The fortress: material that never gets to matter

A piece up, two pieces up, and still drawn. Some positions cannot be broken into, and recognising one is a skill worth more than most tactics.

The claim No way in, ever, not just no way in yet
Watch out Engines score fortresses badly

A fortress is a position the stronger side cannot break into, whatever the material count says. The defender arranges king and pieces so that every route in is covered, and then simply waits. There is no counterplay to find and no accuracy to sustain: once the wall is built, shuffling a king between two squares holds the draw forever.

It is worth learning from both ends. If you are losing, a fortress is often the only thing left, and it is frequently worth giving up more material to reach one. If you are winning, a fortress is the trap you walk into by trading carelessly, so before every simplification ask what shape the position will have afterwards.

The simplest fortress there is: the wrong bishop

White is a bishop and a pawn up. The dots are every square the bishop can reach: eleven of them, all light. The pawn promotes on h8, which is dark, so the bishop can never attack the corner or the king sitting in it. Black shuffles between h8 and g8 and that is the whole defence.

The natural winning attempt: bring the king in and take g8 away.

1.Kf7

Stalemate. Black is not in check and has no legal move, so the game is drawn on the spot. Pushing the pawn with 1.h7 instead is also stalemate, and everything else leaves the king shuffling in the corner forever.

A fortress is a judgement, not a rule

There is no list to memorise, and that is the honest part. Calling a position a fortress is a claim you have to prove to yourself: name every square the attacker could come in on, show that something covers each of them, and check every pawn break and every sacrifice that could open a line. Miss one entry square and it is not a fortress, it is a bad position you have decided to stop working on.

The shapes do repeat, though. A rook pawn with a bishop of the wrong colour, as above. A wall of pawns standing on squares the enemy bishop cannot touch. A king and a minor piece holding a corner that a queen cannot force. A rook that will happily give itself up for the last pawn the moment it advances. If a position looks like one of those, count again before you trade.

One practical warning. Engines are unreliable here, and always have been. A fortress is defined by what can never happen, and a search that looks twenty or thirty moves ahead cannot see never, so you will be shown a comfortable advantage in a position that is a dead draw. If the evaluation never moves while the moves shuffle, that is the tell. It is one of the few places where your own eyes beat the machine, and it is related to the other cases where material stops meaning anything: see insufficient material and perpetual check.

Questions people ask

What counts as a fortress in chess?

A position where the side with more material has no way to make progress at all, because every entry square is covered and no pawn break or sacrifice opens a line. The defender does not have to keep finding good moves, which is what separates a fortress from a difficult defence: once it is set up, waiting moves hold it forever and the fifty-move rule eventually ends the game.

Why do chess engines misjudge fortresses?

Because an engine evaluates by searching ahead, and a fortress is a statement about what will still be true in a hundred moves. Nothing changes inside the search, so the engine keeps scoring the extra piece it can see and cannot see that the piece is useless. Endgame tablebases give the true result once few enough pieces are left, and they frequently disagree with the evaluation.

Fortress sits under Endgame words in the reference. Nearby: