Dvoretsky's Endgame Manual Read online

  1 Qb4+ Ka2 2 Qc3 Kb1 3 Qb3+ Ka1! 4 Qe3 Kb1 5 Qd3 Kb2 6 Qe2!? Ka1!= (but not 6…Kb1? 7 Kc4! c1Q+ 8 Kb3+–).

  The win is possible only if the white king stands so close that it can help the white queen mate the enemy king.

  Let’s put the black king on d2. Now, in order to reach its stalemate haven, it will have to cross the c1-square, giving White the tempo he needs to win: 1 Qd4+ Ke2 2 Qc3 Kd1 3 Qd3+ Kc1 4 Kc4! Kb2 5 Qd2 Kb1 6 Kb3+–.

  1 Qb3+ Ka1! 2 Qd1+ Kb2 3 Qd2+ Kb1 4 Kb4! a1Q 5 Kb3+–

  Starting with the white king at e4, the mate is delivered in somewhat different fashion: 1 Qb3+ Ka1 2 Qc3+ Kb1 3 Kd3! a1Q 4 Qc2#.

  With the king any farther from the pawn, there is no win. I shall limit myself to just that general observation – I do not think it makes any sense to reproduce the “winning zone” for each and every position of the black pawn which I have seen in other endgame texts. It is not worth memorizing – once you have mastered the winning and drawing mechanisms, you can easily figure out for yourself at the board what sort of position you’re facing.

  Of course, there are exceptions, in which the standard evaluations and techniques are no longer sufficient.

  J. Timman, 1980

  1 c5! e4 (1…f2 2 Bc4=) 2 Bd1!! e3 (2…f2 3 Be2=) 3 B×f3 K×f3 4 K×c6 e2 5 Kd7! e1Q 6 c6 ( 7 c7), and after 6…Qd1(d2)+ 7 Kc8=, Black cannot prevent 8 c7.

  In the final position, it is very important that White’s king is on d7. This is why 1 Bd1? c5! would have been a mistake, since the king can’t get to d7. And 1 Bc2? f2 2 Bd3 Kf3! (2…e4? 3 Be2=) 3 K×c6 e4 4 Bf1 e3 5 c5 e2–+ is also hopeless.

  And Black’s king must be drawn to f3 – with the king still on g3, Black wins by 6…Qd1+! 7 Kc8 Qg4+. After 1 c5! e4, the move 2 Bd1!! solves this problem, while 2 K×c6? e3 3 Bc4 e2 4 B×e2 fe 5 Kd7 e1Q leaves the black king on g3. White does no better with 3 Bd5 e2 4 B×f3 e1Q 5 Kd7 – Black manages to bring up his king by: 5…Qd2+ 6 Kc8 Kf4! 7 c6 Ke5 8 c7 Qb4! 9 Bb7 (9 Kd7 Qd6+ 10 Kc8 Qb6! 11 Kd7 Qe6+ 12 Kd8 Kd6–+) 9…Qf8+ 10 Kd7 Qd6+ 11 Kc8 Ke6 12 Kb8 Kd7 13 Bc8+ Kc6 14 Bb7+ Kb6–+.

  N. Elkies, 1986

  When is the right time to break with c4-c5? Right now it would obviously be premature: 1 c5? N×h6+ 2 Kf8 bc, or 2…Nf5 3 cb Nd6, and draws.

  And on 1 Kg7? N×h6 2 K×h6 Ke3! 3 c5 bc 4 b6 c4 5 b7 c3 6 b8Q c2 7 Kg5! (threatening the check at f4) 7…Kd2!, the white king is too far from the pawn. The single tempo that White gets when Black’s king occupies the c1-square is insufficient to win.

  1 h7!! Nf6+ 2 Kg7 N×h7 3 K×h7

  Now we have virtually the same position as in the preceding variation, with the king standing even farther from the queenside. But here, the h6-square is open!

  3…Ke3! 4 c5 bc 5 b6 c4 6 b7 c3 7 b8Q c2 8 Qh2!! c1Q (8…Kd3 9 Qf4+–) 9 Qh6+.

  Exercises

  Pawn Races

  Let’s examine the sort of situation where both players advance simultaneously and queen at the same time, or almost at the same time. Here, the following outcomes are possible:

  (1) One rook’s pawn prevents the other rook’s pawn from queening;

  (2) The pawn queens with check, and thereby prevents the enemy pawn from queening; or

  (3) We get a “queen vs. pawn (or pawns)” endgame.

