The Weirdest Chess Move

This post will attempt to answer the question: "What is the weirdest possible chess move?"

More specifically, we'll consider all possible chess move notations, using standard algebraic chess notation. So among possible written moves such as "Ne6", "Qxd7", etc., we want to decide which is the most unusual. I could imagine two different ways of making "weirdest" more precise.

1. Consider the list of all possible chess games (using standard rules), written down using standard chess notation. Which move notation occurs the fewest number of times in all possible games?
2. Look at all tournament chess games (with standard rules, etc.) that have occurred in history. Alternatively, let a powerful chess engine such as Stockfish play itself many times and write down all the games. Which move notation occurs with the lowest frequency?

Approach (1) is at least a well-defined question, because the number of chess games is finite, though incredibly large. The reason for this is that by the 50-move rule, any chess game in which there is no capture or pawn advance in 50 moves in automatically a draw; thus pieces cannot shuffle around the board forever and every game must end. One estimate for the number of possible chess games is known as the Shannon number, named after the mathematician Claude Shannon. The estimate is 10¹²⁰, which vastly exceeds even the number of atoms in the observable Universe! Therefore, it's impossible in practice to enumerate all chess games.

Approach (2) is a little better, but if we use only games that have actually occurred, or a computer sample size small enough to be computationally feasible, it might be too small for any of the "weirdest" moves to appear (we'll see this later). Moreover, we've made some choices about what a sample of representative "good" chess games might look like. I prefer approach (2), though, since the spirit of the challenge is to find what moves come up the least often in a game where both players play sensibly. In any case, my proposed "weirdest move" has not happened in any game that I could find.

The question I'm considering is a little unusual. Instead of looking at possible chess positions, the problem only asks about how a particular move is written down, so the only information we get about the position is from what the notation conveys. This includes the piece moving, the destination square, whether the move is a capture, whether it delivers a check or checkmate, etc.

For example, in the above position, black has just made the move "Nxf3+". The "N" indicates the knight has moved, "f3" is the destination square, "x" means the move was a capture, and the "+" means that the move delivers a check (but not checkmate). However, if we only see the symbols "Nxf3+" in isolation, we don't know what other pieces are on the board, where the knight came from, or where exactly the enemy king is (just that it's in range of the knight's attack).

There are some other notations for special moves in chess, such as castling: kingside castling is denoted "O-O", and queenside castling "O-O-O". The promotion of a pawn is denoted by an "=" sign, for example "e8=Q". In constrast, en passant does not have a special symbol in standard notation.

The above image shows a position after black plays "a5". This is an opportunity for white to capture en passant, meaning that white's b-pawn can move to a6 and capture the black pawn on a5 on the next move as if it had moved one square instead of two. However, this is notated simply "bxa6", so if we only know the notation and not the position, we can't distinguish en passant from an ordinary pawn capture.

Let's try a few weird moves. We'll start with castling. Castling on its own happens in almost every game, but it is usually a move to improve king safety in the opening, so it is almost unheard of to check or checkmate with a castling move. Hence, we have our first candidate weirdest move:

Candidate 1: O-O-O#

The notation here means: queenside castle and checkmate. It's an incredibly rare move, but it (nearly) happened in a high-profile historical game! In a 1912 game between international master Edward Lasker (distantly related to world chess champion Emanuel Lasker) and twice British chess champion George Alan Thomas, the following position was reached after black's 17th move, "Kg1".

Lasker had the opportunity to play checkmate in one move with "O-O-O#", our candidate weirdest move! He instead played "Kd2#", which is still checkmate since it allows the white rook on a1 to see the enemy king. Moving the king to deliver checkmate is already a bit uncommon, but the queenside castle would have been stranger still. You might be wondering how on earth the black king arrived on g1 in this game! It was in fact the culmination a remarkable series of forced moves in which the black king went from g8 to g1. The full game is well worth checking out (see here for an explanation of the full game).

Therefore, queenside castle checkmate is rare, but it did almost occur in a game involving two very high caliber players, so maybe we can do better. There's another uncommon move type that's natural to try: underpromotion.

When a pawn reaches the end of the board, it can become a queen, but the player also has the option to trade their pawn for a different piece instead (rook, knight, or bishop). This is called an underpromotion. Since the queen is the most powerful piece in chess, there are very few cases in which a player would "legitimately" underpromote, i.e. where the underpromotion is a superior move to promotion to a queen. Most occurrences of underpromotion in real games are "just for fun", for example when the promoted piece would be captured either way. However, there are some exceptions.

The above position occurred in a game between Vladlen Zurakhov and Alexander Koblents in a 1956 USSR semifinal event. Koblents has just played his knight to f5 and is threatening to win the pawn on g7, which would bring the game to a knight vs. two pawns position that is a theoretical draw. Therefore, white must promote the g pawn. However, if white plays "g8=Q", black will simply win the queen by forking the queen and king with the move "Ne7+". The only move that wins for white is "g8=N", an underpromotion to a knight. This defends against the fork, and the new knight can go on to help one of the two remaining pawns to queen promotion (this is still a difficult win, but Zurakhov did end up winning the game). Since knights can move and attack in ways that queens cannot, it makes sense that knight underpromotions are sometimes practical. Indeed, promoting to a knight is next most common after promoting to a queen. This leads us to our second candidate:

Candidate 2: Bishop underpromotion (e.g., "a8=B")

The rarest underpromotion of all is to a bishop. The bishop's value in chess is roughly equal to the knight's as the lowest of the (non-pawn) pieces, and unlike a knight, the bishop does not move in a way that the queen cannot. In particular, I didn't include a "#" (checkmate symbol) on the example move in Candidate 2, because if a bishop promotion move comes with checkmate, the same move with a queen promotion instead would also be mate. Therefore, no "legitimate" underpromotion occurs that way. I could also have made the move a check or capture, but this makes relatively little difference.

Though some endgame compositions have featured this type of move, there are once again precious few examples in real games. One of the only ones on record is in the following game between Aron Reshko and Oleg Kaminski in the 1972 Leningrad chess championship.

In this position, it's white to move. Though Reshko has a pawn close to promotion, Kaminski has a trick up his sleeve. First, if white does not address the threat on the a7 pawn, the black queen will simply win it in the next move, leading to equal material and a likely draw. If white moves to protect the pawn on its current square by taking the queen to a4 or b8, black can play the queen to f7 for checkmate!

Therefore, white must promote the pawn (or throw in a check with "Qg6+", but the queen must simply return and this gains nothing). However, promotion to a queen or a rook lead to an immediate draw because of black's defensive resource "Qf7+", sacrificing the queen! White's only move is to take the queen, and the resulting position is stalemate, a draw.

The above image shows the hypothetical variation 61. a8=Q Qf7+ 62. Qxf7. Though white is now up two full queens, black has no legal moves and the game is a draw by stalemate. A rook on a8 would also guard the h8 square, so it's again stalemate. Promoting to a knight in the initial position avoids this problem, but it's not too hard to see that the knight can't escape the corner without black's queen capturing it. The only winning strategy, which Reshko found in this position, is to promote to a bishop. Soon after, the players traded queens, and with an extra bishop, white went on to win.

While the bishop underpromotion is the rarest type of move in chess, it is not the weirdest move in terms of notation. I'll reveal my pick for the game's weirdest move in the next post.