TCEC Season 17 CPU League 1 Game 138 – Komodo MCTS-Ethereal

From TCEC wiki
Revision as of 01:23, 12 February 2020 by Skiminki (talk | contribs) (Game overview)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Game Season 17, CPU League 1, Game 138
Result 1-0
ECO C00
Links TCEC archive
Lichess



This game was the second encounter of Komodo MCTS and Ethereal in the league. The previous game with the same opening, colors reversed was TCEC Season 17 CPU League 1 Game 18 – Ethereal-Komodo MCTS.


Game overview

    
    
    
    
    
    
    
    
  
   
      
     
       
        

   
♜♞ ♛♚♝ ♜
 ♟ ♞ ♟♟♟
  ♟ ♟   
 ♟ ♟♙   
   ♙    
        
♙♙♙♘♕♙♙♙
♖ ♗ ♘♖♔ 
Position after 10. Ne1


1. d4 a6 2. e4 e6 3. Nf3 d5 4. Nbd2 Bd7 5. Bd3 Bb5 6. Bxb5+ axb5 7. O-O Nf6 8. Qe2 c6

End of opening book moves. No games in the Lichess masters database.

9. e5 Nfd7 10. Ne1 (diagram)

The game is in a juncture with many different options for both sides. White has choices to make, such as whether to focus on the queenside or the kingside. In this game, white went for the queenside pressure, whereas in the reverse, kingside attack was the main plan. But the next move is on black:

  • 10... b4. Prepares for the c5 push by moving the b-pawn from the harms way. Then, for example, 11. Nd3 Be7 12. Qg4 g6. The last move weakens black's kingside dark squares, but also stops the immediate threats and discourages a future f5 by white. Black should now have time for slower b6 and c5 to obtain good pawn structure on the queenside. 12... Kf8 was also an interesting option here, giving up the castling rights but maintaining no weaknesses on the king side.
  • 10... Be7 11. Nd3 b4 12. Nb3 Na6 13. f4 O-O was the line played in the reverse game. It should be noted that many variations starting with 10... b4 and 10... Be7 are transposing by move ordering.
  • 10... Na6 11. Qg4 c5 12. c3 b4 attacking white's center
  • 10... c5 11. dxc5 Bxc5 12. Qxb5 to accelerate piece development.

In the game, black went along with perhaps a slightly dubious choice, attacking the center at the cost of a healthy queenside pawn structure. (See also editor's notes below.)


    
    
    
    
    
    
    
    
    
    
      
       
      
       
   
   
♜♞  ♚  ♜
 ♟   ♟♟♟
 ♝  ♟   
    ♞   
♙ ♟     
      ♙ 
 ♙♙♘ ♙ ♙
♖ ♗ ♘♖ ♔
Position after 18. g3


10... Qc7 11. Nb3 c5 12. dxc5 Bxc5 13. Qxb5 Ba7 14. Kh1 Qc4

A creative way to block the white queen, essentially forcing a queen trade. If the white queen moves away from the a6-f1 diagonal, the rook on f1 is hanging.

15. a4 Bb6

The immediate problem for black is that the queenside is tangled. The quick solution would not have worked, as 15... Qxb5? 16. axb5 with a threat to pin the bishop on a7. The continuations are very favorable for white. For example: 16... Nb6 17. Be3 N8d7 18. Nd3 Ke7 19. Bd4 Rhc8 20. Nbc5 Nc4 21. Rxa7 Rxa7 22. Nxd7 Ra5 23. N7c5 Rxb5 leaving white with 2 pieces for a rook and strong pressure on the Q-side.

However, black does not have to exchange the queens. Instead, black can continue developing with natural moves such as Nc6 and O-O, to connect the rooks and to then allow Qxb5 with better terms. Because of this, white does not wait for black's development and exchanges the queens immediately.

16. Qxc4 dxc4 17. Nd2 Nxe5 18. g3 (diagram)

In the game, black pushed the c-pawn a move later after castling. However, if that was the plan, pushing the c-pawn right away was also an option: 18... c3 19. Ne4 h5 20. b3 O-O. The difference is that now black gets to blockade white's queenside pawn front with the advanced c3-pawn, at least for a while. Taking with 19. bxc3 was not really an option for white, as it would have ruined the queenside pawn structure.

