The Maróczy Bind: Karpov v Kavalek 1974, Checked by Stockfish

14.Na4+0.22about equal

Karpov played 14.Na4, offering the queen trade with the knight on its way back to c3. After 14...Qxd2+ 15.Kxd2 White's king guards c3 and e3, the pin is gone, and the knight is free to head for d5.

Stockfish's first choice too. Depth 26.

What it revealed

Karpov chose the endgame on purpose. In the bind, Black's activity rests on the queen; without it, the cramped side has fewer pieces to break with, and White's king becomes a central defender of c4 and e4.

• ✓ Verified Stockfish rates 14.Na4 its first choice, and keeps White's edge at about a quarter of a pawn after the queen trade; the other moves are all close to level.
• ◐ Inferred The king on d2 is the point: it defends c3 and e3, so White's rooks are free for the c-file and, later, the h-file.

17.Nd5+0.18about equal

Karpov played 17.Nd5, the bind's signature move. The knight is safe from pawns on d5, it hits e7 and b6, and it unpins the c-file. Black can only remove it by giving the bishop on e6 for it.

Stockfish preferred 17.g4 (+0.37); the game move gave up about 0.2 pawns. Depth 29.

What it revealed

Karpov chose the move that changes the structure on White's terms. If Black takes on d5, a white pawn arrives there and the space grows; if not, the knight stays as the best minor piece on the board.

• ✓ Verified Stockfish prefers 17.Be3 or 17.g4 by a small margin. All three moves keep a small edge for White, between a tenth and a half of a pawn; the choice is one of style, not of soundness.
• ◐ Inferred Black's 17...Kf8 guards e7 so that the knight has nothing to hit; two moves later Kavalek gave the bishop for it anyway.

25...b5+0.61White slightly better

Kavalek played 25...b5, the classic break against the bind. After 26.g4 bxc4+ 27.Rxc4 Rxc4 28.bxc4 the queenside is open and White has a weak pawn on c4, the first concrete target Black has had all game.

Stockfish's first choice too. Depth 29.

What it revealed

The break was not only right but close to forced: every quiet move lost material. It shows the correct defence against a bind, which is to strike at the pawn base while the opponent's pieces are busy on the other wing, before the squeeze becomes a zugzwang.

• ✓ Verified Stockfish rates 25...b5 its first choice, and every other move it was offered loses material. After Karpov's 26.g4 (Stockfish preferred 26.h6) its estimate of White's edge falls to about a fifth of a pawn: close to level.
• ✓ Verified Karolyi, who wrote two volumes on Karpov's strategic wins, counted only one small inaccuracy by Karpov in the whole game, 31.Kc3. Stockfish adds 26.g4 to that list.^[3][4]

28...Nc5++2.67White winning

Kavalek played 28...Nc5+, and after 29.Bxc5 Rxc5 the minor pieces were gone except for opposite-coloured bishops: White's light-squared one, Black's dark-squared one. Endings with bishops of opposite colours are famous for being drawn.

Stockfish preferred 28...gxh5 (+0.04); the game move gave up about 2.6 pawns. Depth 28.

What it revealed

Black traded the best piece for White's worst one and let the h-pawn live. After 30.h6 the bishop on f8 has no moves that matter: it cannot reach the long diagonal, and the pawns on f4 and h6 cover its squares.

What Stockfish adds This is the biggest swing of the game by Stockfish's count, bigger than the famous 37...Bg7 in the line it led to. The rule that opposite-coloured bishops draw needs one condition: both bishops must be able to move. After 30.h6 Bf8 Black's bishop has two squares, both covered, and the ending is close to rook and bishop against rook.

• ✓ Verified Stockfish rates 28...Nc5+ almost three pawns worse than 28...gxh5. After the game move White's edge is well over two pawns; after taking on h5 it is zero.
• ◐ Inferred The knight on e4 was Black's only active minor piece; the bishop on e3 was White's least useful one. Trading them changed which pieces were left to play the ending.

37...Bg7+5.91White completely winning

Kavalek played 37...Bg7 to cover h8. White simply took on d5 with 38.Bxd5, kept the c4 pawn and the h7 pawn, and the rest was technique.

Stockfish preferred 37...Rxc4+ (+1.97); the game move gave up about 3.9 pawns. Depth 21.

What it revealed

Under pressure, Black went for the most direct defence of the promotion square and gave White the time to clean up. The check on c4 would have won a pawn first, because the bishop could still come to g7 on the next move.

What Stockfish adds Stockfish agrees that 37...Rxc4+ was much better, but it does not call it a draw: after the check it still rates White about two pawns better. Black's defensive chances were real mainly because the position was hard for White to play with the clock running.

• ✓ Verified Kavalek called 37...Bg7 a time-pressure blunder and gave 37...Rxc4+ as the move that kept good drawing chances.^[3][4]
• ✓ Verified Dvoretsky called 37...Bg7 the decisive mistake and recommended 37...Rxc4+ too.^[5]
• ? Speculative The time control was 40 moves in two and a half hours, so Black likely had only minutes for this decision: perhaps why the simplest-looking defence was chosen.

31.Kc3+0.24about equal

Karpov played 31.Kc3, keeping the a-pawn. After 31...fxg4 32.Bxg4 Black's kingside pawns were loosened and the bishop could later come out via g7.

Stockfish preferred 31.g5 (+3.10); the game move gave up about 2.9 pawns. Depth 28.

What it revealed

Karpov chose safety over a pawn sacrifice whose point was hard to prove at the board. It gave Black real defensive chances again, but only for a few moves.

What Stockfish adds Stockfish agrees with Karolyi and Dvoretsky: 31.g5 keeps about three pawns' advantage, while 31.Kc3 lets it fall to about a third of a pawn. It was the right debate, but it began only after Black's biggest error, 28...Nc5+, which the notes pass over, and it shows that Karpov, too, let the win slip for a few moves.

• ✓ Verified Dvoretsky preferred 31.g5, admitted that the reason was hard to prove at the board, and then proved it at length: the pawn on a2 matters less than the dead bishop.^[5]