A surprising piece sacrifice in Garry Kasparov’s pet Najdorf

The Najdorf Sicilian is a rare guest on our site. But with German prodigy Vincent Keymer having made it his main defense against 1.e4 as a German chess site we should open our playbook on “the queen of openings”.

This correspondence game is somewhat different from most others already published, not only regarding the opening. Usually we pick encounters that highlight how humans still understand some aspects of the game much better than the machines. This game is about one human pushing his machine’s horizon and the other failing to do so – resulting in an highly unusual piece sacrifice that his machine on its own didn’t see coming.

Also the game is a hot candidate for the “Most useless pair of bishops” award 😉


For you to replay and analyze more conveniently I’ve put all the moves and lines below into a Lichess study. The whole thing with text commentary as seen below also represents the latest addition to our database with annotated games and opening surveys.

Garry Kasparow used to be the leading authority on the Najdorf for decades. After quitting professional chess he perpetuated his collected Najdorf wisdom by recording a number of Najdorf DVDs for ChessBase. There he also deals with the line played in the game below that was his Najdorf pet line in the later years of his career.

Villain (1.577) – Hero (2.206)
Lechenicher Schachserver 2017, Sicilian (Najdorf)

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6


The point of …a6 is not only to control the b5 square and to prepare leashing out with …b5. It is also a useful waiting move that prevents White from going g4 Keres style immediately (as he would after 5…e6). 5…a6 forces White to show his hand.

6. Be3

White sets himself up for an English attack: f3, Qd2, 0-0-0, g4 and so on. As so often in the Najdorf Black has to make the principal choice if he wants to go …e6 or …e5. At this point he can chose a third option that used to be Kasparov’s pet move in the final years of his chess career and is still quite hot.

6… Ng4


Compared to White’s three pieces Black has developed one, and now he moves it again instead of further developing his army. Every beginner’s manual would despise this move that was first played by US grandmaster Walter Browne in 1970 and later adopted and heavily analyzed by Kasparov. The point of …Ng4 is to avoid facing an English attack. White cannot exploit Black moving the knight a second time, he has to take care of his Be3 now, either 7.Bc1 or 7.Bg5, everything else makes no sense. After 7.Bc1 Nf6 White either needs to play something else than 8.Be3, or he has to concede a draw by repetition.

7. Bg5 

After 7.Bg5 the game takes an individual course. Had the villain played 7.Bc1 Nf6 8.Be3, it would have been upon me to decide between …e6 and …e5. Draw by repetition against a much weaker opponent was no option, especially in this preliminary round game of the server anniversary tournament, in which the points scored in the prelims will also count in the final. This system forces everyone to crush mercilessly from the get-go in order to win the tournament.

7… h6 8. Bh4

(8. Bc1 is a rare, but interesting alternative. White invests a tempo to provoke …h6, a mark that he will later use when his g2-g4-g5 attack is rolling.)

8… g5 9. Bg3 Bg7


The black setup is provocative and doubleedged. He has losened his structure quite a bit in order to get activity and black square control. With this approach Black forces himself to continue actively and precisely. Once White manages to take over and dictate the course of action, the question where to go with the black king and how to keep things together in general will not be answered easily.

10. Be2 h5


11. h4

11.Bxg4 and 11.Nf5 are good alternatives. The text move immediately undermines Black’s losened kingside. In the long run h5 may be a weakness.

11… Nc6


12. Nb3

From now on the option …Bxc3 bxc3 will be tempting each move. It completely shatters White’s queenside and highlights that both kings aren’t feeling safe. But giving up a bishop as proud as the one on g7 is not an easy thing to do.

(12. Nf5?! Bxf5 13. exf5 Qa5 and Black is happy.)

12… gxh4 13. Bxh4


13… Be6 

This is still a very theoretical position, and I was surprised to see my opponent coming at me like this. Usually the guys with a medium to low rating don’t care much for theory, but this one seemed to push it. I was expecting 14. Qd2, the move most commonly played among grandmasters, and hadn’t made up my mind yet how to respond. 14…Qb6 is the most popular response, 14…Rc8 seems to be a good move as well, but there may also be alternatives and room for investigation.