  Or, if both pawns queen, then:

  (4) One queen is lost to a “skewer” check along the file or diagonal;

  (5) Mate follows;

  (6) The queens are exchanged, after which we once again have a pawn ending; or

  (7) We get a queen ending (either an elementary one, or one with some play to it).

  In order to get an idea of all these possibilities (except perhaps the last one), we shall present a sizeable number of examples. In the previous chapter we have already seen an ending which transposed into a “queen vs. pawns” endgame; and earlier, we also saw cases where the king fell into check, or the queen was lost to a skewer check (see exercises 1/4, 1/7, 1/8, 1/10). Quite often, the chief problem of a position is either to draw the enemy king onto a bad square, or to avoid such a square with one’s own king.

  G. Walker, 1841

  1 b4 f5 2 b5 f4 3 b6 f3 4 b7 f2 5 b8Q f1Q 6 Qb5+! Q×b5 7 K×b5 Kg4 8 a4, and the h-pawn will never become a queen.

  This simple example illustrates Points 1 and 6 of the above list; the following example is for Points 2 and 4 (perhaps the most important ones).

  J. Moravec, 1925

  The only move to draw is 1 Kd5! Kg2! (1…K×h2? 2 Ke4 Kg2 3 Ke3+–) 2 h4, and White’s pawn queens immediately after Black’s.

  On 1 Kf5? Kg2! the black pawn queens with check; while on 1 Ke5? Kg2! White’s queen will be lost after …Qa1+.

  N. Grigoriev, 1928

  Black’s king is in the square of the f-pawn, so the hasty 1 f4? Kb5! leads only to a draw. White has to block the king’s path to the kingside (“shouldering”!).

  1 Kd4! b5

  The other defensive plan gets instructively refuted: 1…Kb5 2 Kd5! Ka6 3 f4 Kb7 4 f5 Kc7 5 Ke6 Kd8 6 Kf7! b5 7 f6 b4 8 Kg7 b3 9 f7 b2 10 f8Q+. In a practical game, nearly every pawn for some reason ends up queening with check; there are times, however, when you have to work a little bit for it!

  Interestingly, if we place the pawn on b7 in the starting position, Black saves himself by 1…Kb5! 2 Kd5 Kb6! 3 Kd6 Ka7 4 f4 b5.

  2 f4 b4 3 f5 b3

  Now the enemy king must be drawn to a checkable square.

  4 Kc3! Ka3 5 f6 b2 6 f7 b1Q 7 f8Q+, mating or winning the queen.

  Tragicomedies

  Ljubojevic – Browne

  Amsterdam 1972

  Recognize this position? Yes, it is the Grigoriev study we just examined, except with colors reversed and the black king positioned differently (which has no meaning here). 1…Kd5! would have won; instead, GM Browne played 1…f5??, and after 2 Kb4, a draw was agreed.

  Mohr – Conquest

  Gausdal 1989

  After Conquest’s move, 1…Kc1?, the position became drawn: 2 g7 b1Q 3 g8Q=.

  Black could have won by 1…Nd5! 2 K×d5 (2 g7 Ne7 3 Ke6 Ng8 4 Kf7 Kc2 leads to a won “queen vs. knight’s pawn” endgame) 2…Kc1 3 g7 b1Q 4 g8Q Qb3+.

  Gavrikov – Kharitonov

  USSR ch(1), Sverdlovsk 1984

  The winning idea is 1 Kc5! K×h5 2 b4 Kg4 3 a4 h5 4 b5 ab 5 a5!, when the white pawn queens, while preventing the black one from doing so.

  The game line was 1 Ka5? K×h5 2 K×a6 Kg4 3 b4 h5 4 b5 h4 5 b6 h3 6 b7 h2 7 b8Q h1Q, with a drawn queen endgame.

  Golombek – Keres

  Margate 1939

  1…Be5 2 Kd2 Bg3 3 K×c2 B×h4 would have won. Even easier was 1…Bc3! with zugzwang: if 2 Ke3, then 2…Be1.

  Instead, Keres played 1…Bb2?, and his opponent resigned, believing that after 2 Kd2 c1Q+ 3 N×c1 B×c1+ 4 K×c1 Ke5 5 Kc2 K×e4, his attack on the a5-pawn would come too late: 6 Kb3 Kf4 7 Ka4 Kg4 8 K×a5 K×h4 9 a4 K×g5 10 Kb6 h4 11 a5 h3 12 a6 h2 13 a7 h1Q–+.

