TCEC Season 17 CPU League 1 Game 18 – Ethereal-Komodo MCTS
|Game||Season 17, CPU League 1, Game 18|
|Links|| TCEC archive|
|Position after 18. Bh6|
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 Be7 11. Nd3 b4 12. Nb3 Na6 13. f4
The more direct 13. Qg4 here would have been interesting, as it would have held back black from castling due to the simple mating threat: 13. Qg4 O-O?? 14. Bh6 g6 15. Bxf8.
13... O-O 14. Be3 b6
Black prepares for the queenside play, but allows white to set up pieces for an attack on the kingside.
15. f5 exf5 16. Rxf5 c5
Here white has multiple different options available. Some examples:
- Defusing the 17... c4 threat of winning material with a line such as 17. Nf2 c4 18. Nd2 Nc7 19. Ng4. But this would also slowed down the attack a bit.
- Sacrificing for the h-file pressure with a line such as 17. Rh5 c4 18. Qg4 g6 19. Rh6 Kg7 (knights still forked by the c4-pawn)
- But taking the pawn would be have been dangerous for white: 17. dxc5 bxc5 18. Nd2 c4 19. Nf2 g6 20. Rf4 Nxe5.
In the game, white chose another sharp line, offering two pieces for a rook, a pawn, and an attack.
17. Qg4 c4 18. Bh6 (diagram) g6 19. Nf4 cxb3 20. axb3 Nc7
Black has no time to save the rook: 20... Re8 21. Rxf7 Kxf7?? 22. Qe6#. The most resilient defense would still have lost a critical amount of material: 21... Bf8 22. Rxd7 and black has no time to save the queen, as 22... Qb8?? 23. Nxg6 Bxh6 24. Nf8+ Kxf8 25. Rf1+ Bf4 26. Qg7#
|Position after 28... Ne8|
21. Rf1 Qc8
Sets up an exposure threat of Qxf5 to eventually defuse white's attack. But again, there was no time to save the rook, as 21... Re8 22. Rxf7 Kxf7 23. e6+ Kg8 24. Nxg6 and the eventual mate cannot be stopped after 20 more moves with best play.
Another tempting idea of stepping aside with the king simply does not work. This is because the f7 square is too weak: 21... Kh8 22. Bxf8 Nxf8 23. Rxf7 Qe8 24. e6. Black still has two minors for the rook, but white's attack is too strong for black to survive.
22. h3 Nc5
Now black is concretely threatening Qxc5. Note that the knight on c5 is not really hanging, since 23. dxc5?? Bxc5+ 24. Kh2 Qxf5 25. Qxf5 gxf5 26. Bxf8 Rxf8 would leave black piece up with negligible compensation for white.
23. Bxf8 Bxf8
Again, white had an interesting option: 24. Rxf7.
- 24... Qxg4 25. hxg4 Kxf7 26. Nxd5+ Kg7 27. Nxc7 Rc8 and white has three pawns for a piece, of which two are connected passers. Likely a draw.
- 24... Kxf7 25. Ne6+ Ke7 26. Qh4+ Kxe6 27. Rf6+ Kd7 28. Qxh7+ Kd8 29. Rxf8+ Ne8 30. Qxg6 Ra1+ 31. Kh2 Qe6 32. Rxe8+ Qxe8 33. Qxb6+ Kd7 34. Qd6+ Kc8 35. Qxc5+. This long line would be an interesting endgame with queen and up to 7 non-advanced pawns versus queen and rook. Both Leela and Stockfish consider this equal.
The above example lines are not forced.
In the game, white parries the Qxc5 threat creatively.
24. e6 N5xe6 25. Rxf7 Nxd4
Following the theme of many earlier lines, the rook on f7 was poisoned. 25... Kxf7 26. Nxe6+ Ke7 27. Nxc7 Qxg4 28. hxg4. White is pawn up and much better.
The rook move lights up the Nxg6 threat. If black does nothing, the rook on the 7th rank will decide the game. For example: 26... b5 27. Nxg6 with unstoppable mating threats. Black's best attempt at a defense loses the queen with 27... Bg7 28. Ne7+ Kh8 29. Nxc8. Mate is also incoming in 5 more moves.
Counters the threat with the exposition threat Nf5+ threatening to win the rook on d7. White king side-steps the threat.
