Improve your worst piece · For club players · 19 min read
Improve Your Worst Piece: Karpov's 24.Nb1 v Spassky, 1974
Leningrad 1974: Karpov retreats a knight to its first rank and wins in eleven more moves. Stockfish 19 agrees with the famous move, and shows Spassky lost the same race on the other side.
- threat
- plan
- target
- weakness
- Black's idle pieces
In short
- When a position has no concrete plan, find your least useful piece and give it a job: every other piece gains from it.
- In game 9 of the 1974 Candidates semi-final, Anatoly Karpov's knight on c3 was shut in by pawns on a4, e4 and c6, so Karpov sent it back to b1 and round by d2 to f3.
- Stockfish 19 picks 24.Nb1 as the best move in the position, so the famous retreat is not only style: it is the strongest plan White had.
- The game was decided by the same rule on the other side: Spassky's worst pieces, the knight on a6 and the bishop on d8, never found a job, and the engine's defence 28...Nb8 was a retreat of exactly Karpov's kind.
- To use it: name your worst piece, name the square where it would work, find the route, and make sure the opponent has nothing better to do while it travels.
How we know what we say
- Verified From the record: Stockfish's numbers, the players' own notes or interviews, or a cited source.
- Inferred Read from the moves and the position: a reasoned reconstruction, not the player's words.
- Speculative A guess about mood, intent or practical motives: plausible, not proven.
Part 1The setting: Leningrad 1974, a former champion against the rising star
In the spring of 1974 Anatoly Karpov, then 22, played Boris Spassky in the semi-final of the Candidates matches in Leningrad. Spassky had been world champion until losing it to Bobby Fischer in 1972. The winner of the match would go on to the Candidates final, and the winner of that would challenge Fischer.
Spassky won the first game, but Karpov won games 3 and 6 and came to game 9 a point ahead. The ninth game was Karpov's third win, and Mark Weeks's history of the cycle calls it Karpov's best game of the match. Karpov won game 11 as well and took the match with four wins, one loss and six draws. [1][5] Verified
| Game | White | Winner | Match score (Karpov first) |
|---|---|---|---|
| 1 | Karpov | Spassky | 0–1 |
| 3 | Karpov | Karpov | 1.5–1.5 |
| 6 | Spassky | Karpov | 3.5–2.5 |
| 9 | Karpov | Karpov | 5.5–3.5 |
| 11 | Karpov | Karpov | 7–4 |
The game has no sacrifice and no attack on the king. It is remembered for one quiet retreat, 24.Nb1, which Sam Collins's book on Karpov and Raymond Keene's column both call one of the most memorable moves in chess history. The win came from making one bad piece good while the opponent's bad pieces stayed bad. [3][4] Verified
Part 2What does improving your worst piece mean?
Improving your worst piece means this: when there is no tactic and no obvious plan, find the piece that does the least for your position, and move it to a square where it has a job. A position is only as strong as its weakest-placed piece, so lifting that one piece often lifts everything. Coaches often credit the idea to the trainer Mark Dvoretsky, who called it the principle of the worst piece.
Three questions make it practical. Which piece is worst (the one controlling the fewest useful squares and doing the fewest jobs)? Where would it work (an outpost, an open file, a diagonal aimed at a weakness)? And how does it get there? The route is often backwards first.
- weakness
- outpost
- Covered by c6
- threat
- target
- Rook on the 2nd rank
The same bad knight in two positions. On the left Black has nothing to do, so a slow trip costs nothing and Stockfish rates White a little better. On the right Black's rook is already active, the b2 pawn comes first, and Stockfish rates White more than two pawns worse.
Part 3How the Scheveningen left White's knight stuck on c3
Spassky chose the Scheveningen Sicilian, with pawns on d6 and e6, and Karpov answered with the classical 6.Be2 and 8.f4. Black's 10...a5 and 11...Nb4 put a knight on b4 that White's a-pawn, already on a4, can no longer chase, at the cost of the b5 square. After 12...Bc6 and 13.Nd4, the bishops came off on c6 and Black's pawns were left as c6, e5 and a5.