  But White’s king can also attack the other black pawn: 6 Kc3! Kf4 7 Kd4 Kg4 8 Ke5 K×h4 9 Kf6 Kg4 10 K×g6 h4 11 Kf6 h3 12 g6, with a draw.

  Exercises

  The Active King

  King activity is the most important factor in the evaluation of position in a pawn endgame. In fact, not just in pawn endgames – in any endgame. But in pawn endgames, where there are no other pieces on the board, this is perhaps an especially important factor.

  The influence the degree of king activity has on the battle’s outcome is obvious in many of the preceding and succeeding examples. Here, we examine two vitally important means of exploiting an active king’s position: playing for zugzwang, and the widening of the beachhead.

  Zugzwang

  M. Dvoretsky, 2000

  1 g3! Kd7 2 g4 Ke7 3 g5

  3 d4? Kd7 4 Kf5 Kd6 5 Kg6 Kd5 6 K×g7 K×d4 would be less exact, as the pawns both queen. However, White could transpose moves by 3 Kf5 Kd6 4 g5! (4 Kg6? Ke5! 5 K×g7 Kf4 6 Kf6 K×g4=) 4…Ke7 (4…Kd5 5 Kg6 Kd4 6 K×g7 K×d3 7 g6 c5 8 Kf6 c3 9 g7 c3 10 g8Q c2 11 Qg5+–) 5 Kg6 etc.

  Let’s think abo
ut the position after 3 g5. White’s king dominates the board – that is why zugzwang is unavoidable. In fact, whichever pawn Black moves is bound to be lost (3…c5 4 Kd5, or 3…g6 4 d4). Retreating the king to f7 clears the way for his opponent to go after the c6-pawn. That leaves just one move; but after that move, White finally executes his main plan – getting his king to h7.

  3…Kd7 4 Kf5 Ke7 5 Kg6 Kf8 6 Kh7 Kf7 7 d4

  One final, decisive zugzwang.

  If it were Black’s move in the starting position, then after 1…Kd7, 2 g4! would lead to the win. It sometimes makes quite a difference if you have a choice between moving a pawn one or two squares. For a more detailed examination of this, see “Steinitz’s Rule,” and the chapters which follow.

  Widening the Beachhead

  Hansen – Nimzovitch

  Randers simul 1925

  Who stands better? White intends to play c2-c3, obtaining an outside passed pawn, which will secure him a decisive advantage (for example, after 1…c5? 2 dc+ K×c5 3 c3).

  Nimzovitch hits upon the correct plan – he activates his king, even if it means sacrificing a pawn.

  1…Kc7! 2 c3 Kb6! 3 cb Kb5 4 Kc3 Ka4

  As Black had foreseen, it is zugzwang. White resigned, since he has to lose all his queenside pawns: 5 Kc2 K×b4 (White still has the outside passed pawn, but the activity of Black’s king means far more here) 6 Kd3 Ka3 7 Kc3 K×a2 8 Kb4 Kb2 9 Kc5 Kc3–+.

  Let’s look at 3 c4 (instead of 3 cb). White will continue by exchanging pawns at d5. It is not hard to see that b5-b3 and a4-b2 are corresponding squares; after that, we can establish a third pair of corresponding squares: a5-c2. Now we understand that Black must inevitably take advantage of this correspondence (since he can wait on either of the equivalent squares b6 and a6, while White cannot).

  3…Ka6! 4 cd cd 5 Kc2 Ka5 6 Kb2 Ka4 7 Kc2 Ka3 8 Kb1 b3 9 Ka1 Kb4 10 Kb2 ba

  In order to win, Black cleared a path for his king towards the center. This is, in fact, what “widening the beachhead” means – trading off pawns, with the idea of clearing a path for the king.

  Let’s examine another classic endgame.

  Cohn – Rubinstein

  St. Petersburg 1909

  With 1 f4!, White would have had an inferior but defensible position. Instead, he decided to exchange rooks, because he had misjudged the pawn endgame.

  1 Rc1? R×c1 2 K×c1 Kf6 3 Kd2 Kg5

  Rubinstein sends his king to h3, in order to tie White’s king to the defense of the weak pawn at h2. It is not difficult to calculate that White’s counterattack with 4 Kd3 Kh4 5 Kc4 comes too late.

  4 Ke2 Kh4 5 Kf1 Kh3 6 Kg1 e5 7 Kh1 b5!

  It is useful to fix the queenside pawns, while Black also leaves himself the reserve tempo a7-a6. White could have prevented this by playing 7 a4!?, but it would not have altered the assessment of the position.