However, here an arguably stronger move than the game continuation was 18... Bc5 unblocking the b-pawn. If white would now continue with the f-pawn push, then 19. f4 Ned7 20. Nxc4 b5 and black would be winning the a-pawn on the next move. This would have resulted in a more balanced position.


    
    
    
    
    
    
    
    
     
    
       
        
     
     
     
       
♜  ♜  ♚ 
 ♟   ♟♟♟
    ♟   
        
♙♙ ♝    
  ♙  ♖♙ 
♖  ♗   ♙
       ♔
Position after 26. Rxf3


18... O-O 19. f4 c3 20. fxe5

Note that 20. bxc3 Ned7 21. Nd3 Rc8 would leave the doubled c-pawns weak, making it less than ideal option for white.

20... cxd2 21. Bxd2 Nc6 22. Nf3 Rfd8 23. b4 Bd4 24. Ra2 Nxe5 25. c3 Nxf3 26. Rxf3 (diagram) Bf6

The previous move was undoubtedly an inaccuracy. Black needs the bishop to slow down white's pawn front, and better moves for that purpose were Ba7 and Be5. To illustrate: 26... Be5 27. a5 Rac8. Now, unlike in the game, 28. Rc2 can be met with 28... Rc4.

The idea behind 27... Ba7 would be somewhat different, supporting the e-pawn push for active counterplay.


    
    
    
    
    
    
    
    
      
     
      
    
       
     
     
        
♜   ♚   
 ♟♝  ♟  
    ♟ ♟ 
♙ ♙  ♜ ♟
 ♙      
   ♖  ♙♙
    ♖♗♔ 
        
Position after 45. c5


27. a5

Now that the is bishop on f6, white has an answer for the rook blockading idea: 27... Rac8 28. Rc2 Rc4. Now 29. Rf4 can be played, as the bishop is not controlling the f4-square. Thus, black did not go for it.

27... Kf8 28. Rc2 Ke8 29. Kg2 Ke7 30. Kf2 Rac8 31. c4 Rd5 32. Ke2

The rook was of course not hanging: 32. cxd5? Rxc2 33. Rd3 e5.

32... Re5+ 33. Be3 h6 34. Kf1 Re4 35. Bc5+ Ke8 36. Re3 Rg4

It is in black's best interest to keep the rooks on board, as more pieces offer generally better chances for counterplay. However, the rook on g4 is getting short of squares.

37. h3 Rg5 38. Rb3 Re5 39. Bf2 Ra8 40. Re2 Rf5 41. Rd3 h5 42. Rd6 Bd8 43. Kg2 Bc7 44. Rd3 g6 45. c5 (diagram) Re5

Now black offers the rook exchange. White's threat was to double on the d-file: 45... g5 46. Red2 g4 47. hxg4 hxg4 48. Rd7


    
    
    
    
    
    
    
    
      
       
       
   
       
      
       
      
♜   ♚   
 ♟      
      ♟ 
♝♙♙  ♟ ♟
    ♟   
      ♙♙
    ♔   
♖   ♗   
Position after 52. Ra1


46. Rxe5 Bxe5 47. Kf3 Bc7 48. Be1 e5 49. b5 f5 50. Ke2 e4 51. Rd1 Bxa5

Black didn't have anything better to stop the pawns, and now white gets to pin and win the bishop. For example: 51... Ke7 52. a6 bxa6 53. b6 Be5 54. b7 Rd8 55. Bc3 Bc7 56. Bf6+ Kxf6 57. Rxd8 Ke7 58. Rd1 Bxg3 with no hope for black for stopping the advanced b/c pawns.

52. Ra1 (diagram) Rc8 53. Rxa5 Rxc5 54. h4

Conversion here still needs care, as black has some counterplay ideas with the e/f pawns. If white was to play the tempting bishop move to prevent b6 driving the rook away, the natural plan would fail: 54. Bf2 Rc2+ 55. Ke1 h4 56. gxh4 f4 57. Ra7?? e3. White would have to give up the bishop to stop the pawns, and the game would likely be a draw.


