Learning from an epic endgame defense

by Albert Silver
5/3/2017 – When faced with a very bad or lost endgame, it is sometimes hard to muster the strength, much less the resources, to truly revert the situation. Yet that is what Hikaru Nakamura did, and against no less an expert than Vladimir Kramnik himself. Somehow the American found ways to keep it alive and justify playing on. Some of the resources he found smacked of black magic even. Here is a look at this epic endgame with detailed analysis by endgame expert GM Karsten Mueller.

Photos by Frederic Friedel



The position above is nothing short of a nightmare for Black, whose turn it is now. First of all, the pawn on d4 is untouchable. Can you see why? Then look at the bishop on a8. It is a sitting duck with nowhere to go! It is hard to imagine a more horrid position, yet Nakamura seems to find much of his best and most resourceful self precisely when things are the most desperate.



Ok, Black has not resigned yet, needless to say, yet it looks more like a question of refusing to accept the inevitable, right? Right?? Try to guess Black's next move.



It is hard to believe, but Black's wriggling and squirming has left White so out of sorts that he has now actually committed a fatal blunder. Black now has a concrete, irrevocable win. It is not a simple win, and even the top engines only truly see the light after hours of computing. Take your time and set up the position for leisurely analysis, or go straight away to Karsten Mueller's complete analysis of this and the rest of the game.

Remember:You can move the pieces on the diagrams above.

The players go over the long battle, which brought praise and admiration from players around the world

Vladimir Kramnik - Hikaru Nakamura (annotated by Karsten Mueller)

