/1KP1kp2/q7/r4P2/4B3/8/8|prevmove=62... Qa5#}}
 +
 
 +
'''49. Be3 Kf7'''
 +
 
 +
To exemplify the blockade, 50. c7 here would be simply met with 50... Qxb7+.
 +
 
 +
'''50. a4 Rh8 51. Rc2 Qe4'''
 +
 
 +
The queen forks the bishop and the a4-pawn. Naturally the bishop is move from the harm's way.
 +
 
 +
'''52. Bd2 Qxa4'''
 +
 
 +
53. c7 here would be met with 53... Qb5+
 +
 
 +
'''53. Rc3 Qd4 54. Be3 Rh2+ 55. Kb3 Qd5+ 56. Rc4 Qd3+ 57. Kb4 Rb2+ 58. Kc5 Ke6'''
 +
 
 +
The simple threat is now 59... Qd6#. This prompts white to give up the rook.
 +
 
 +
'''59. Rb4 Qd5+ 60. Kb6 Rxb4+ 61. Ka7 Ra4+ 62. Kb6 Qa5# 0-1''' ''(diagram)''
 +
 
 +
Black wins by mating.<br clear=all>
  
 
= References =
 
= References =
 
<references />
 
<references />

Latest revision as of 18:44, 16 May 2020

Game Season 18, This is not a SuFi Bonus, Game 39
Result 0-1
ECO A00
Links TCEC archive
Lichess



The first Grob game of the Season 18 "This is not a SuFi" bonus event. Here, Stockfish and Leela played openings from Superfinals of past seasons following the format of playing the openings from both sides. This opening is from Season 12 Superfinal between Stockfish and Komodo, games 45 and 46. Both these games ended also in black wins, as well as the second Grob game in the "This is not a SuFi" bonus event [1].

The time control in the bonus event was 2 hours with 10 seconds increment per move.


Game overview

    
    
    
    
    
    
    
    
   
  
      
       
      
       
  
   
 ♜ ♛♚♝ ♜
♟  ♞♞♟♟♟
♕ ♟     
   ♟    
   ♙  ♝ 
  ♘     
♙♙  ♙♙♗♙
♖ ♗ ♔ ♘♖
Position after 9. Qa6


1. g4

End of the opening book. The Grob opening (1. g4) is rated by the strongest engines as the worst first move. It is sometimes seen by the titled players as a surprise blitz weapon, and in rapid games to a lesser degree. But the opening is practically never seen in a serious game with classical time controls, as its practical merits are hard to justify.

For the opening move, the current SF-dev evaluation is at around -1.1 at depth 56 (330B nodes) and the Leela T42810 evaluation is at expected score of around 33%. Contrasting, the evaluation for the common first moves are at around +0.5 for Stockfish and at around 53-55% for Leela, respectively. The opening does not do any better in the opening database, black winning a bit more than half the time. Needless to say, 1. g4 is not a great opening move by any standard.

The opening has been recently debated at the TalkChess.com forums using deep engine analysis, the question being whether it is objectively still drawn for white.[2] The most recent studies claim that Stockfish with depth 60 analysis would not be able hold the draw against a correspondence-level preparation by black, favoring the argument that 1. g4 is losing against best play.

In the game, black responds with the most principled move, attacking the g4-pawn immediately.

1... d5 2. Bg2 Bxg4 3. c4

The Grob gambit: Fritz gambit line. Leela plays as Grob is arguably supposed to be played, creating some early chaos on the board trying to trick the opponent. But this line is not really a true gambit line, as it will be hard for black to keep the extra pawn. This said, exchanging the g-pawn to a central or queenside pawn is not exactly a great idea, should black be not fooled by the Grob tricks.

3... c6 4. Qb3

White plays the Grob opening in an aggressive manner, attacking the b7-pawn. This position has been seen once in a rapid game at the super-GM level.[3] In that game, black played a bit more conservatively protecting the pawn with the queen but also relinquishing a bit of the advantage by not keeping the center: 4... Qd7 5. cxd5 Nf6 6. Nc3 cxd5 7. Nxd5 Nc6 and the damage to white's position was less than in this game.