(13… Bxc3+ 14. bxc3 Qc7 15. Qd2 Bd7 According to Najdorf guru MVL something like this would be doubleedged, but probably easier to play for White, “who can dream to be ahead on both sides of the board”. Who am I to not follow his advice?)

14. Kf1


This one hit me as a surprise. In the middle of a theoretical debate suddenly my opoonent choses a move that hasn’t been played before, neither OTB nor in correspondence chess – and that doesn’t seem to accomplish anything. 14.Kf1 doesn’t look like an improvement out of his analysis laboratory. But when I broadened my investigation to my database of engine games I was surprised yet again. Among machines 14.Kf1 is the second most popular move, favoured by Stockfish 8 and Stockfish 7. So it seemed that for some reason my opponent had decided to abandon everything theoretical and rely on his machine alone. At least I knew now that I was playing Stockfish since no human would ever look at the position after 13 moves and then decide
that 14.Kf1 must be the way to go.

14… Bxc3

Not an easy decision. 14…Bxc3 gives Black something to play against, creates an imbalance and adds some strategical features to a position that could easily develop into a tactical slugfest once White goes Nd5. For instance after 14…Rc8 15.Nd5! or 14…Nce5 15.Nd5 play will have a much more forced character and will be easy to handle for an engine. So 14…Bxc3 is a try to maximize the human factor in a sense that future decisions won’t be based on tactics alone (at least that’s what I thought).

15. bxc3 Rc8 16. Nd4


Of course Black won’t repair White’s structure via 16…Nxd4, and he also won’t move his Be6 from its best possible square. But that means that after Nxe6 the game may well develop into a duel pair of bishops vs pair of knights that often favours the bishops, one of the reasons why 14…Bxc3 wasn’t an easy decision. However, while the position is not exactly closed the black knights will find plenty of outposts, and the move …fxe6 improves Black’s king position. That guy will find shelter on d7 while his white counterpart on f1 can’t feel comfortable yet.

16… Qa5 17. Rb1 b5 18. Nxe6 fxe6


Black has made a series of normal moves, aimed at fixing the white weaknesses on the queenside, and it looks like the position is on the brink of being better for Black. …Nce5 and …Kd7 are coming, and if White doesn’t do anything Black will start to collect on the queenside.

19. c4 

An attempt to open things up.

(19. Bxg4? hxg4 20. Qxg4 is almost losing for White. The pin on the h file, the open f and g file and the shattered queenside give Black plenty to play against, more than White can handle.)

19… b4

Of course. Black keeps things closed and the opponent’s queenside pawns weak.

20. Rh3


Quite logical, activating a piece and planning to find some freedom and improvement via c2-c3.

(Again 20. Bxg4? hxg4 21. Qxg4 leads to a huge black advantage. Until now and during the next moves it’s a nice feature of the position that Black can constantly present the Ng4 outpost and never can White take it under favourable circumstances although he would like to. Both white bishops make a poor impression anyways, so the pair of bishops isn’t much of a factor. Actually, should there ever be an award for the most useless pair of bishops, this game would be a hot candidate.)

20… Rf8

Eyeing at the white king, a move that can be found by excluding everything else.

(20… Nce5 21. Qd2 and suddenly b4 is a problem.)

(20… Kd7 21. Qd2 Qxa2 22. Rd1 and the option c4-c5 gives White good counterchances.)

(20… Qxa2 21. Ra1 and White is fine.)

21. Kg1 Kd7 22. Qe1


22… Rb8

Frees the Nc6 from defending the b4 pawn and prepares further centralizing and activating the black forces. …Nce5 or …Nd4 along with …Qc5 or …Qe5 and …a5 are now on Black’s agenda while it is hard to see how White wants to improve anything. The verdict on the evaluation is now official: Black is better.