[Event "Kortchnoi ZCC 2017"] [Site "Zurich"] [Date "2017.04.15"] [Round "4"] [White "Kramnik, Vladimir"] [Black "Nakamura, Hikaru"] [Result "1/2-1/2"] [ECO "D05"] [WhiteElo "2811"] [BlackElo "2793"] [Annotator "Karsten Müller"] [PlyCount "127"] [EventDate "2017.??.??"] [SourceDate "2016.11.12"] [SourceVersionDate "2016.11.12"] 1. d4 Nf6 2. Nf3 d5 3. e3 e6 4. Nbd2 c5 5. b3 cxd4 6. exd4 Nc6 7. Bb2 g6 8. Bb5 Bg7 9. O-O O-O 10. Re1 Qb6 11. a4 Bd7 12. c4 a6 13. Bxc6 Bxc6 14. c5 Qc7 15. b4 Ne4 16. Ne5 Nxd2 17. Qxd2 b5 18. cxb6 Qxb6 19. Rec1 Rfc8 20. Rc5 Be8 21. Nd3 Rcb8 22. Bc3 Qd8 23. Ra5 Bf8 24. Nc5 Rb6 25. g3 Bd6 26. h4 Bc7 27. b5 axb5 28. Rxa8 Qxa8 29. axb5 Qc8 30. Ba5 Rxb5 31. Bxc7 Qxc7 32. Qg5 Rb8 33. h5 {The bad bishop strikes back! It looks like White is just winning due to the strong pressure on the dark squares and the bad bishop, but Nakamura's defence shows that bad bishops can be better than they look:} Qd8 $1 34. Qe5 {Of course Kramnik does not want to exchange his attacking queen.} gxh5 $1 {Nakamura correctly stops h5-h6, when the coffin nail would be too dangerous.} ({For example} 34... Rb6 $2 {loses due to} 35. h6 f6 36. Qe1 g5 37. Ra7 Qb8 38. Rg7+ Kh8 39. Kg2 Bg6 40. Qa5 $18) 35. Ra7 Rb6 $1 {Nakamura moves his rook out of harm's way.} (35... Qb6 $2 {runs into} 36. Nxe6 $18) 36. Qxh5 Kg7 37. Qg4+ Kf8 38. Qf4 Kg7 39. Qg4+ Kf8 40. Kg2 h6 41. Qf3 (41. Qe2 $5 {was another dangerous try, but Black should be able to hold after} Qd6 $14) 41... Kg7 42. Qa3 Qf6 { Nakamura activates the queen directly.} ({The computer solution} 42... e5 43. Ra8 Qe7 44. Qa5 Rd6 45. Rc8 e4 46. Qa8 Qf6 $11 {draws as well.}) 43. Ra8 Bb5 $2 {After the very good defensive effort Nakamura now puts his bad bishop in the way of his rook.} (43... Bc6 $1 {should defend, e.g.} 44. Rc8 Kh7 45. Qa7 Rb2 46. Nd3 Rd2 47. Nf4 Qf5 48. Rxc6 Qe4+ 49. Kh3 Qf5+ 50. Kh2 Rxf2+ 51. Kg1 Rb2 52. Qc7 Kg8 53. Rc1 Rb3 54. Ng2 Qe4 55. Qf4 Qxf4 56. Nxf4 Rxg3+ 57. Kf2 Ra3 $14 ) 44. Qa5 $2 {Kramnik had probably missed one of the black bishops retreats: either 44...Bc6 or more likely 46...Bd7! in the line after 45.Rc8.} (44. Qa7 { is the right way to start the attack, e.g.} Bc6 (44... Rc6 $2 45. Qb8 $18) ( 44... Bf1+ $2 45. Kxf1 Rb1+ 46. Kg2 Qf5 47. Qe7 $18) 45. Rc8 Rb2 46. Nd3 Rd2 47. Qa3 Rxd3 (47... Qxd4 $2 48. Qf8+ Kg6 49. Qg8+ Qg7 50. Nf4+ Kf6 51. Nh5+ $18 ) 48. Qxd3 Bd7 49. Rc7 Ba4 50. Rb7 $16) 44... Bc6 45. Qxb6 {Kramnik does not want to exchange rooks, but} (45. Rc8 {can be met by} Rb2 46. Nd3 Bd7 $1 47. Rc7 Rb3 48. Nf4 Qf5 {with equality as} 49. Rxd7 Qe4+ 50. Kh2 Rb2 51. Nh5+ Kg6 52. Nf4+ Kg7 {is a forced draw.}) 45... Bxa8 46. Qc7 h5 $1 ({Of course Black does not fall for the cheapo} 46... Qxd4 $2 47. Nxe6+) 47. f3 Qf5 48. Nd7 Bb7 $5 {Nakamura activates his bad bishop again.} 49. Qc5 ({Taking the bishop with } 49. Qxb7 {leads to perpetual check after} Qc2+ 50. Kh3 Qf5+ 51. Kh4 Qxf3 52. Ne5 Qf6+ 53. Kxh5 Qf5+ 54. Kh4 Qf6+ 55. Kh3 Qf1+ $11) 49... Ba6 50. Kf2 Kh7 51. Ne5 $2 {The wrong knight road.} (51. Nf8+ $1 Kg7 52. Nd7 {transposes to the game and is called for.}) ({Of course not} 51. Qf8 $4 Qc2+ 52. Kg1 Qd1+ 53. Kh2 Qe2+ 54. Kg1 Qf1+ 55. Kh2 Qf2+ 56. Kh1 Qxf3+ $19) 51... Kg7 $2 {Nakamura misses the moment.} (51... Qb1 {wins as Albert Silver discovered after forgeting to turn his Komodo 10 engine off. After more than four hours, including access to 6-piece endgame tablebases, it began displaying a +4 evaluation, leaving no doubt as to how things stood.} 52. Nxf7 Kg7 $3 {This move and the role of Black's king in the following attack is really amazing. White has no good option now:} 53. Qc7 (53. Nd8 Qf1+ 54. Ke3 Qe1+ 55. Kf4 Qd2+ 56. Ke5 Qg5+ 57. Kxe6 Qxd8 58. Qa7+ Kg6 59. Qf7+ Kh6 60. Qf4+ Qg5 61. Qf8+ Qg7 62. Qf4+ Kh7 $19) (53. Ne5 $2 Qf1+ 54. Ke3 Qe2+ 55. Kf4 Qd2#) (53. Qa7 $2 Qf1+ 54. Ke3 Qe2+ 55. Kf4 Kf6 {Black's king closes the escape road himself.} 56. g4 Qd2+ 57. Kg3 h4+ 58. Kxh4 Qh2#) (53. Qe7 $2 Qf1+ 54. Ke3 Qc1+ 55. Kf2 Qd2+ 56. Kg1 Qxd4+ 57. Kg2 Qd2+ 58. Kh1 Qe1+ 59. Kh2 Qf2+ 60. Kh1 (60. Kh3 Bf1+ 61. Kh4 Qh2+ 62. Kg5 Qxg3+ 63. Kxh5 Qxf3+ 64. Kh4 Qf4+ 65. Kh5 Be2#) 60... Qxf3+ 61. Kg1 Qxf7 $19) (53. Nd6 {is met by} Kg6 $3 54. Qc3 h4 55. gxh4 Qf1+ 56. Kg3 Qg1+ 57. Kf4 Qh2+ 58. Ke3 Qxd6 $19) 53... Qf1+ 54. Ke3 Qd3+ 55. Kf2 Qxd4+ 56. Kg2 Kf6 57. Nd8 Qd2+ 58. Kh1 Qe1+ 59. Kh2 Qf2+ 60. Kh1 Qxf3+ 61. Kh2 Qf2+ 62. Kh1 Qf5 $19) 52. Nd7 $1 {Now Kramnik's counterplay is just in time to stave off disaster.} Bc4 53. Qf8+ Kh7 54. Ne5 Qc2+ 55. Kg1 ({Of course not} 55. Ke3 $4 Qe2+ 56. Kf4 Qd2#) 55... Qd1+ 56. Kh2 Qd2+ 57. Kh3 Bf1+ 58. Kh4 Qh2+ 59. Kg5 Qxg3+ 60. Ng4 $1 hxg4 61. Qxf7+ Kh8 62. Qf8+ Kh7 63. Qf7+ Kh8 64. Qf8+ 1/2-1/2