    
    
    
    
    
    
    
    
        
        
      
     
      
      
       
       
        
        
 ♟  ♚   
     ♟♖♟
    ♟  ♙
    ♗ ♙ 
♜       
      ♔ 
Position after 66. Kg1


54... Kd7 55. Ke3

As keeping the b-pawn with advantage is not so easy, white goes for a more straightforward solution and offers the b-pawn for the rook's access to the kingside pawns.

55... b6 56. Ra7+ Ke6 57. Kf4 Rxb5 58. Rg7 g5+ 59. Rxg5 Rb1 60. Bd2 Rf1+ 61. Ke3 Ra1 62. Ke2 Ra3 63. Be3 Ra2+ 64. Kf1 Ra1+ 65. Kf2 Ra2+ 66. Kg1 (diagram) Ra3

Black gives up the b6 pawn without a fight, which is a sign of an engine fighting the losing game.

Regardless of the b5-pawn, black was losing anyways. For example: 66... b5 67. Rxh5 b4 68. Rh6+ Kd5 69. Rb6 Kc4 70. h5 Kd3 71. Bc5 e3 72. Rxb4 Ra1+ 73. Kg2 e2 74. Bf2 e1=Q 75. Bxe1 Rxe1 76. Rf4. The extra two connected passed pawns prove an easy win for white.


    
    
    
    
    
    
    
    
        
        
       
      
       
      
      
       
        
        
    ♚   
     ♖ ♙
      ♙ 
     ♟♔ 
    ♟♗  
 ♜      
Position after 77... e2


67. Bxb6 Rc3 68. Ba5 Rf3 69. Be1 Ra3 70. Kf1 Ra2 71. Rxh5 Rh2 72. Bf2 Rh1+ 73. Kg2 Rb1 74. g4 f4

Final try for black to make some active counterplay with the connected passers. But as the white king and bishop are in the vicinity of the pawns, white is well prepared to sacrifice the bishop for both black's pawns.

75. Rf5 f3+

The e-pawn push 75... e3 76. Rxf4 exf2 77. Kxf2 was completely winning for white as well.

76. Kg3 e3

Here white has 10 winning moves to choose from. The most natural move was one of them: 77. Bxe3 Rf1 78. Bf2 Rh1 79. Rxf3. White chose to delay the sacrifice and push the h-pawn instead.

77. h5 e2 1-0. (diagram)

The game was adjudicated by the TCEC win rule. A possible continuation: 77... e2 78. h6 Rb8 79. Rxf3 Rb1 80. Re3+ Kf6 81. Rxe2 Rb3+ 82. Kf4 Kg6 83. Re6+ Kf7 84. Rb6 Ra3 85. g5 Ra4+ 86. Kg3 Ra3+ 87. Kg2 Ra5 88. g6+ Ke7 89. h7 Rg5+ 90. Bg3 Rh5 91. g7 Rc5 92. g8=Q Rc2+ 93. Bf2 Rxf2+ 94. Kxf2 Kd7 95. Qf7+ Kd8 96. h8=Q#

Editor's notes

This is a game where it is somewhat difficult to pinpoint the exact moment when the game was decided.

Move 10 for black seems to be critical for the long term direction of the game. Analysis with SF11 to depth 43 and 48.7G nodes offered the following choices: 10... b4 0.00; 10... Be7 0.00; 10... Na6 +0.23; 10... Qc7 +0.25; 10... c5 +0.49. Leela T42810 after 10M nodes largely agree with: 10... b4 +0.34; 10... Be7 +0.36; 10... Qc7 +0.36; 10... Na6 +0.39; 10... c5 +0.46. Cato's book also knows the position with the following statistics: 10... Be7 (29%, 31 games), 10... Na6 (23%, 22 games), 10... c5 (38%, 21 games), 10... Nb6 (25%, 6 games) and 10... Qg5 (0%, 1 games). The percentage is the average score for black. So definitely, white has some advantage in this position.

Black had an opportunity to undermine the queenside 3-2 pawn majority with 18... Bc5. Instead, black allowed the 3-1 majority by castling, which was arguably a waste of an important tempo. Without any redeeming factors, holding such majority is generally hard. The deciding inaccuracy was perhaps 26... Bf6, although it is not quite certain whether the game was already objectively lost before this move, or whether the game was still salvageable with precise play after the move.