TCEC Season 17 Division P Game 6: Stockfish-Komodo

From TCEC wiki
Revision as of 11:23, 23 March 2020 by Skiminki (talk | contribs) (Game overview)
Jump to: navigation, search
Game Season 17, Division Premier, Game 6
Result 1-0
ECO B20
Links TCEC archive
Lichess



Game overview

    
    
    
    
    
    
    
    
   
    
    
     
       
    
    
   
♜ ♝♛ ♜♚ 
♟♟    ♝♟
  ♞ ♟♟♟ 
   ♟♙♞  
   ♙    
 ♙♘  ♘♙ 
♙    ♙♗♙
♖ ♗♕♖ ♔ 
Position after 13. b3


1. e4 c5 2. d3 e6 3. Nf3 Nc6 4. g3 g6 5. Bg2 Bg7 6. O-O d6 7. Re1 Nge7 8. c3 O-O

End of book. Lichess Masters database has 62 games with 41% for white, 34% drawn, and 25% for black.[1]

9. d4 cxd4 10. cxd4 d5 11. e5 f6

The move 11... f6 may look a bit weird at first, but black needs to do something and challenging the center is a plan. It is also not that easy for white to use the a2-g8 diagonal to take advantage of the exposed f7 square. This also indirectly adds pressure on white's pawn on d4.

Pushing the f6-pawn was not forced. Alternatively, black could have tried to take some initiative on the king side with something such as 11... h6 12. Nc3 Nf5 13. Bh3 Bd7 14. Bxf5 exf5 15. Nxd5 g5. But that would have allowed white to maintain some central control.

12. Nc3

Taking the f6-pawn would nullify white's minor advantage. Superficially, it looks like that black might have trouble with the backward e6-pawn, but tactics will solve this issue. For example: 12. exf6 Bxf6 13. Bh6 Rf7 14. Ng5 Bxg5 15. Bxg5 Qb6 16. Nd2 Nxd4 17. Bxe7 Rxe7 18. Bxd5 Rf7 19. Bg2 Qxb2. In this line, the position remains still somewhat sharp. Black is a pawn up and aiming at the f2 square, but also has some problems with the long a8-h1 diagonal and piece development. Computer analysis suggests this is close to equal.

12... Nf5 13. b3(N) (diagram)

Leela (T42810) suggests a different plan for white, instead of 13. b3. Leela would suggest more direct approach to claim some space both on the kingside and the queenside, starting with 13. g4. And indeed, g4 is also the top choice in the Lichess Masters database, although only 7 games before 13. b3. A possible continuation: 13. g4 Nfe7 14. exf6 Rxf6 15. Rb1 Rf7 16. b4 a6 17. a4 b5 leading the game in a quite different direction.


    
    
    
    
    
    
    
    
      
   
      
    
      
   
      
      
  ♜   ♚ 
 ♟ ♝♛♜ ♟
    ♟  ♝
♟  ♟♙♟  
   ♘   ♙
♙♙ ♕  ♙♔
 ♗   ♙  
 ♖  ♖   
Position after 30. Kh3


13... Bd7 14. Ba3 Re8 15. Bb2 Rc8 16. Rb1 Rf8 17. h4 Qe7 18. Ne2 a5 19. a3 Bh6 20. Nh2 Bg7 21. Nf3 Bh6 22. Qd3 fxe5 23. dxe5 Qc5 24. Bh3 Ng7 25. Bg4 Rf7 26. Ned4 Nxd4 27. Nxd4 Nf5 28. Bxf5 gxf5 29. Kh2 Qe7 30. Kh3 (diagram) f4

This is where the game becomes interesting. (This is in editor's opinion, but see also the video analysis below for the previous moves.) 30... f4 is a somewhat committal move, especially as white is not strongly encouraged to take the f-pawn. And thus, it may become a long-term weakness. However, white was likely going to play Bc1 to challenge the bishop on h6 if black deferred f4, and then f4 was practically forced anyways.