All of Karsten Mueller's books and DVDs on the endgame have not only been met with praise and enthusiasm by readers and experts alike, such as acknowledgements by the late Mark Dvoretsky, but have established him as one of the leading authorities and instructors on the subject. You will find many such DVDs in the ChessBase Shop, such as the one below covering this particular piece combination, or others such as his "Golden Guidelines of Endgame Play" summarizing the best and most important takeaway rules on this phase of the game.

Chess Endgames 4 - Strategical Endgames 
By Dr. Karsten Müller

In the fourth volume of the endgame series a lot of themes with more material are included like rook and minor piece vs rook (and minor piece), rook vs two minor pieces, double rook endings, queen and knight vs queen and bishop, the bishop pair and endgame principles.

Typical topics are: the so-called Fischer endgame rook and bishop vs rook and knight, which occurs quite frequently in practical play and is very often favourable for rook and bishop; with opposite coloured bishops the presence of one rook each greatly increases the winning chances of the attacker compared to the pure opposite coloured bishop endgame (see DVD 1); when are queen and knight stronger than queen and bishop; which drawing methods exist in the pawnless endgame rook and bishop vs rook, which is quite often won over the board. Those who have always found studying endgames with chess textbooks too dry and tedious will enjoy this DVD with its comfortable training system and benefit enormously – in brilliant endgames at the actual board.

Complete video running time: approx. 6 hours.
Delivery: download, post.
Price: €29.90; €25.13 without VAT (for customers outside the EU); $28.41

Born in the US, he grew up in Paris, France, where he completed his Baccalaureat, and after college moved to Rio de Janeiro, Brazil. He had a peak rating of 2240 FIDE, and was a key designer of Chess Assistant 6. In 2010 he joined the ChessBase family as an editor and writer at ChessBase News. He is also a passionate photographer with work appearing in numerous publications.
Feedback and mail to our news service Please use this account if you want to contribute to or comment on our news page service


Rules for reader comments


Not registered yet? Register

benedictralph benedictralph 5/4/2017 03:44
It's nice to see, for once, a computer being acknowledged for being able to do or see something in chess a human probably couldn't.
Offramp Offramp 5/4/2017 04:34
The win for Black after 51...Qb1 is amazing.
Bertman Bertman 5/4/2017 08:00
@benedictralph - I don't know that a human could not. There is no question a human could not crunch all the moves to conclusively declare the position winning, but it is not unthinkable to intuit the move and main line. It is what humans do anyhow, no matter how well trained in calculation.
TMMM TMMM 5/5/2017 01:45
I would not call that position a "nightmare" for black; he is up a pawn, and with White's exposed king, the bishop stuck in the complete opposite corner, and White's weak d-pawn, black has decent chances to hold the draw with a perpetual, should white go after the bishop. You could even argue white has to be more careful than black; if black manages to consolidate his bad bishop, he may even think about winning...

(What would more qualify as a nightmare for me would be being down a pawn, and with the other side having the initiative. Having to defend that precisely for hours to get at most a draw is psychologically much more difficult.)
Karbuncle Karbuncle 5/5/2017 03:18
51...Qb1 will be instantly preferred by the top engines. It's the 2nd move of 52...Kg7!! than can be missed due to selective search pruning. Silver's engine was simply allowed to run long enough to investigate previously pruned lines. That being said, it also depends on whether or not you analyze with one variation or multiple choices. Using just two PVs will force the engine to investigate Kg7 immediately.