10.Nb3 a5 11.a4 Nb4 12.Bf3 Bc6 13.Nd4 g6 14.Rf2 e5 15.Nxc6 bxc6 16.fxe5 dxe5 +0.75White slightly better
From the Scheveningen to the game's structure: Black's knight settles on b4, the light-squared bishops come off, and the f- and d-files open. Inferred
- outpost
- weakness
- target
- c6 covers these
- pawn chain or line of force
Black's c6 and e5 pawns are weak, but they also take the best squares from White's knight. Karpov's first job was to stop Black improving: 17.Qf1, 18.h3 and 19.Bg4 were all moves against Black's ideas (...Ng4, ...Bc5), as both Collins and Kapnisis point out. [2][3] Verified
Part 4Two plans from the same position
After 19.Bg4 h5 20.Bxd7 Qxd7, each side has one piece doing well and one doing badly. White has an open d-file, a target on c6 and a queen that wants c4, but a knight with no future on c3. Black has the b4 knight, but a dark-squared bishop on e7 that looks only at its own e5 pawn.
- plan
- target
- weakness
- The open d-file
- idea not played
- threat
- weakness
- target
- outpost
- pawn chain or line of force
The same position, two plans. Both sides know which of their pieces is worst. White's knight has a long road; Black's bishop has a short one, but it leads to h4, where there is nothing to attack.
Black's counterplay depends on the knight on b4: while it stands there, c2 needs a guard, and every white piece guarding c2 is not attacking c6 or f7. So White's plan has two halves: bring the knight back into play, and drive Black's best piece away with c2-c3. Inferred
Part 5The critical moments
Three decisions built White's win: in the first Black chooses how to fix its own worst piece, in the next two White fixes its worst piece and spoils Black's best. Try each position before opening the answer.
Critical moment 1 of 4 · Move 21, Black to play
Black's worst piece: activate it, or trade queens?
- threat
- weakness
- target
White's queen has just come to c4, hitting c6 and f7. Black's bishop on e7 is hemmed in by its own e5 pawn. Do you activate the bishop, defend, or offer a queen trade?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 21...Qe6 22.Qxe6 fxe6 | +0.70White slightly better | Stockfish's choice. Black trades the queens and accepts doubled e-pawns; the c6 and f7 weaknesses matter much less without White's queen, and Black is only somewhat worse. Verified |
| 21...Rad8 | +1.81White clearly better | Collins calls it more logical than the game move, and it contests the d-file. But Stockfish rates it even a little worse than the game, at nearly two pawns for White. Verified |
| 21...Qd6 | +2.89White winning | Covers c6 and keeps the queens on, but the queen stands on the open d-file in front of White's rooks. Stockfish rates it more than two and a half pawns for White. Verified |
What Spassky played
21...Bh4+1.43White clearly better
Spassky played 21...Bh4, giving the bad bishop a diagonal outside the pawn chain. The bishop eyes g3 and f2, and Black kept the queens on rather than accept a long, passive ending.
Stockfish preferred 21...Qe6 (+0.67); the game move gave up about 0.8 pawns. Depth 25.
What it revealed
Spassky also applied the worst-piece rule, but chose a square with nothing to attack. On h4 the bishop can be pushed away by g2-g3, and it ends the game back on d8, where it started its trip.
- Verified Collins calls 21...Bh4 an odd move and says the queen trade 21...Qe6 22.Qxe6 fxe6 might have been a better try than risking a middlegame against Karpov, even though the endgame would be a long defence.[3]
- Speculative Spassky may have preferred to keep queens on because a long defence of doubled pawns against Karpov, a point down in the match, looked like playing for a draw at best.
- target
- weakness
- idea not played
- weakness
- target
Two ways to deal with Black's worst piece. The trade takes White's best piece off with it; the bishop trip leaves White's queen dominant.
