PINAKA Learn Chess Intently Learn Chess Play Now

Variants

Chess 960: why it exists and how it works

Nine hundred and sixty starting positions, one set of rules, and a castling rule that nearly every explanation gets wrong. Here is the version that is actually accurate.

Chess 960 is chess with the back rank shuffled. The pawns start where they always start; the eight pieces behind them are rearranged, subject to two constraints that exist to keep the game playable rather than to make it strange. Everything else — pawn moves, captures, en passant, promotion, checkmate — is ordinary chess.

The name is not decoration. It is the exact number of legal starting arrangements, and the fact that the count is exact rather than approximate is a hint about how the variant was designed: this is not a loose house rule, it is a precisely specified system with a numbering scheme, which is why two players anywhere in the world can agree to play "position 518" and mean the same board.

The two constraints

Shuffling eight pieces onto eight squares gives 40,320 permutations. Only 960 of them are legal, because of two requirements:

  1. The bishops must be on opposite colours. One bishop starts on a light square and one on a dark square. Without this, some positions would give one side a pair of same-coloured bishops, which is a materially different game rather than a different arrangement of the same game.
  2. The king must start somewhere between the two rooks. This is what makes castling definable. If the king were outside both rooks, there would be no coherent way to say which rook is the "kingside" rook, and the castling rule would collapse.

Divide 40,320 by the constraints and you land on exactly 960. That is where the number comes from, and it is worth internalising because it explains the rules that follow: every awkward case in Chess 960 castling exists to preserve the king-between-rooks invariant.

How the positions are numbered

The standard numbering is Scharnagl's, and it produces every position exactly once with no gaps. It is worth knowing the shape of it, because it turns "position 518" from an arbitrary label into something you can reconstruct:

Sixteen times six times ten is 960, and the digits fall out in a fixed order. You do not need to do this arithmetic at the board; the point is that the numbering is a complete, deterministic enumeration rather than a list someone compiled. PINAKA uses the same scheme, so a game labelled SP #518 is the same position a tournament director would set up if they announced 518.

One practical note: the numbering is a convention, not a rule of chess. Two players can agree to play any legal arrangement and call it whatever they like. The numbering matters because it makes positions reproducible, which is what allows a 960 opening you found to be shared, revisited, and refuted.

The castling rule, properly

This is where most explanations fail, usually by describing the king's final square as though it were the same as orthodox chess and leaving it there. The accurate statement is about destinations, not distances.

After castling, the king and rook occupy the same squares they would occupy in ordinary chess. Kingside: the king ends on g1 and the rook on f1. Queenside: the king ends on c1 and the rook on d1. That is identical in every one of the 960 positions, which is the whole reason the rule is written this way — the outcome of castling is a single standard shape, whatever the starting arrangement.

What varies is the movement required to get there, and in Chess 960 it can be strange:

The conditions are otherwise familiar. Neither the king nor the castling rook may have moved earlier in the game. Every square between the king's starting square and its destination, and between the rook's starting square and its destination, must be empty apart from the king and that rook. The king may not be in check, may not pass through a square attacked by the opponent, and may not finish on an attacked square. The rook's path is not subject to an attack test — only the king's is, exactly as in orthodox chess.

The one-line version

Castling in Chess 960 is a move that ends with the king on c1 or g1 and the rook on d1 or f1, subject to the usual "nobody has moved, nothing in the way, king not in check and not passing through check" conditions. Everything else is detail about how they get there.

Why the variant exists

The argument for Chess 960 is not that ordinary chess is broken. It is narrower and more interesting than that: in ordinary chess, the opening is a solved collaboration. Two strong players can play twenty moves of theory neither of them has independently derived, and the first genuinely new position arrives with both clocks already half spent. The player with the better memory or the better database wins a contest that was not really about chess.

Shuffling the back rank removes that. In position 743 there is no theory, because the position has been played a thousandth as often as the Spanish, and the principles that survive are the ones that were always doing the real work: develop toward the centre, get the king safe, do not create weaknesses you cannot repair, find the pawn break that matches your piece placement.

This is why the variant is a better teacher than it looks. In orthodox chess you can play a good opening badly — follow the moves, arrive at a decent position, and never understand why it is decent. In Chess 960 you cannot hide, because the first ten moves require you to make judgements rather than recall them. Players who are strong in 960 are usually strong for reasons they can articulate.

