Chess tournaments Chess strategy Computer chess Chess players FIDE Chess variants Chess rules and history

Two knights endgame

The two knights endgame is a chess endgame with a king and two knights versus a king, possibly with some other material. The material with the defending king is usually one pawn, but some positions studied involve additional pawns or other pieces. In contrast to a king plus two bishops (on opposite-colored squares), or a bishop and a knight, a king and two knights cannot force checkmate against a lone king. (However, the superior side can force stalemate.) Although there are checkmate positions, the superior side cannot force them against proper (and easy) defense (Speelman, Tisdall & Wade 1993:11).

a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Checkmate position, but it cannot be forced (Seirawan 2003:17). The knight on d2 could be on c3 or a3 instead, and the white king could be on a3 instead.

On the other hand, if the lone king has a pawn (and sometimes with more pawns), then checkmate can be forced in some cases. These positions were studied extensively by A. A. Troitzky. If the defender's pawn is securely blocked on or before the "Troitzky line", the stronger side can force checkmate, although it may require up to 115 moves with optimal play. The technique (when it is possible) is to block the pawn with one knight and use the king and other knight to force the opposing king into a corner or near the other knight. Then when the block on the pawn is removed, the knight can be used to checkmate (Dvoretsky 2006:280).

Two knights cannot force checkmate

Although there are checkmate positions with two knights against a king, they cannot be forced. Edmar Mednis stated that this inability to force checkmate is "one of the great injustices of chess" (Mednis 1996:40).

Unlike some other theoretically drawn endgames, such as a rook and bishop versus rook, the defender has an easy task in all endings with two knights versus a lone king. The player simply has to avoid moving into a position in which he or she can be checkmated on the next move, and always has another move available in such situations (Speelman, Tisdall & Wade 1993:11).

Three knights and a king can force checkmate against a lone king within twenty moves (unless the defending king can win one of the knights) (Fine 1941:5-6).

In the corner

Keres
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Two knights cannot force checkmate

The player with the lone king has to make a blunder to be checkmated. In this position, 1.Ne7 or 1.Nh6 immediately stalemates Black. White can try instead:

1. Nf8 Kg8
2. Nd7 Kh8
3. Nd6 Kg8
4. Nf6+

and now if Black moves 4...Kh8?? then 5.Nf7# is checkmate, but if Black moves

4... Kf8

then White has made no progress (Keres 1984:2-3).

Berger
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Draw with either side to move

Johann Berger gave this position, a draw with either side to move. With White to move:

1. Nf5 Kh8
2. Ng5 Kg8
3. Ne7 Kf8! (Black just avoids 3...Kh8? which leads to a checkmate on the next move with 4.Nf7#)
4. Kf6 Ke8

and White has made no progress. With Black to move:

1... Kh8
2. Nf7 Kg8
3. Nh6 Kh8
4. Ng5

gives stalemate (Guliev 2003:74).

On the edge

a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White can also try for a mate on the edge of the board

There are also checkmate positions with the inferior side's king on the edge of the board (instead of the corner), but again they cannot be forced ("Chess program Wilhelm".  + "Nalimov Engame Tablebases". AutoChess. ). In the position at right, White can try 1. Nb6+, hoping for 1...Kd8?? 2.Ne6#. Black can easily avoid this with, for example, 1... Kc7. This possible checkmate is the basis of some problems (see below).

Examples from games

Benko vs. Bronstein, 1949
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Position before 106. Nh2+

In this position from a 1949 game between Pal Benko and David Bronstein, Black had just underpromoted to a knight (104...f2-f1=N+ 105.Kd2-c3 Kg2-f3). Black did not promote to a queen or any other piece because White could fork Black's king and his newly promoted piece (104...f1=Q 105.Ne3+) immediately after the promotion. White made the humorous move

106. Nh2+

forking Black's king and knight, but sacrificing the knight. Black responded

106... Nxh2

and a draw was agreed (Benko 2007:133).

Another example is the eighth game from the 1981 World Chess Championship match between Anatoly Karpov and Viktor Korchnoi.

Troitzky line

The Troitzky line
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Troitzky line positions when White has the two knights and Black the pawn.

Müller and Lamprecht 2001
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Black loses no matter where the kings are. (Müller & Lamprecht 2001)

Kling & Horwitz 1851.
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to play and win.

1. Kh4 Kg2 2. Kg4 Kg1 3. Kh3 Kh1 4. Ng3+ Kg1 5. Nf3#

The pawn does not move; it assists mate by blocking the king's escape.

Whilst two knights cannot force checkmate (with the help of their king) against a lone king, a decrease in material advantage allowing the defending king to have a pawn can actually cause his demise. The reason that checkmate can be forced is that the pawn gives the defender a piece to move and deprives him of a stalemate defense (Müller & Lamprecht 2001:19-20). Another reason is that the pawn can block its own king's path without necessarily moving (e.g. Kling & Horwitz position right).

The Troitzky (or Troitsky) line (or Troitzky position) is a key motif in chess endgame theory in the rare but theoretically interesting ending of two knights versus a pawn.

The line, assuming White has the two knights and Black the pawn, is shown left.