Consider the following line trying to avoid the f-pawn push: 30... a4 31. bxa4 Rc4 32. Bc1. Now the f-pawn must be pushed. If insisting not to push with, e.g., 32... Bxc1?, then 33. Rexc1 Qc5 34. Rxc4 dxc4 35. Qe3 and white is a pawn up with some dangerous kingside potential.


    
    
    
    
    
    
    
    
      
    
       
     
   
      
     
      
    ♛  ♚
 ♟   ♜♝♟
    ♟   
   ♟♙  ♙
♝ ♜♘ ♟♙ 
♙      ♔
 ♗ ♕ ♙  
  ♖   ♖ 
Position after 36... Rc4


31. Qd2

White prevents black from playing fxg3 or f3 as the bishop on h6 would be hanging. Here 31. Bc1 could also have been played for a similar effect.

Taking the f4-pawn would be a huge mistake: 31. gxf4?? Rxf4 32. Rg1+ Kh8 33. Qg3 Qf7 and white is practically lost. White's defenses would get quickly overloaded against the threats on the pawns on f2 and h4, and the knight on d4 (by Bf8-c5) to begin with.

31... a4 32. bxa4 Bxa4 33. Rg1

White threatens to win a tempo with gxf4+ to prevent black from taking black immediately.

33... Kh8 34. g4

White pushes the g4-pawn with the claim that f4 is becoming a weakness.

34... Bg7 35. h5 Qe8 36. Rbc1 (diagram)

Black is now starting to feel the pressure, as white threatens to seize the control of the c-file, and with that, the queenside. Black's options are not great:

  • Attempting to stabilize: 36... Bd7 37. Rxc8 Qxc8 38. Bc3 Bxe5 39. Rc1 Kg8 40. Nf3 Bxc3 41. Rxc3 Qd8 42. Ne5 Rf8 43. Rf3. But white is getting a strong attacking initiative with the practically unchallengeable knight on e5 and connected g4/h5 pawns. Black's f4-pawn is also subject to be lost soon.
  • Rook exchange to try to relieve some pressure: 36... Rxc1 37. Rxc1 Kg8 38. Qb4 Bd7 39. Qd6 Bc8 40. Bc3 b6 41. Bb4 Rb7 (protecting against the Qb8 threat) 42. Rc7 Rxc7 43. Qxc7. Here white has a crushing initiative with the b6-pawn likely getting won by white, and the a-pawn running.

Given the alternatives, black opted for arguably the best option, temporarily giving up a pawn to try to ease the pressure. However, strategically black is already in a difficult spot given the pawn structure.


    
    
    
    
    
    
    
    
     
     
       
      
      
        
      
       
  ♕  ♝ ♚
 ♟  ♛  ♟
    ♟   
    ♙  ♙
     ♟♙ 
        
     ♙♔ 
♗       
Position after 49. Qc8


36... Rc4 37. Rxc4 dxc4 38. Qb4 Bc6 39. Qxc4 Bd5 40. Qd3 Rd7 41. Qe2 Rd8 42. Nb5 Bf3 43. Qxf3 Qxb5 44. Rd1 Rxd1 45. Qxd1 Bf8

White's bishop was not hanging, of course, as 45... Qxb2?? 46. Qd8+ Bf8 47. Qxf8# would have been an embarrassing mate.

46. Ba1 Qc5

For similar backrank issues, the a3-pawn was still untouchable: 46... Bxa3?? 47. Qd8+ Kg7 48. Qf6+ Kg8 49. Qxe6+ is winning for white with the advanced e-pawn and the open black king. However, the queen now on c5 will win the a-pawn for the threat of Qxf2.

47. Qd8 Qxa3+ 48. Kg2 Qe7 49. Qc8 (diagram)

White avoids the queen trade and maintains the pin. The plan for white here is simple in principle, but somewhat complex to execute, as black's counterplay must be prevented. The goal for white is to win one of the weak pawns b7, e6, or f4, and with that, the game. The black king is boxed in, so white effectively has an extra piece, king, available for capturing the pawns. White's pawn structure is also significantly healthier, and white's only weakness, the e5-pawn, is not an easy target for black.