(22… Qxa2 is playable, but it gives up the nice grip Black has on the position. After 23. Rhb3 (or 23. c3) White will find some counterplay. If Black insists on being greedy and plays 23… Qxc2? after 24. f3! White even wins. The Ng4 is hanging, and Rbb2 threatens to win the Qc2.)

23. c5?

In combination with 24.c3 an attempt to free himself, open the position and finally get some play. From a human point of view this looks like a good practical choice, so in an OTB game I’d be reluctant to mark 23.c5 with a “?”. However, this is a correspondence game, and White now runs into some cool, deep tactics that he could have foreseen had he invested more time in order to investigate more carefully and deeper.

(To just hold still and try to survive somehow may be the best approach, for instance 23. Bg3 Qc5 24. Rd1 Nd4 25. Rd2 a5 and Black is more than happy, but there’s nothing decisive going on yet. It will be hard to make progress.)

23… Qxc5


24. c3

(After 24. Bxa6 Qxc2 the white position is much worse than the engine suggests at first.)

24… bxc3 25. Rc1

(25. Rxb8 Rxb8 26. Rxc3 doesn’t quite work. 26… Qe5 27. Bg3 Qxe4 with a large black advantage.)

25… Nd4!

(25… c2? 26. Rc3 is rather unclear.)

26. Bxg4

(26. Rhxc3 Qxc3 27. Rxc3 Nxe2+ 28. Qxe2 Rb1+ 29. Qf1 Rxf1+ 30. Kxf1 Rf4 (threatening …Nh2+) is almost hopeless for White being two pawns down.)


26… c2!!

Not your everyday piece sacrifice, this one is as counterintuitive as they come. Instead of taking back Black pushes his c passer. White is almost defenseless against Black doubling on the b file, then go …Rb1, …Rxc1 and …Rb1 again. The only thing he can do is resacrifice his light squared bishop on c2.

27. Bf3

(27. Bd1 cxd1=Q 28. Rxd1 (28. Qxd1? Qxc1-+) 28… Rb2 is similar to the game.)

27… Rb5!


The only square on the b file to prepare doubling. I’m not getting into lines now, too complicated, but Black needs to be extremely precise since White still had some defensive mechanisms going for him. For instance after 27…Rb2? 28.e5! the idea of closing the 5th rank in order to play Bg5 and cover the c1 square would have saved him. I suggest to look into this with an engine yourself if you’re interested.

28. Kh2 Rfb8 29. Bd1


Now he has to give back the piece.

29… Rb1 30. Bxc2 Qxc2 31. Rxc2 Rxe1 32. Rd2 Nc6


White is facing a dilemma. Obviously his e pawn is hanging. If he protects it via 33.f3 his Rh3 is cut off from the queenside, and Black can double on the b file, intrude on the 2nd rank, eventually capture the white a pawn and create a passer on the a file. If White counterattacks the h5 pawn instead via 33.Bg5 after 33…Rxe4 34.Rxh5 e5 Black has a mighty rolling pawn mass in the centre while White has only captured the h pawn that was weak anyways. Whatever White does, this endgame is much worse for him than it may look at first glance.

33. f3

(33. Bg5 Rxe4 34. Rxh5 e5)

33… Reb1 34. Bg5 R1b2 35. a3 Rxd2


White resigned.

Before executing 35…Rxd2 I had put in a long analysis session in order to make sure I can steer this towards a win. I wasn’t 100% sure yet, but well beyond 90 😉 To my surprise a few days later my opponent resigned, probably justified, but a bit too early for my taste. This endgame isn’t trivial yet. The following moves until 42…Rxa4 were my main line that leads to the point where I think it should be a win:

36. Bxd2 Rb2 37. Bc1 Ra2 38. Rxh5 Nd4 39. Kh3 e5 40. Rh8 Ra1 41. Bh6 a5 42. a4 Rxa4


Black should be able to convert.

5 1 vote
Article Rating
Benachrichtige mich bei

Inline Feedbacks
View all comments