Underpromotion is the act of promoting a pawn to a knight, rook, or bishop rather than to a queen — and doing it on purpose. The rules of promotion give you four choices when a pawn reaches the last rank, but the queen is so obviously the best of them that the other three are usually irrelevant. A player underpromotes only when the queen, for once, is the wrong answer: when a knight does something the queen geometrically cannot, or when the queen would win so completely that it draws the game by stalemate. These situations are vanishingly rare in practical play and slightly less rare in composed studies, which is exactly why they fascinate.

The queen is the default for a simple reason: it moves like a rook and a bishop combined, so it dominates every other promotion piece on open lines. Promote to a rook and you have given up the diagonals for nothing; promote to a bishop and you have given up the files and ranks for nothing. The queen is strictly stronger than either on the same square, and there is never a positional reason to prefer them. So the entire case for underpromotion rests on two narrow ideas: that the knight is not a subset of the queen, and that being slightly weaker can occasionally be an advantage.

Why the knight is the interesting one

Of the three underpromotion choices, only the knight is genuinely different from the queen. A rook and a bishop each cover a strict subset of the queen’s squares; whatever they can attack, the queen can attack too. The knight does not. It jumps. It reaches squares no queen, rook, or bishop standing on the same square could ever reach, and it delivers checks along geometries the queen has no access to.

That is the whole basis of knight underpromotion. There is essentially one circumstance in which promoting to a knight beats promoting to a queen: when the knight, on the promotion square, immediately does something the queen cannot — almost always a check or a fork. The classic motif runs like this. A pawn reaches the last rank; promoting to a queen does nothing decisive; but promoting to a knight gives check, and that check forks the enemy king and queen (or king and rook). The opponent must answer the check by moving the king, and the knight then captures the forked piece. No queen move from that square could have attacked both targets at once with tempo.

In notation this looks like e8=N+ or fxg1=N+ — the advance or capture to the last rank, the promotion to knight, and the check that justifies it. When you see =N on a score sheet, you are almost always looking at a fork or a check, because there is no other reason to choose a horse over a queen.

It is worth being precise about how uncommon this is. In serious over-the-board play, knight underpromotion appears a handful of times per decade across the entire grandmaster record; it is far more at home in composed problems, where the composer builds the exact geometry that makes the knight indispensable. The motif is real; the frequency is not.

The rook, and the stalemate trap

The second reason to underpromote has nothing to do with extra power and everything to do with having too much. Sometimes promoting to a queen wins so emphatically that the opponent is left with no legal move and is not in check — which is stalemate, and stalemate is a draw. Promoting to a rook instead leaves the opponent a legal move, keeps the position a clear win, and sidesteps the draw entirely.

The most celebrated example is the Saavedra position, a study from the late nineteenth century refined by the Spanish priest Fernando Saavedra. White has a king and a single c-pawn against a lone black king and rook. The pawn races to promote; Black harasses with a string of rook checks; White’s king walks down the board to escape them. At the critical moment White is about to queen, and a queen seems to win on the spot — but Black has a resource. With the queen on the board Black plays …Rc4+, forcing White to capture the rook, and with no other black piece to move the position is stalemate. A draw, snatched from a winning position.

The solution is to underpromote to a rook. With a rook the stalemate trick evaporates: the rook does not cover the squares the queen did, so capturing on c4 leaves Black a legal move, and Black is instead lost to a mating threat the rook delivers all the same. White wins. The Saavedra is, deservedly, one of the most reprinted studies in the literature precisely because it teaches the lesson in a handful of moves: a queen can be too strong, and a rook can win where she cannot.

Rook underpromotion to dodge stalemate does occasionally surface in real endgames, almost always in pawn races where the defending king is boxed into a corner. The rule of thumb is simple: before you reach for the queen, check whether the opponent has any legal reply. If the answer is no and they are not in check, promote to a rook.

The bishop, almost never

Bishop underpromotion is the rarest of the four by a wide margin. Like the rook, the bishop covers only a subset of the queen’s squares, so it offers no new geometry; unlike the rook, it is even weaker. The only conceivable reason to choose it is the same stalemate-avoidance logic that motivates the rook — and in nearly every such position a rook serves at least as well. A bishop promotion that genuinely beats both a queen and a rook requires a position so finely tuned that it is the near-exclusive property of composed studies, where a problemist has engineered the one configuration in which the bishop’s diagonal coverage is exactly what avoids stalemate while still winning. In actual games, bishop promotions are curiosities, sometimes played as flourishes when the win is trivial by any route. If you ever find a real reason to choose a bishop over both a queen and a rook, you have stumbled into a study come to life.

What it tells us about the game

Underpromotion is the exception that proves how good the queen is. Three of the four promotion choices exist only to handle what the queen mishandles — two because she is too sweeping, one because she cannot jump. The motif rewards the player who, at the moment of triumph, pauses to ask not “what is strongest?” but “what does this exact position need?” Usually the answer is the same, and you make a queen. Once in a long while it is not — and the player who notices promotes to a knight and forks the king, or promotes to a rook and avoids the stalemate, winning a game the obvious move would have drawn.

For the rule itself — the four legal choices, the notation, the edge cases — see Caissly’s reference page on pawn promotion.

References

Cross-links inside Caissly: the rule of pawn promotion sets out the four legal choices; stalemate is the draw that rook underpromotion is most often played to avoid.

Issue Nº 009 · The Magazine · The Caissly Editorial