White's concrete winning plan is to maneuver the bishop to f6 to restrict black's pieces, fix the kingside pawn structure with f3 to stop immediate counterplay on g4, and then begin a king walk to take the b-pawn or possibly the f4-pawn. White is objectively winning. (SF11+: 49... Kg7 +4.81 depth 63, 107B nodes.)


    
    
    
    
    
    
    
    
      
       
     
      
    
       
        
       
     ♝♚ 
   ♛    
    ♟♗ ♟
    ♙  ♙
 ♟♕  ♟♙ 
     ♙  
        
    ♔   
Position after 73. Ke1


49... Kg8 50. Bd4 Kf7 51. Bb6

White is concretely threatening Bd8 to chase the queen away from protecting the b-pawn, and then capture it with a check.

51... Qg5

As white's threat is to win either of the b7/e6-pawns, black goes for active counterplay.

52. Qc7+

White protects the e5 pawn. Taking the b-pawn here would have been premature: 52. Qxb7+?? Kg8. White will lose the e5 pawn due to double e5/g4 attack by black, resulting in a likely draw.

52... Kg8 53. f3

White solidifies the pawn structure and protects g4.

53... Qh4 54. Bf2

Preventing counterplay is imperative, so the b7-pawn must wait. Taking the pawn now would have given up most of the advantage: 54. Qxb7 Qe1 and white cannot protect the e5-pawn. 55. Bc7?? would lose immediately to Qe2+ mating and 55. Qe4 would end in a forced draw.

54... Qe7 55. Qc8 b5

Staying passive would not work for black, as white would be able to enforce the bishop exchange and then win the f4 pawn by force, analogous to the game continuation.

56. Kh3 h6 57. Bh4 Qf7 58. Bf6

Bishop is installed in its ideal square for now.

58... b4 59. Kg2 Kh7 60. Qc4 Kg8

Now white plays some shuffling moves before initiating the king walk. Shuffling is likely due to that white dictates the direction of the game and black is simply reacting. Thus, transpositions via shuffling do not change the overall evaluation. White could have played 61. Kf1 already here.

61. Qc8 Kh7 62. Qc2+ Kg8 63. Kh2 Qe8 64. Qd3 Qf7 65. Kh3 Qh7 66. Qc4 Qd7 67. Qc2 Qf7 68. Qc8 Kh7 69. Kg2 Kg8 70. Kf1

Now the king walk finally begins.

70... Kh7 71. Qc4

White still needs to be careful to keep the winning advantage. Continuing the king walk carelessly would allow black to draw: 71. Ke2?? Bg7 72. Bxg7 Kxg7. White can no longer make progress.

71... Kg8 72. Ke2 Qd7 73. Ke1 (diagram)

White spends a tempo to force black to play a losing move due to a zugzwang. In the game, black vacates the d-file with the queen, which is the most resilient option.

Black could have kept the queen on the d-file, preventing white king's access to it. But then white would win by other means. For example:

  • 73... Kh7 74. Qc2+ Kg8 75. Qg6+ Bg7 76. Bxg7 Qxg7 77. Qxg7+ Kxg7 78. Kd2. The resulting king-pawn endgame is easily winning for white.
  • 73... Be7 74. Qc2 Bxf6 75. exf6 Kf8 76. Qg6 Qa7 77. Qxh6+ Kg8 78. Qxf4. Hopeless for black with two pawns down and white king close enough to intercept the b-pawn even if black was able to exchange the queens.
  • 73... Bg7 74. Bxg7 Kxg7 75. Qxb4 Qa7 76. Qxf4. Black has some checks, but eventually white is able to escape them. Two pawns up is a win for white.