Critical moment 2 of 4 · Move 24, White to play
The knight that has no square
- weakness
- outpost
- target
- Taken by c6
White's pieces are well placed except one: the knight on c3 is blocked by the c6 pawn and White's own pawns on a4 and e4. Black's knight on b4 still pins down c2. What would you do?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 24.Bc5 24...Qg5 25.Rxd8+ Rxd8 | +0.61White slightly better | Attacks the queen and eyes the b4 knight, but after the trades on d8 Black's pieces get active. Stockfish rates it under a pawn for White, well below the game move. Verified |
| 24.Kh2 | +0.09about equal | Natural, but the rook on d2 is guarded only by the bishop, so 24...Rxd2 25.Bxd2 Rd8 takes the d-file. Stockfish rates it close to level: the advantage disappears. Verified |
| 24.Re2 | +0.76White slightly better | Clears d2 for a knight but gives up the d-file for now, and the knight would still need b1 first. Stockfish rates it under a pawn for White. Verified |
What Karpov played
24.Nb1+2.16White winning
Karpov played 24.Nb1. The knight steps back to the first rank, guards the rook on d2 a second time, and heads for d2 and f3, while c2-c3 is now possible because the knight is no longer in the c-pawn's way.
Stockfish's first choice too. Depth 21.
What it revealed
One move did three jobs: it started the knight's road to the kingside, freed the c-pawn to chase Black's best piece, and guarded the d2 rook. The backward move was the most active one on the board.
- Verified Collins and Keene both describe the idea the same way: re-route the knight from c3, where the c6 pawn and White's own a4 and e4 pawns limit it, to the kingside, while protecting the rook and preparing c2-c3 to drive away Black's only good piece.[3][4]
- Verified According to Jan Timman, as quoted by Collins, Karpov's trainer Semyon Furman predicted 24.Nb1 in the press room.[3]
- Verified Stockfish 19 picks 24.Nb1 as its first choice at about two pawns for White, and every alternative we checked (24.Bc5, 24.Kh2, 24.Re2) scores under a pawn. The quiet retreat was close to the only way to keep the advantage.
Critical moment 3 of 4 · Move 26, White to play
Chase the good knight now, or first finish the trip?
- plan
- outpost
- key square
- weakness
White's knight is on b1 and the king has gone to h2. Black's knight on b4 is still the best black piece. Should White kick it away at once, or first clear a road for the knight?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 26.Re2 | +2.75White winning | Kapnisis gives this move: the rook leaves d2 free for the knight first, and the c-pawn push can wait. It is Stockfish's first choice in its deepest search, about half a pawn more than the game. Verified |
| 26.Rxd8 26...Rxd8 | +2.86White winning | Trades a pair of rooks and lets Black take the d-file for a moment. Stockfish rates it slightly above the game move, close to three pawns for White. Verified |
| 26.g3 26...Be7 | +2.70White winning | Drives the bishop back at once, at the cost of loosening the king's cover; the bishop simply returns to e7. Stockfish rates it about the same as the game move. Verified |
What Karpov played
26.c3+2.43White winning
Karpov played 26.c3. The knight on b4 has only one square where it is not simply lost, a6, and from there it guards nothing in the centre. Stockfish's best reply was 26...Rxd2, which Black did not play.
Stockfish preferred 26.Re2 (+3.01); the game move gave up about 0.6 pawns. Depth 21.
What it revealed
Karpov chose to make Black's best piece bad before making White's worst piece good. The order cost a little by Stockfish's count, but the plan stayed the same: one side's pieces got better while the other side's got worse.
- Verified Collins suggests Black should have answered 26.c3 with 26...Rxd2, trading a pair of rooks even though it brings the white knight to d2, where it wants to go. Stockfish agrees that 26...Rxd2 was Black's best reply.[3]
- Verified Stockfish rates 26.c3 about half a pawn below 26.Re2, but White stays more than two pawns better either way: the order mattered less than the plan.
