Chess960 (Fischer Random): how to play well when there is no opening theory

Short answer

In Chess960 the back rank is drawn from 960 possible arrangements and opening theory ceases to exist. This guide covers what changes and what does not, the three-question reading you make before the first move, the data that corrects the myth of the weak f7 square, how the castling rules work, and where development starts when the pieces are born on strange squares.

The essentials in a few lines

• The 960 positions come from a simple count: bishops on opposite colours, the queen on one of the six remaining squares, two knights among the five that are left — and the king necessarily between the two rooks.
• The weak square is not always f7. In the drawn starting position, the most frequently undefended pawns are a2 and h2, in 228 of the 960 positions each; f2 is undefended in only 36.
• Castling has a fixed destination in every draw: king to g1 and rook to f1 on the king's side, king to c1 and rook to d1 on the queen's side — no matter where the two pieces start.
• A bishop in the corner is the piece that demands a plan: in 420 of the 960 positions at least one bishop starts on a1 or h1, and it does not exist until the pawn in front of it moves.
• The order that works in any draw is always the same: find the undefended pawn, activate the worst-placed piece, and get the king safe before opening the centre.

What changes and what does not

The count is easy to redo: four light squares for one bishop, four dark for the other, six remaining squares for the queen and ten ways to place the two knights among the five that are left. Four times four times six times ten is 960 — and the king and rooks multiply nothing, because the three squares left over already determine the only legal arrangement: king in the middle, rooks at the ends.

What changes is memory alone. There is no main line, no home-prepared novelty and no repertoire: moves 1 to 15, which in standard chess come out of a book, have to be solved at the board by the person playing them. That is why Fischer proposed the variant, and why it came back as an official tournament format.

What does not change is almost everything else: piece values, endgame technique, tactical motifs, the importance of the centre, king safety. A rook and pawn against rook endgame is the same endgame. People who play Chess960 well are, almost always, people who play chess well — with the difference that the advantage of having studied openings disappears, and the advantage of understanding positions remains.

The reading you do before move one

The diagram position is one of the 960 and serves as a good example. The king starts on d1, between the rook on c1 and the rook on h1; the bishops are on a1 and f1, the knights on e1 and g1, and the queen on b1. None of that is good or bad in itself — it is the list of facts the game starts from, and it has to be read rather than felt.

The first question has an objective answer here: f2 is the only pawn no white piece defends. The king on d1 is too far away, the knight on e1 covers c2, d3, f3 and g2 but not f2, and the bishop on f1 defends e2 and g2. That is the square where the game can end early if Black gets two pieces pointing at it before White solves the problem.

The third question is the one that saves the most time later. With the king on d1 and rooks on c1 and h1, castling on the king's side is a long trip — the king crosses three squares — and requires clearing e1, f1 and g1 first. Knowing that in the first minute changes the whole order of development: the pieces on that side come out first because the king has to pass through.

The undefended pawn — and the f7 myth

The reflex of looking at f7 comes from standard chess, where the king on e1 defends f2 and the f7 pawn is attacked by almost every developed bishop. In Chess960 that reflex gets in the way: in nearly 80% of positions the king starts somewhere other than e1, and the fragile pawn moves with it.

The frequency of a2 and h2 has a structural explanation. The rooks start in the corner in only 108 of the 960 arrangements; in the rest, the a- and h-pawns come to depend on some other piece, and often there is none. That is why edge attacks, rare in standard chess, appear early in Chess960 with a naturalness that unsettles newcomers.

The practical consequence is a counting rule, not a memory one: before move one, run through all eight pawns and count how many pieces defend each. It takes under a minute, and the answer usually decides the first move of the game — which is nearly always the move that fixes the undefended square, not the one that occupies the centre.

Development: the order that works in any draw

The line in the diagram is the one the engine picks in the position from the previous section, and it shows the order clearly. White opens with 1.c4 and 2.e3, two pawn moves that simultaneously free the rook on c1, the bishop on f1 and the queen's diagonal — no piece has moved yet, and three are already breathing.

The third move is the instructive one: 3.Nh3. A knight on the rim looks like a mistake, and in standard chess it usually is. Here it solves the problem the opening reading raised — from h3 the knight defends exactly f2, the undefended pawn — and it also supports the f2-f4 advance that follows. It is the sort of move no book teaches and the reading of the position hands to you.

After 4.Be2 Nh6 5.f4 f5 the position has a shape: White has taken central space, fixed the weak square and keeps both rooks ready for whichever files open. It is not an opening with a name, and it does not need to be — it is development with a reason behind every move, which is all Chess960 asks for.

Corner bishops and the castling rules

Two consequences confuse first-timers. If the king already starts on g1, castling that side barely moves anything: the king stays put and only the rook travels. And since the king may cross four or five squares, castling sometimes requires emptying half the rank first — which makes it a multi-move plan rather than a decision taken on the spot.

The second diagram covers the other subject of this section: bishops born in the corners. Here both are, on a1 and h1, and neither exists until the pawns on b2 and g2 move. This is not a rare case — in 420 of the 960 positions at least one bishop starts in a corner, and in 60 of them both do.

The line 1.b4 b5 2.e4 Nd6 3.d3 f5 4.g3 handles it directly: the pawn moves b4 and g3 do not fight for the centre, they open both long diagonals and turn two dead pieces into two long-range ones. Anyone who tries to play this position centre-first discovers on move ten that he is playing two pieces short.

How to train Chess960

The basic exercise takes three minutes per position. Draw an arrangement and write down which pawn is undefended, which piece is worst placed, which way the king is going and which two pawn moves you would play first. Then compare with what an engine recommends over the first few moves. Ten repetitions teach more than thirty games played on impulse.

In a game, the clock is this variant's specific enemy: it is easy to spend ten minutes on move one and reach the middlegame with nothing left. The practical rule that works is to set a ceiling — two or three minutes for the opening reading — and accept that the first two moves will be reasonable rather than optimal.

Finally, review your games by phase. With no memorised opening, the typical Chess960 error is not forgetting a line: it is developing in the wrong order, or opening the centre with the king still in the middle. Both show up in the aggregate of your games with a clarity that no single game provides.

Frequently asked questions