In the game, black played a sharper move, maintaining the center while giving up the b7-pawn. The resulting semi-Slav central pawn structure severely restricts white's light-square bishop, questioning the merits of white's opening moves.

4... e6 5. Qxb7 Nd7 6. Nc3

The c6 pawn was poisoned and taking it would have been dangerous, as black would get some dangerous initiative by better development: 6. Qxc6? Rc8 7. Qa6 Rxc4.

6... Ne7 7. cxd5 exd5 8. d4 Rb8 9. Qa6 (diagram)

The theme of the poisoned pawn repeats: Taking a7 would come with a hefty price, as black would get the d4-pawn in exchange. For example: 9. Qxa7? Nf5 10. Nf3 Ra8 11. Qb7 Bxf3 12. exf3 Nxd4. Here white is probably objectively losing against black's initiative.


    
    
    
    
    
    
    
    
     
    
     
    
       
     
     
    
  ♛  ♜♚ 
♟   ♞♟♟ 
  ♜♝   ♟
  ♙♟♞  ♙
♘       
 ♙ ♕ ♘  
♙   ♙♙  
♖ ♗ ♔  ♖
Position after 20... Ne5


9... Rb6 10. Qd3 Ng6 11. h3

As is typical for Leela, white fights for some counterplay.

11... Be6 12. Nf3 Bd6 13. h4 h6(N)[4]

The last remaining game in the opening database was a correspondence game from 2014.[5] The game continued with 13... O-O 14. b3 Qf6 15. Kf1 h6 16. h5 Bf5 17. Qd1 Nf4 with white being in time to mount defenses. In the game, white was able to exchange enough pieces to hold against black's pressure.

Here black plays in a bit more direct manner.

14. Bh3

White forces the exchange of the bad bishop and tries to mount some counterplay on the open kingside files. But this plan is a bit double-edged, as it also exchanges a defender, and consequently, the white king will have difficulties in finding safe squares.

14... Bxh3 15. Rxh3 O-O 16. h5 Ne7 17. Rh1 c5

It is logical for black to start opening the position to take advantage of better development and white's king safety issues. 17... c5 also creates the immediate threat of c4, which white promptly discourages.

18. b3 Qc8 19. Na4 Rc6 20. dxc5 Ne5 (diagram)

An important intermezzo that allows black to develop the remaining bishop to a better square with tempo. This tempo is used to maneuver the rook on c6 to e6 with a plan of doubling the rooks on the e-file.


    
    
    
    
    
    
    
    
       
     
     
   
      
      
   
      
      ♚ 
♟    ♟♟ 
  ♞  ♜ ♟
 ♘♙♟♜  ♙
   ♝  ♛ 
 ♙ ♕    
♙  ♗♙♙ ♖
  ♖  ♔  
Position after 28... Qg4


21. Nxe5 Bxe5 22. Rb1 Re6 23. Bd2 Re8 24. Rc1 Nc6

Now that the rooks are doubled on the semi-open e-file, white's king safety is being seriously questioned. Black has multiple additional resources to add in the attack, as the light pieces are nicely centralized. The pawn on d5 is currently untouchable. It controls some important central squares and is ready to advance. The white king sidesteps the rook battery.

Note that taking the d5-pawn would be a big mistake. For example: 25. Qxd5? Rd8 26. Qg2 Qd7 27. Rd1 Bf4 28. e3 Qd3 29. Nb2 Qc2 30. Nc4 Nd4. White's position would collapse soon.

25. Kf1 Bd4 26. Nc3 Rf6 27. Rh2 Re5 28. Nb5 Qg4 (diagram) 29. Nxd4

White tries to relieve some of the pressure, but the white defenses are already getting overloaded.


    
    
    
    
    
    
    
    
       
     
      
     
       
       
    
        
      ♚ 
♟    ♟♟ 
  ♜    ♟
  ♙♟   ♜
 ♙      
  ♖     
♙  ♗♔♙  
        
Position after 36. Ke2


29... Nxd4 30. Rg2 Qh4 31. Qg3

White enforces the queens off the board with the immediate 32. Qxg7# threat. However, black wins material in the tactics afterwards.