Part 6The turning point: Black's worst piece stays worst
After 26...Na6 27.Re2 Rf8 28.Nd2 the knight is one step from f3, and now the question is Black's: the knight on a6 does nothing for the centre, and the bishop on h4 is about to be chased.
Critical moment 4 of 4 · Move 28, Black to play
Two bad pieces: which one to fix?
- threat
- weakness
- target
White's knight will land on f3 next move, hitting the h4 bishop and the e5 pawn. Black's knight on a6 does nothing. Which black piece should move, and where to?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 28...Nb8 29.Nf3 Bf6 | +3.68White winning | Stockfish's choice: the knight goes back to b8 to come to d7, where it guards e5 and c5, and the bishop goes to f6. White is still winning by the engine's count, at three to four pawns, but Black's pieces get work. Verified |
| 28...Nc7 29.Bc5 | +3.85White winning | Kapnisis warns that 29.Bc5 hits the rook on f8 and wins material. Stockfish scores it about four pawns for White, close to the knight retreat but with the knight still far from d7. Verified |
| 28...Qxb2 29.Nf3 | +8.40White completely winning | Grabs a pawn, but Collins shows it drops a piece after 29.Nf3, with the h4 bishop attacked and the queen far from home. Stockfish rates it more than eight pawns for White. Verified |
What Spassky played
28...Bd8+5.22White completely winning
Spassky played 28...Bd8, bringing the bishop back to guard e7 and f6 before the knight arrived. The a6 knight stayed where it was.
Stockfish preferred 28...Nb8 (+3.49); the game move gave up about 1.7 pawns. Depth 22.
What it revealed
Black saved the wrong piece. The bishop on d8 sits behind its own pawns and blocks the rooks, the knight on a6 still has no job, and White's knight reaches f3 with nothing to stop it.
What Stockfish adds Stockfish rates 28...Bd8 close to two pawns below its choice, 28...Nb8, which re-routes Black's worst piece to d7. After 28...Bd8 29.Nf3 the engine already gives White more than four pawns' worth of advantage.
- Verified Stockfish 19 rates 28...Bd8 nearly two pawns below 28...Nb8, the biggest single loss of the game.
- Inferred 28...Nb8 is the same manoeuvre as 24.Nb1: a knight retreating to its first rank to come back to a better square. Black needed Karpov's idea, a few moves later.
- threat
- outpost
- target
- weakness
- White's grip
30.Rd2 Be7 31.Qe6 Rad8 32.Rxd8 Bxd8 33.Rd1 Nb8 34.Bc5 Rh8 35.Rxd8 +7.42White completely winning
The finish. The e2 rook, now White's least active piece, goes to d2; the rooks are traded; the bishop comes to c5. Black resigned after 35.Rxd8 in view of 35...Rxd8 36.Be7, when the rook and the f6 pawn are both hit. Verified
Part 7What Spassky could have done
Black's real chances came early. At move 21, 21...Qe6 traded White's best piece and kept the disadvantage under a pawn by Stockfish's count. At move 26, 26...Rxd2 27.Nxd2 traded a pair of rooks, which helps the side with less space. By move 28 the engine already rates White winning, but the best resistance was the worst-piece rule applied by the defender: 28...Nb8, the knight going back to b8 and then to d7, where it guards e5 and covers c5 and f6.
- plan
- outpost
- key square
- pawn chain or line of force
None of these saves the game by force, but each follows the article's idea: the queen trade removes White's best piece, the rook trade leaves White fewer pieces to improve, and the knight retreat fixes Black's worst one. Spassky's 21...Bh4 and 28...Bd8 moved a bad piece without giving it a job. Inferred
28...Nb8 29.Nf3 Bf6 30.Ref2 Nd7 31.Bg5 c5 32.Bxf6+ Nxf6 +3.11White winning
Stockfish's defence after 28.Nd2: the knight reaches d7, the dark-squared bishops come off, and the knight recaptures on f6. White is still about three pawns better, but Black's pieces finally all do something. Verified
Part 8A different angle: the race between two worst pieces
The usual account of this game is about one move: 24.Nb1, Karpov's knight going backwards. Stockfish 19 goes further than confirming it: every other move we checked lets the advantage shrink below a pawn, because only the knight on b1 guards the d2 rook against ...Rxd2 and ...Rd8. What the engine adds is the other half of the story: the game was a race between two worst pieces, and Black never really started it.