  8 Kg1 f5

  Black’s further plan is to “widen the beachhead” – clear a path for his king to the queenside via pawn exchanges.

  9 Kh1 g5 10 Kg1 h5 11 Kh1 g4

  11…h4 12 Kg1 g4 13 fg K×g4 14 Kg2 h3+ is also strong.

  12 e4 fe! 13 fe

  13 fg hg 14 Kg1 e3 15 fe e4 16 Kg1 g3 is no better.

  13…h4 14 Kg1 g3 15 hg hg White resigned.

  12 fg (instead of 12 e4) 12…fg! 13 Kg1 e4 14 Kh1 h4 15 Kg1 g3 changes nothing – Black still wins. However, taking with the other pawn – 12…hg? – would have been a serious inaccuracy: 13 Kg1 f4 14 ef ef 15 Kh1.

  Here, widening the beachhead no longer wins: 15…g3? 16 hg fg 17 fg (17 Kg1=) 17…K×g3 18 Kg1 Kf3 19 Kf1 Ke3 20 Ke1 Kd3 21 Kd1 Kc3 22 a4!=.

  The right plan, 15…f3! 16 Kg1 Kh4, was pointed out by Jonathan Mestel.

  17 Kh1 Kg5 18 h3 gh 19 Kh2 Kg4 20 Kg1 Kf4 21 Kh2 Ke4 22 K×h3 (22 Kg3 h2!) 22…Kd3 23 Kg4 Ke2 24 Kg3 a6–+ (here is where the reserve tempo comes in handy!)

  17 Kf1 Kh5! 18 Ke1 Kg5 19 Kf1 (19 Kd2 Kh4–+) 19…Kf4 20 Ke1 Ke4 21 Kd2 Kd4 22 Kc2 Kc4 23 Kd2 Kb3 24 Ke3 K×a3 25 Kf4 K×b4 26 K×g4 a5–+.

  Tragicomedies

  Horowitz – Denker

  Philadelphia 1936

  Here is how the game actually ended: 1 Kh7 Kf7 2 Kh8 Kf8 3 g5 Black resigned.

  Zinar has shown that every move played by both sides was a mistake – except for the last one. His analysis follows:

  (1) White should not take his king into the corner. The correct plan to exploit his advantage is – widening the beachhead!

  1 g5! hg 2 K×g5 Kf7 3 h5 Ke7 4 Kg6 Kf8 5 h6! Kg8! 6 Kh5! gh 7 K×h6 Kf7 8 Kh7 (the opposition) 8…Kf6 9 Kg8 (now, an outflanking) 9…Kg5 10 Kf7 Kf4 11 Ke6 K×e4 12 K×d6 Kf4 13 K×c5 e4 14 d6 e3 15 d7 e2 16 d8Q e1Q 17 Qf6+, with an easily won queen endgame.

  (2) With 1…h5! (instead of 1…Kf7?), Black would have drawn: 2 g5 Kf7 3 Kh8 Kg6! 4 Kg8 stalemate; or 2 gh Kf7 3 h6 g6! 6 Kh8 Kf8=.

  (3) But 2 Kh8? lets slip the win. Also insufficient was 2 g5? h5! 3 g6+ (3 Kh8 Kg6!) 3…Kf6 4 Kg8 K×g6 5 Kf8 Kf6 6 Ke8 g5 7 hg+ K×g5 8 Ke7 h4 9 K×d6 h3 10 Kc7 h2 11 d6 h1Q 12 d7 Qh7 13 Kc8 Qh3=.

  The right move was 2 h5! Kf6 (we have already seen what happens after 2…Kf8 3 Kg6 Kg8 4 Kf5 Kf7 5 g5 hg 6 K×g5) 3 Kg8 g6 (3…g5 4 Kh7) 4 Kf8! gh 5 gh Kg5 6 Ke7 K×h5 7 K×d6 Kg4 8 K×e5+–.

  (4) And retreating the king to f8 was the decisive mistake. Black could still have drawn with 2…h5! 3 g5 Kg6!, or 3 gh Kf8 4 h6 g6!.

  Exercises

  The King Routes

  In this section, we shall examine some different types of king maneuvers.

  Zigzag

  The laws of geometry, as we have known them since grade school, have no relevance on the chessboard. There, a straight line is not the shortest distance between two points (or squares) – if the king follows a broken-line path, it is by no means longer. This phenomenon is exploited both in the Réti idea we have already examined, and in the “shoulder block” we shall learn later on.