What it is not good for

Two honest caveats. First, the variant does not eliminate preparation; it relocates it. Strong 960 players prepare types of position rather than lines — which pawn breaks follow from which bishop placements — and that is still preparation, just at a higher level of abstraction.

Second, the positions are not equally interesting. Some arrangements produce cramped, awkward games where both sides spend fifteen moves untangling their own pieces. The 960 are legal and reproducible, not curated. If you play enough of them you will find a few you dislike, and that is a property of the system rather than a flaw in it.

The first five moves, when there is no theory to lean on

In standard chess, most players have something prepared for the first eight moves whether they admit it or not. In 960 nothing is prepared, and the first few moves expose what you actually understand. The principles that hold up:

Find the pawn moves that open lines for pieces you already have. Not the pawn moves that would be thematic in a standard opening — the ones that connect your rooks, free your bishops, or contest the centre squares your opponent's pieces are aimed at. Because the back rank is scrambled, the pawn structure often has to do the work that development would normally do.

Count the squares between the rooks. The single most useful habit in the first minute of a 960 game is checking how many pieces sit between your king and each rook, because it tells you which side you can castle to and how soon. A position where one wing requires moving three pieces and the other requires moving one is not symmetric in practice, however symmetric it looks.

Do not develop toward the opponent's bishop. In some arrangements a bishop starts on a diagonal that points straight at a square your knight wants. Moving the knight there on move three because it is a natural square is a mistake that costs a tempo in every standard opening you know, and you will make it in 960 because the cue is unfamiliar.

Treat the centre as contested from move one. The usual "occupy, then support" rhythm depends on knowing which pawn moves are safe. Here you have to work it out, and the answer changes every game, so the useful habit is to ask which central pawn move is supported by a piece rather than which one is standard.

What transfers from standard chess, and what does not

Almost everything about the middlegame and the endgame transfers unchanged, which is why strong standard players are strong at 960 within a handful of games. What does not transfer is narrower and more specific than people expect:

Transfers and does not transfer
Skill In 960
Tactical patterns, mate nets, endgame technique Identical. The pieces move the same way and the geometry is unchanged.
Pawn-structure understanding, weak squares, outposts Identical. Structures differ from game to game but the judgements do not.
Calculation and candidate-move discipline Identical, and more heavily tested, because there is no memory shortcut.
Named opening lines Gone entirely. There is no Sicilian in 960; there are 959 other starting positions, and each has its own theory that nobody has written.
Typical development sequences Gone. The sequence has to be derived from the arrangement in front of you.
Castling intuition Actively misleading for the first few games. Destinations are fixed; the path is not.
Opening "feel" — the sense that a position is normal Gone, and this is the part that takes longest to replace.

The last row is the interesting one. Most of what a strong club player calls opening knowledge is not a set of lines; it is a feeling for which positions are comfortable. 960 removes that feeling completely and forces the slower, more explicit reasoning that principles are supposed to produce. That is uncomfortable, and it is the point.

Three mistakes almost everyone makes in their first ten games

Castling too late because the path is unfamiliar. The king sits in the middle for fifteen moves while you work out which rook can move where, and by then the centre has opened. Decide on move two which side you are castling to, and treat the pieces in the way as a to-do list.

Assuming a piece is developed when it is not. A bishop that starts on b1 in a 960 position may be on an open diagonal immediately, or may be entombed behind its own pawns. The file and rank tell you nothing about whether a piece is doing anything; only the lines do.

Playing the arrangement you got last time. Players who play a lot of 960 start recognising arrangements and importing plans from a previous game with the same numbering. This is exactly the memorisation the variant exists to remove, and it fails because the opponent's plan in that arrangement is not the same as the last opponent's.

How to start

Do not read about the variant; play it, and play the same position twice — once from each side. That is the single most useful 960 exercise, because it forces you to see the position's imbalances rather than your own plan. Then note which of the ordinary opening principles you actually applied, and which you only apply when a book tells you to.

If castling confuses you at first, that is normal and it passes. The trick is to stop thinking about the move as a sequence and start thinking about it as a destination: two squares, one shape, thirty-two possible starting arrangements of the two pieces involved. Once that reframing sticks, the awkward cases stop being awkward.

Next in the library

>