Knight Endgames

Put a knight on a light square and move it: it lands on a dark square. Move it again: light. The color flips every single move. So a knight needs an even number of moves to come back to where it started, and it can never return to the same square in an odd number of moves.

A king can triangulate. A bishop can shuffle along a diagonal and hand the move back. A knight can do neither: it cannot pass, and it cannot spend one move going nowhere. That is why zugzwang with knights is so awkward to engineer, and why the side that wants to be to move often simply cannot arrange it.

The practical consequence: do not count on winning a knight ending by waiting. If your plan needs the opponent to run out of moves, check the square colors first. Very often the tempo you need does not exist.

Botvinnik's dictum, and it is a good one. A knight is short-range: it cannot switch wings in one move, it cannot create threats on both sides of the board at once, and it defends slowly. So knight endings behave like king-and-pawn endings with a slightly heavier piece attached: the extra pawn usually wins, and king activity usually decides.

This cuts both ways. Attacking, look for a passed pawn on the wing away from the enemy knight: the knight needs several moves to arrive, and it may simply be too slow. Defending, keep your knight centralized, where it reaches both wings, and keep your king in front of the pawns exactly as you would in a pawn ending.

Because the position resembles a pawn ending, the concepts from that family transfer directly: the opposition, passed pawns and zugzwang all still apply, with the tempo caveat above.

Stick a rook in front of a passed pawn and it goes passive. Stick a bishop there and it does nothing but stare down one diagonal. Stick a knight there and it blockades the pawn while still attacking things around it: the knight's moves radiate outward from the blockade square, so it keeps working.

The diagram at the top is the pure form. The knight on f7 does two jobs in one: it attacks the pawn on d6 and it covers the promotion square d8. The pawn cannot advance, and it cannot be defended into promoting. Black simply holds.

A lone knight can stop a single passed pawn on almost any file, provided it arrives in time: it must reach a square from which it hits either the pawn or its path. Get the knight in front of the pawn, not behind it: from behind it is always one move too slow. Against a rook pawn even this often fails, which is the next section.

Everything above breaks against an a-pawn or h-pawn. The edge of the board takes half the knight's moves away, and a knight in the corner is a knight in a cage.

A knight fighting a rook pawn on the seventh rank can find itself with the pawn attacked, the queening square covered, and still be lost, because it has no safe square to return to and cannot lose a tempo to fix its footing. This is the same family of problem as the wrong bishop: the rook pawn is the piece-endgame's great equalizer, and it defeats bishops and knights for different geometric reasons.

The practical rule, and note it runs the opposite way to the bishop: if you are the one pushing a passed pawn against a knight, steer for the rook pawn, because that is where the knight is weakest. If you are the one holding with the knight, avoid it, and if you cannot, get the knight in front of the pawn early. Play the rook-pawn drill below; it is the position that decides whether you actually understand knight endings or merely like them.