  Here, we shall speak of a technique closely connected with the simultaneous advance of pawns we just studied. To be more exact: we shall be speaking of two techniques, which look very similar. Let’s call them “zigzag.”

  N. Grigoriev, 1928

  The direct 1 g4? leads only to a draw: 1…b5 2 g5 b4 3 g6 b3+ 4 Kc3 b2 5 g7 b1Q 6 g8Q+ Ka1!=.

  1 Kc3! Ka3 2 Kc4 Ka4 3 g4 b5+ 4 Kd3!

  Here is the zigzag! The king returns to c2, while avoiding the pawn check.

  4…Ka3 5 g5 b4 6 g6 b3 7 g7 b2 8 Kc2! (drawing the king into check) 8…Ka2 9 g8Q+.

  The other form of zigzag occurs when the king has to avoid a check from a newly-promoted queen.

  J. Moravec, 1952

  White is outside the square of the a-pawn. His only hope is Réti’s idea.

  1 Kg4 a5 2 Kf5! a4

  Otherwise, the king gets into the square. Now White would lose by 3 e6? Kd8 4 Kf6 Ke8; and 3 Kf6? a3 4 e6 a2 5 e7 a1Q+ is also bad. The king must avoid the f6-square.

  3 Kg6! a3 4 e6 a2 5 e7 Kd7 6 Kf7=.

  Exercises

  How should this game end?

  The Pendulum

  1 Kg3! Ke3 2 Kg2! Ke2 (2…f4 3 Kf1) 3 Kg3=

  This elementary defensive technique appears frequently.

  Exercises

  Shouldering

  Quite often, one must choose a route for the king that gives a “shouldering” to the enemy king – that is, it prevents the enemy from arriving in time at an important part of the board.

  Schlage – Ahues

  Berlin 1921

  White must inevitably win the pawn at a7. Black can save himself only if he can succeed in locking the white king into the corner with …Kc7.

  The game was drawn after 1 Ke6 Kc3 2 Kd6? Kd4 3 Kc6 Ke5 4 Kb7 Kd6 5 K×a7 Kc7.

  Maizelis demonstrated a win for White by 1 Ke6! Kc3 2 Kd5!+–

  White’s king approaches the a7-pawn while simultaneously “shouldering” the enemy king, keeping it from approaching the c7-square.

  J. Moravec, 1940

  White only gets a draw out of 1 Ka2? Kg2 2 Kb3 Kf3 3 Kc4 Ke4 4 b4 Ke5 5 Kc5 (White’s king does manage to
shoulder the enemy king, but here this is insufficient) 5…Ke6 6 Kb6 (6 b5 Kd7 7 Kb6 Kc8=) 6…Kd5 7 K×b7 Kc4=.

  It is important to keep Black’s king farther away from the pawns; and for this, White needs to meet him halfway.

  1 Kb1! Kg2 2 Kc2 Kf3 3 Kd3! Kf4 4 Kd4 Kf5 5 Kd5 Kf6 6 Kd6 Kf7

  If 6…Kf5, then 7 b4 Ke4 8 b5 Kd4 9 b6 with the idea of 10 Kc7+–.

  7 b4 Ke8 8 Kc7 b5 9 Kc6+–.

  Tragicomedies

  Rogers – Shirov

  Groningen 1990

  Black would have drawn had he continued 1…f5+! 2 K×f5 (2 Kd4 Kg2) 2…K×f3 3 R×h2 Kg3 4…f3. 1…f6! is also possible: 2 Rh8 f5+ (or even 2…Kg2).

  Shirov decided instead to pick up the rook for his h-pawn, but he misjudged the pawn ending.

  1…Kg2?? 2 K×f4 h1Q 3 R×h1 K×h1 4 Kg3!

  Black resigned. His king is squeezed into the corner, giving White time to push his f3-pawn forward, after which he can win Black’s pawn. For example: 4…Kg1 5 f4 Kf1 6 f5! (but not 6 Kf3? f5!=) 6…Ke2 7 Kf4 Kd3 8 Ke5 Ke3 9 f6!+–

  Exercises

  Breakthrough

  A breakthrough occurs when one or more pawns are sacrificed in order to create a passed pawn and promote it.

  Let’s examine a few of the standard structures in which a pawn breakthrough is possible.

  White to move wins by 1 b6! cb (1…ab 2 c6) 2 a6! ba 3 c6

  Black to move has only one way to parry the threatened breakthrough: by 1…b6! (both 1…a6? 2 c6! and 1…c6? 2 a6! are bad).