The threat prompting the queen exchange was 31... Rxf2+ 32. Rxf2 Qh1#

31... Qxg3 32. Rxg3 Nxe2

Black wins a pawn and forks the rooks.

33. Rgc3 Nxc3 34. Rxc3 Rc6 35. b4 Rxh5 36. Ke2 (diagram) f6

Black makes room for Kf7-e6 and supports g5 with the h-pawn push. This also goes for a sharp pawn race.

Here black had multiple ways to convert the game. Perhaps the simplest would have been to put a halt to all white queenside counterplay and only then begin pushing with the kingside pawns: 36... d4 37. Rc4 Rh3 38. Rxd4 Ra3 39. Rd8+ Kh7 40. Be3 a6. As now the white queenside pawns are blockaded for the time being, black would have plenty of time to start pushing the kingside pawns.


    
    
    
    
    
    
    
    
      
      
      
       
      
       
      
        
    ♜ ♚ 
 ♙    ♟ 
  ♙  ♟  
   ♛    
   ♗ ♙  
  ♖     
♙♔      
        
Position after 48... Qd5


37. Kd3 Rc8 38. Be3 Rh4 39. f4

Natural move to prevent the black rook attacking b4. A similarly natural response would be to either play g5, intending to create connected passers, or Rh1 to start attacking the white pawns from behind and to make room for the h-pawn push. However, black goes for a faster conversion with the rush of getting the h-pawn promoted.

39... h5 40. c6 Rh1 41. b5 h4 42. Bxa7 h3 43. b6 h2 44. b7 Re8 45. Kc2

Even if the pawns on c6 and b7 look menacing or even winning, white is simply not in time to decide the game with the pawns queening. An example line: 45. c7 Rhe1 46. b8=Q h1=Q 47. Kd4 Qe4+ 48. Kc5 Rb1 49. a3 Qe1 50. Rb3 Qe7+ 51. Kc6 Qe6+ 52. Kc5 Rc1+ 53. Kb5 Qd7+ 54. Kb6 Rc6+ 55. Kb7 Rxb8+ 56. Bxb8 d4 57. f5 Rc1. White cannot both stop the d-pawn advancing and Qc6+ mating.

45... Rc1+

A cute little tactic to exchange the rook for a queen.

46. Kxc1 h1=Q+ 47. Kb2 d4 48. Bxd4 Qd5 (diagram)

Black gave up the d-pawn to dynamically blockade the b/c pawns. Any attempt to push either of them would lose them. The rest is now technique, as black uses the better mobility of the heavy pieces to overrun white with material-winning threats.


    
    
    
    
    
    
    
    
        
      
    
       
      
       
        
        
        
 ♙    ♟ 
 ♔♙ ♚♟  
♛       
♜    ♙  
    ♗   
        
        
Position after 62... Qa5#


49. Be3 Kf7

To exemplify the blockade, 50. c7 here would be simply met with 50... Qxb7+.

50. a4 Rh8 51. Rc2 Qe4

The queen forks the bishop and the a4-pawn. Naturally the bishop is move from the harm's way.

52. Bd2 Qxa4

53. c7 here would be met with 53... Qb5+

53. Rc3 Qd4 54. Be3 Rh2+ 55. Kb3 Qd5+ 56. Rc4 Qd3+ 57. Kb4 Rb2+ 58. Kc5 Ke6

The simple threat is now 59... Qd6#. This prompts white to give up the rook.

59. Rb4 Qd5+ 60. Kb6 Rxb4+ 61. Ka7 Ra4+ 62. Kb6 Qa5# 0-1 (diagram)

Black wins by mating.

References

  1. TCEC Season 18 This is not a SuFi Bonus Game 40: Stockfish-LeelaCPU
  2. 1.g4 opening is losing? http://talkchess.com/forum3/viewtopic.php?f=2&t=72084
  3. Chess.com PRO League Group Stage, Round 10: Nakamura-Mamedyarov ½-½, March 2019 (rapid 15m+2spm). https://www.chessgames.com/perl/chessgame?gid=1963060
  4. Lichess masters database. https://lichess.org/NlyX5cNC#25
  5. Ohtake-Parsons ½-½, 2014. https://lichess.org/9qWOVKzC