73... Qf7 74. Kd2 Qe8 75. Kc1 Qd7 76. Qc2

With the threat of Qg6+ to tie up the black queen for defending duties.


    
    
    
    
    
    
    
    
       
       
       
      
     
      
        
        
       ♚
 ♛      
       ♟
    ♙  ♙
 ♔   ♟♙ 
   ♕ ♙  
        
        
Position after 88... Qb7+


76... Qf7 77. Kb2 Bg7

Black finally had enough and started to crumble.

White had a beautiful forcing win if black was to resist: 77... Qe8 78. Kb3 Qf7 79. Qc8 Kh7 80. Bd8 Kg8 81. Bc7 Kh7 82. Bd6 Bg7 83. Qc2+ Kg8 84. Qc7 enforcing the queen trade 84... Qe8 85. Qb8 Kf7 86. Qxe8+ Kxe8 87. Kxb4, and a tablebase win follows. Black could not have exchanged the bishops, as 82... Bxd6 83. exd6 e5+ 84. Kxb4 and the passed d-pawn will win the game quickly for white.

78. Bxg7 Kxg7 79. Kb3 Qe8 80. Qe4

White protects the pawn on e5 while maintaining the Qg6+ threat.

Again, overzealous pawn capture would have lost the winning advantage: 80. Kxb4?? Qb8+. If white wants to avoid perpetuals, then the e5-pawn will be lost: 81. Kc4 Qc7+ 82. Kd3 Qxe5. The point here is that there is no Qg6+, since the white king blocks the queen.

80... Qf7 81. Qb1

Now a bit more shuffling. Immediate Kxb4 was also good.

81... Qe8 82. Qd3 Qf7 83. Qe4 Kh8 84. Qg6 Qf8

Alternative was 84... Qxg6 85. hxg6 Kg7 86. Kxb4 Kxg6 87. Kc5, but white wins the e6 pawn by force, as black is not in time for counterplay. 87... h5 88. gxh5+ Kxh5 89. Kd6 and white is faster to queen. 89... Kh4 90. Kxe6 Kg3 91. Kf5 Kxf3 92. e6 and so on.

85. Qxe6 Qf7

Final trick offering the queen, but white will not fall for such an easy stalemate trap.

86. Qc4 Qb7 87. Qd3

Black can no longer protect against all threats. For example: Kxb4, e-pawn running, and Qg6.

87... Qa8 88. Kxb4 Qb7+ 1-0. (diagram)

White wins by the TCEC win rule with mate in 29 more moves. For example: 89. Qb5 Qxf3 90. Qe8+ Kg7 91. Qg6+ Kf8 92. Qxh6+ Ke8 93. Qg6+ Kd7 94. h6 Qb7+ 95. Kc4 Qc7+ 96. Kd3 Qc1 97. Qf7+ Kd8 98. Qd5+ Ke7 99. h7 Qd1+ 100. Kc4 Qa4+ 101. Kc3 Qa3+ 102. Kd2 Qe3+ 103. Kc2 Qe2+ 104. Kb3 Qe3+ 105. Ka4 Qa7+ 106. Kb5 Qb8+ 107. Kc4 { Now black finally runs out of usable checks.} 107... Qh8 108. g5 Qc8+ 109. Qc5+ Qxc5+ 110. Kxc5 Ke6 111. h8=Q Kf5 112. e6 f3 113. e7 f2 114. Qh3+ Kxg5 115. e8=Q f1=Q 116. Qg3+ Kf5 117. Qee5#.

Editor's notes

This game is a notable example on the endgame advantage Stockfish currently enjoys over most other contemporary AB engines. Stockfish crossed the +3 mark on move 46. Ba1 whereas Komodo did that only 30 moves later on 76... Qf7. Arguably, the advantage is a bit exaggerated here due to shuffling, but it is still remarkable.

Analysis

Chesspuzzler's analysis of the game


References

  1. https://lichess.org/TX6gyXju#17, accessed on 2020-03-22