The Russian theoretician Troitsky made a detailed study of this endgame and discovered the following rule:

If the pawn is securely blockaded by a white knight no further down than the line, then Black loses, no matter where the kings are.

An example of the application of this rule is given in the diagram Müller and Lamprecht right; "... the position would be lost no matter where the kings are". (Müller & Lamprecht 2001)

However, the checkmate procedure is difficult and long. In fact, it can require up to 115 moves by White, so in competition often a draw by the fifty-move rule will occur first (but see this article and Second Troitzky line section for the zone where the win can be forced within fifty moves).

Troitsky showed that "on any placement of the black king, White undoubtedly wins only against black pawns standing on [the Troitzky line] and above" (Rabinovich 2012:88).

John Nunn analyzed the endgame of two knights versus a pawn with an endgame tablebase and stated that "the analysis of Troitsky and others is astonishingly accurate" (Nunn 1995:265). He undertook this checking after the very ending occurred in a critical variation of his post mortem analysis of a game he lost to Korchnoi in the 1980 Phillips and Drew Tournament in London. Neither player knew whether the position was a win for the player with the knights (Korchnoi). A spectating Grandmaster, M. Quinteros, commented that Bent Larsen (also playing in the event) might know.

Even when the position is a theoretical win, it is very complicated and difficult to play correctly. Even grandmasters fail to win it. Andor Lilienthal failed to win it twice in a six-year period, see Norman vs. Lilienthal and Smyslov vs. Lilienthal. But a fine win is in a game by Seitz, see Znosko-Borovsky vs. Seitz (Giddins 2012:26).

Two knights versus pawn is sometimes called the "Halley's Comet" endgame.

Examples

a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move wins

This diagram shows an example of how having the pawn makes things worse for Black (here Black's pawn is past the Troitsky line), by making Black have a move available instead of being stalemated.

1. Ne4 d2
2. Nf6+ Kh8
3. Ne7 (if Black did not have the pawn at this point, the game would be a draw because of stalemate)
3... d1=Q
4. Ng6#

If Black did not have the pawn move available, White could not force checkmate.

a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Black to move wins in 115 moves

The longest win is this position that requires 115 moves, starting with 1... Ne7 (this is not the only possible example of position where 115 moves are required to win).

Pawn beyond the Troitsky line

Chéron, 1955
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White wins with either side to move

In this study by André Chéron, White wins even though the pawn is well beyond the Troitsky line (Müller & Lamprecht 2001:20).

Averbakh & Chekhover, position #251
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Drawing area marked with "x"

White to play draws. Black to play loses. (Averbakh & Chekhover 1977:119-120).

In the situation with Black's rook pawn blockaded on h3, if the black king can enter and remain in the area marked with crosses in the diagram to the right, the game is a draw. Otherwise, White can force the black king into one of the corners not located in the drawing zone and deliver checkmate. Black cannot be checkmated in the a8-corner because the knight on h2 is too far away to help deliver mate: Black draws by pushing the pawn as soon as White moves the knight on h2. White to play in the diagram can try to prevent Black to enter the drawing zone with 1.Ke6, but Black then plays 1...Kg5 aiming to attack the knight on h2. White is compelled to stop this with 2.Ke5 which allows Black to return to the initial position with 2...Kg6, and White has made no progress (Averbakh & Chekhover 1977:119-120).

Topalov versus Karpov

Topalov vs. Karpov, 2000
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White won after 74. Ne2, even though the pawn was past the Troitzky line.

Anatoly Karpov lost an endgame with a pawn versus two knights to Veselin Topalov although he had a theoretical draw with a pawn past the Troitzky line; because of its rarity, Karpov seemed not to know the theory of drawing and headed for the wrong corner. (Depending on the position of the pawn, checkmate can be forced only in certain corners (Troitzky 2006).) In this "rapid play" time control, the position in the game was initially a draw, but Karpov made a bad move which resulted in a lost position. Topalov later made a bad move, making the position a draw, but Karpov made another bad move, resulting in a lost position again.

Wang versus Anand

Wang vs. Anand, 2009
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Position after 61. Kxa5

This position from a blindfold game between Yue Wang and Viswanathan Anand leads to an example with a forced win even though the pawn is past the Troitsky line. The game continued

61... Kc5,

blocking the pawn with the wrong piece. Black should have played 61...Ne4 62. c4 Nc5!, blocking the pawn on the Troitsky line with a knight, with a forced win. The game continued:

62. c4 Ne4
63. Ka4 Nd4
64. Ka5.

Black still has a forced win in this position, even after letting the pawn advance past the Troitsky line:

64... Nc6+
65. Ka6 Kd6!!
66. c5+ Kc7

and Black has a forced checkmate in 58 more moves (Soltis 2010:42). However, the actual game was drawn.

Second Troitzky line

a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Second Troitzky line

Karsten Müller asked for the "second Troitzky line", which corresponds to where the knights can win without the fifty-move rule coming into effect. If Black's pawn is blocked by a white knight on or behind one of the dots, White can force a win within fifty moves. If the pawn can be blocked on or behind one of the Xs, White can force a win within fifty moves more than 99 percent of the time.