27. Kh2 Nf5 28. Nh5
Protecting the rook on d7 dynamically via the f6+ fork. Black's next move enters a forced sequence leading to an interesting queen versus rook, bishop, and knight endgame.
28... Ne8 (diagram) 29. Rxf5 Qxd7 30. Rf8+ Kxf8 31. Qxd7 gxh5
End of the forced sequence. The game enters now in a very interesting phase, essentially the white queen trying to escort the two connected passed pawns while black attempts to get the pieces coordinated (see the diagram after 36. Qxh5). Generally, a coordinated rook/bishop/knight combination is a bit stronger than a lone queen, which means that white's game is essentially on a timer. Black will lose two pawns before initial stabilization.
In this position, the black bishop is importantly anchored on the c5 square, providing many protected dark squares for the black rook and knight.
It should be noted that a KRBNvKQ endgame without pawns is generally a draw if the position starts without unparryable threats by either side.
|Position after 36. Qxh5|
32. Qf5+ Ke7 33. Qxh7+ Ke6 34. Qg8+ Kd6 35. Qg6+ Kd7 36. Qxh5 (diagram) Nf6 37. Qf5+ Ke7 38. g4 Rf8 39. Kg2 Bd6 40. Qg5 Kd7 41. c4
The c-pawn push was necessary to limit black's queenside play and the black king being able to enter the white side. For example: 41. h4 Kc6 42. Qg6 Ne4 43. g5 Rf2+ 44. Kh3 Kc5 45. Qe8 Kd4. Black is preparing a mating net, which prevents white making progress: 46. g6 Nf6 47. c3+ Kd3 48. Qb5+ Kc2 and Rh2# is coming. Similarly, the h-pawn push on move 46 would end badly for white, although not as quickly.
However, the c-pawn push now has a drawback that the anchor square (c5) for the black bishop can no longer be challenged. White could have also tried 41. c3, which would have kept the possibility to challenge the anchor square alive.
|Position after 47... Ke6|
Taking the pawn with 41... bxc3 42. bxc3 Kc6 43. c4 would open up the queenside, giving white queen more opportunities to harass.
42. Qg6 Bc5 43. g5 Ng8 44. Qe4 Ne7 45. h4 Rf5 46. g6 d3
The sacrifice of the d-pawn is important, since it makes the f2-square protected for the rook, important for achieving coordination behind the white pawns.
47. Qxd3+ Ke6 (diagram) 48. g7
White could have won a tempo here by playing 48. Qg3 before pushing the g-pawn. The point being that the black king would have to do a detour via the f6 square.
|Position after 69. Qf6+|
48... Kf7 49. Qg3 Kg8 50. Qg4 Rf6 51. Kh3 Rg6
Because white wasted the tempo, black now wins the g-pawn easily.
52. Qf4 Kxg7 53. h5 Rf6 54. Qc1
A bit weird computer move, wasting another tempo. This move is arguably a small imprecision. White could have played Qg5+ immediately, if that was the plan.
54... Nf5 55. Qg5+ Kf7 56. Qg4 Nd4
More precise was 56... Ne7 57. Qe4 Rf5 58. Qg4, which would have been close to winning for black, as the black pieces are getting towards white's h-pawn.
57. Qd7+ Kf8 58. Qd8+ Kg7 59. Kg4 Nc6 60. Qd7+ Ne7 61. Qc7 Rf5 62. Qb8
This was black's last chance to win the h5-pawn and the game. The key is to cut the white king from the vicinity of the h-pawn, and provide enough protection for the black king to advance. 62... Be3 63. Qe8 Rf4+ 64. Kh3 Bc5 65. Qd7 Kf6 66. Qd2 Rf2 67. Qh6+ Kf5 68. Kg3 Ke5 69. Qg7+ Rf6 70. Kg4 Nf5 71. Qd7 Nh6+ 72. Kg3 Rd6 73. Qe8+ Kf5 74. Qe2 Kg5. Black would now win the h5-pawn, after which black would slowly and unstoppably start advancing to the queenside.
62... Rf7 63. Qe5+
But now, white is just in time with the h-pawn and is able to create enough counterplay to hold the draw.
63... Kf8 64. h6 Rh7 65. Kh5 Kg8 66. Qg3+ Kf7 67. Qh3 Kg8 68. Qe6+ Kf8 69. Qf6+ ½-½. (diagram)
The game was adjudicated by the TCEC draw rule. The white pawn is now too advanced for black to overcome white's initiative with the less mobile pieces.