Knight and Pawn vs Knight

Knight and pawn against knight is a draw in the overwhelming majority of positions: a knight that blockades the pawn — or even just controls a single square on its path — can usually hold forever, and it always has the emergency exit of giving itself up for the pawn. The extra pawn wins in two situations worth knowing cold: when the pawn has reached the seventh rank with the defending king far away, and — the great exception — when the pawn is a rook pawn, whose edge geometry turns the normally nimble defending knight into a trapped bystander. Every boundary position below is tablebase-verified.

I teach this ending right after knight vs pawn, because it is the same story with one twist: the attacker now owns a knight whose only real job is to evict the defender from the pawn's path. Whether that eviction can succeed before the defense reorganizes decides everything.

The Blockade Rule: One Square Is Enough

Start with the standard picture. White: Ke5, pawn d5, Nc3. Black: Kd7, Nd6 — the knight sits directly in front of the pawn, defended by its king. White to move, and the tablebase confirms the draw:

The blockade works because a knight, unlike a bishop, cannot lose a move — but also does not need to. Black's setup has no weakness to probe: the knight on d6 is defended, the pawn can never advance, and White's knight has no way to attack d6 that the black king cannot answer. If White ever engineers an eviction by force, Black has the standard fallback that governs this whole ending: the knight sacrifices itself for the pawn the moment the pawn becomes more valuable than the knight. Knight versus knight is a dead draw, so the defender's calculations are wonderfully simple — count whether the pawn can be stopped, and if not, count whether it can be eaten.

The defending priorities, in order: king in front of the pawn (best — then even losing the knight for nothing may not matter), knight blockading with king support (the diagram), and knight controlling one square on the pawn's path from a safe distance (adequate — the pawn cannot cross a guarded square, and every attempt to chase the knight costs the attacker tempi that the defending king uses to arrive).

When the Pawn Wins: The Seventh Rank

The defense above has one prerequisite hiding in plain sight: the defending pieces got organized in time. Let the pawn reach the seventh with the defender's king out of play, and no knight trickery saves it. White: Kd7, pawn c7, Na1. Black: Kg2, Ne4 — the black knight is close, the black king is not. Black to move, and White wins:

Black's checks run out with brutal speed: 1...Nc5+ 2.Kc6! and the knight, attacked, has no square that keeps c8 covered — the squares that guard c8 (a7, b6, d6, e7) are either unreachable or controlled by the white king. After 2...Ne6 3.c8=Q Nd8+ 4.Kd7 the last fork attempt fails and the queen wins trivially. Compare this with the blockade diagram and you see the whole evaluation method for this ending: the pawn on the seventh needs only one safe tempo; a pawn further back must cross several guarded squares. The attacker's plan is therefore always the same — advance the pawn with king escort before the defending king arrives, using your own knight to shield or to chase the defender off the path. The defender's plan is the mirror: king toward the pawn at maximum speed, knight delaying with checks.

A useful rule for quick evaluation: if the defending king reaches any square in front of the pawn, the game is drawn, full stop. If only the defending knight gets there, the game is drawn unless the pawn is already very advanced or the knight can be attacked by both king and knight at once. If neither defender arrives in time — the attacker wins, and the winning technique is no harder than the line above.

The Rook-Pawn Exception

Everything changes on the a- and h-files, and it changes against the defender — the opposite of most endgame lore, where rook pawns are the drawing files. A knight's power comes from its eight-square wheel; at the edge of the board the wheel loses half its spokes, and a blockading knight can be caged. White: Kb5, pawn a6, Nd1. Black: Kg2, Na7 — the knight has blockaded the pawn on a7, textbook procedure anywhere else. White to move wins:

1.Kb6! attacks the knight, and its troubles are immediate: every square in the knight's wheel except c8 is either off the board or covered by the white king. After 1...Nc8+ 2.Kc7 the knight must return — 2...Na7 — because abandoning the pawn's path loses instantly. Now comes the elegant part: 3.Nc3!, and White's knight takes away b5, the caged knight's last flight square. The bars close: c8 and c6 are covered by the king, b5 by the knight, and the a7-knight has literally no safe move. White strolls — 3...Kf1 4.Kb7 Ke1 5.Kxa7 — and the pawn promotes at leisure. Even the fact that Black's king then wins the c3-knight does not matter; the resulting king and pawn ending is elementary.

This is why, of all pawns, the rook pawn gives the attacker the best winning chances in knight endings — a paradox worth engraving, because it inverts the rule you learned in every other chapter of endgame theory. The defending knight in front of a rook pawn has at most two escape routes instead of the usual four or more, and the attacker's knight only needs to cover one of them. When defending against a rook pawn, prefer to fight it from the side or behind, keep your knight off the edge, and above all get the king there — the corner is a fine home for a defending king and a death trap for a defending knight.

The Knight-Trick Draws

The defender's tactical arsenal deserves its own mention, because it saves positions that look resignable. The knight's forking power means the attacker's final step — promotion — is exactly when catastrophe strikes: a knight controlling the promotion square from one hop away often wins the new queen by a check-and-fork, a theme demonstrated in detail (with a tablebase-verified example) in knight endgames. Checks are the second resource: as the pawn's escort king steps forward it wanders through a minefield of knight checks, and each check either drives it back or drags it onto a forkable square. This is why the winning method in the c7-pawn diagram had to be so precise — Kc6 was the only square from which every check could be answered while keeping the eviction going.

For the attacker the lesson is defensive alertness in victory: before you promote, look at every knight check within two moves and confirm none of them forks king and queen-to-be. More extra-pawn wins are thrown away at this final hurdle than anywhere else in the knight ending — the defender should play on to the very end, and against a careless opponent, zugzwang-proof shuffling plus one well-timed fork rescues the half point remarkably often.

Frequently Asked Questions

Is knight and pawn versus knight a draw?

Usually yes. A defending knight that blockades the pawn or controls one square on its path holds indefinitely, and it can always sacrifice itself for the pawn as a last resort, since knight versus knight is a trivial draw. The attacker wins mainly when the pawn reaches the seventh rank before the defending king is close, or with a rook pawn that traps the blockading knight at the edge.

Why is the rook pawn winning more often in knight endings?

At the edge of the board a knight loses half of its mobility. A knight blockading a rook pawn has only two or three escape squares, so the attacking king and knight can cover them all and simply capture it — the caging maneuver shown in this article. On central files the blockading knight always has spare squares and cannot be trapped.

Where should the defending king go?

Straight toward the pawn's path, as fast as possible. Any defending king on a square in front of the pawn makes the position a dead draw regardless of what happens to the knight. The knight alone can usually hold too, but the king's arrival is what makes the draw unconditional.

Can the defending knight just be chased away from the blockade?

Not easily on central files: the chasing king walks into checks, and a knight defended by its own king cannot be evicted by the enemy king at all. The attacker needs his own knight to join the eviction, and the tempo cost usually lets the defending king arrive. The exception is the board edge, where the defending knight's shortage of squares makes eviction and even outright trapping realistic.

When should the defender give the knight for the pawn?

The moment the pawn can no longer be stopped by other means, and not before. The trade produces a bare-kings draw, so the only mistake is timing: sacrificing while the blockade still held is unnecessary, and waiting one move too long lets the pawn queen. Count the pawn's remaining squares and the knight's routes to them every single move.

Does the fork trick really save lost-looking positions?

Yes — a knight one hop from the promotion square frequently wins the freshly promoted queen with a check-and-fork, converting a lost position into a draw on the spot. It is the single most common swindle in knight endings, and the attacker must check for it before every promotion.