More pawns

Fine & Benko, diagram 201
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move wins in 96 moves
Fine, ECE #1778
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move wins

Two knights can win in some cases when the defender has more than one pawn. First the knights should blockade the pawns and then capture all except one. The knights cannot set up an effective blockade against four connected pawns, so the position generally results in a draw. Five or more pawns usually win against two knights (Fine & Benko 2003:101).

Example from game

Motwani vs. I. Gurevich
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
Position after Black's 72nd move

In this 1991 game between Paul Motwani and Ilya Gurevich, Black has blockaded the white pawns. In ten moves, Black won the pawn on d4. There were some inaccuracies on both sides, but White resigned on move 99 (Speelman, Tisdall & Wade 1993:114).

Position of mutual zugzwang

Troitsky
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move draws; Black to move loses

There are positions of mutual zugzwang in the endgame with two knights versus one pawn. In this position, White to move draws but Black to move loses. With Black to move:

1... Kh7
2. Ne4 d2
3. Nf6+ Kh8
4. Ne7 (or 4.Nh4) d1=Q
5. Ng6#

With White to move, Black draws with correct play. White cannot put Black in zugzwang:

1. Kf6 Kh7
2. Kf7 Kh8
3. Kg6 Kg8
4. Ng7 Kf8
5. Kf6 Kg8
6. Ne6 Kh7! (but not 6...Kh8? because White wins after 7.Kg6!, which puts Black to move)
7. Kg5 Kg8
8. Kg6 Kh8

and White has no way to force a win (Averbakh & Chekhover 1977:106).

Checkmate in problems

The possible checkmate on the edge of the board is the basis of some composed chess problems, as well as variations of the checkmate with two knights against a pawn.

Angos

Angos, 2005
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move and mate in four

In this problem by Alex Angos, White checkmates in four moves:

1. Ne6! Nd8
2. Nf6+ Kh8
3. Ng5 N-any (Black is in zugzwang and any knight move must abandon the protection of the f7-square)
4. Nf7# (Angos 2005:46).
Berger, 1890
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move and mate in four

A similar problem was composed by Johann Berger in 1890. The solution is:

1. Nf7! Nd6
2. Nh6+ Kh8
3. Ng5

followed by

4. Ngf7# (Matanović 1993:492-93).

de Musset

Alfred de Musset, 1849
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move and mate in three

In this composition by Alfred de Musset, White checkmates on the edge of the board in three moves with:

1. Rd7 Nxd7
2. Nc6 N-any
3. Nf6# (Hooper & Whyld 1992).

Sobolevsky

P. Sobolevsky, Shakhmaty v SSSR, 1951
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move and win

In this study composed by Sobolevsky, White wins by checkmating with two knights:

1. Nh8+ Kg8
2. Kxg2 Bf4
3. Ng6 Bh6!
4. Ng5 Bg7!
5. Ne7+ Kh8
6. Nf7+ Kh7
7. Bh4! Bf6!
8. Ng5+ Kh6
9. Ng8+ Kh5
10. Nxf6+! Kxh4
11. Nf3# (Nunn 1981:6).

Nadanian

Ashot Nadanian, ChessBase, 2009
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move and win

In this study composed by Ashot Nadanian, White wins by checkmating with two knights:

1. Rg8!! Rxg8

If 1...Re7, then 2.N6f5! Re1 3.Rxg6+ Kxh5 4.Rxh6+ Kg5 5.Nf3+ and White wins.

2. Ne4+ Kxh5
3. Ne6

and checkmate on the next move, due to zugzwang; two white knights deliver four different checkmates:

History

The first known composition where two knights win against one pawn is, according to Lafora, by Gioachino Greco in 1620. In 1780, Chapais did a partial analysis of three positions with the pawn on f4 or h4 (Troitzky 2006:200). In 1851 Horwitz and Kling published three positions where the knights win against one pawn and two positions where they win against two pawns (Horwitz & Kling 1986:64-68). The analysis by Chapais was revised by Guretsky-Cornitz and others, and included by Johann Berger in Theory and Practice of the Endgame, first published in 1891. However, the analysis by Guretsky-Cornitz was incorrect and the original analysis by Chapais was in principle, correct (Troitzky 2006:200). Troitsky started studying the endgame in the early 20th century and published his extensive analysis in 1937 (Mednis 1996:43). Modern computer analysis found it to be very accurate (Nunn 1995:265).

Master games with this ending are rare - Troitzky knew of only six when he published his analysis in 1937. In the first four (from c. 1890 to 1913), the weaker side brought about the ending to obtain a draw from an opponent who did not know how to win. The first master game with a win was in 1931 when Adolf Seitz beat Eugene Znosko-Borovsky (Troitzky 2006:197-99).

Horwitz & Kling, 1851
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move checkmates in six moves

Horwitz & Kling, 1851
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move wins

Pollock vs. Showalter, c. 1890
a b c d e f g h
8
8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
a b c d e f g h
White to move. Pollock refused to continue and agreed to a draw six moves later, but White has a winning position (Troitzky 2006:197).

Read more:

COMMENTS
Tabletop games: Rules and Strategy