- plan
- idea not played
- target
- Targets from f3
White's knight needed three knight moves and two clearing moves (c3 and Re2) to reach f3. Black's knight, pushed to a6 by 26.c3, needed only ...Nb8 and ...Nd7 to come back, but Spassky spent those moves on ...Kg7, ...Rf8 and ...Bd8, and the knight did nothing until 33...Nb8, when the game was gone. Inferred
The engine also settles a detail the annotators see differently. Kapnisis points out that 30.Ng5 was a tactical win that Karpov most likely missed, while Collins and Keene praise 30.Rd2. Stockfish rates both as winning by more than five pawns, and the game move is its first choice: the quiet worst-piece move did not cost anything against the tactic. [2][3][4] Verified
30.Ng5 fxg5 31.Rxf8 Kxf8 32.Qe6 +7.04White completely winning
Kapnisis's 30.Ng5: the knight is given up to open the f-file, and after 31.Rxf8 Kxf8 32.Qe6 the queen invades on the light squares. Stockfish confirms that White wins this way as well, by about seven pawns. Verified
| Move | What happened | Stockfish |
|---|---|---|
| 21...Bh4 | Black keeps queens on instead of 21...Qe6 | about three-quarters of a pawn lost |
| 24.Nb1 | The knight starts its trip: Stockfish's first choice; the alternatives score under a pawn | about two pawns for White |
| 26.c3 | Black's best piece goes to a6; 26.Re2 was slightly stronger | about half a pawn below best |
| 28...Bd8 | Black saves the bishop, not the knight; 28...Nb8 was the defence | nearly two pawns lost |
| 29...f6 | Black weakens e6 for good | more than a pawn lost |
Part 9How to improve your worst piece in your own games
The worst-piece routine
- Check for threats firstIf either side has a tactic or a direct threat, deal with that. The rule is for quiet positions.
- Name the worst pieceAsk of each piece: how many useful squares does it control, and what does it attack or defend? The piece with the fewest answers is the worst; pieces blocked by their own pawns are the usual suspects.
- Name its best squareAn outpost no pawn can attack, an open file, a diagonal aimed at a weakness, a square next to a target.
- Find the route, even backwardsTrace the path square by square. If the direct path is blocked, look behind: a knight on c3 may need b1 and d2.
- Check the cost of the tripCount the moves and ask what your opponent does with them. If the answer is nothing dangerous, go.
- Spoil the opponent's best pieceOn the way, look for a pawn move or exchange that makes the opponent's best piece worse, like Karpov's c2-c3 against the b4 knight.
- Is the position quiet: no threats, no hanging pieces?
- Which of my pieces does the least?
- Where would it work, and how many moves does it take?
- What does my opponent do in those moves?
- Can I make the opponent's best piece worse on the way?
The same thinking runs through other Karpov games on this site: see prophylaxis against Unzicker and the knight against the bad bishop. For the modern version of taking squares from the enemy pieces, read restriction in Ding v Gukesh.
Your turnPractise it
Set each position up, decide your plan before you look, then play it out against Stockfish to see whether it holds.
White to play (after 23...Rfd8). Every white piece works except the knight on c3. Find the move that gives it a future and keeps the rook on d2 safe.
Hint
The knight cannot go forward: b5 and d5 are covered by the c6 pawn. Which square behind it guards d2?
A good answer
24.Nb1, Karpov's move and Stockfish's first choice, at about two pawns for White. The knight guards the d2 rook (so ...Rxd2 is met by Nxd2), frees the c-pawn for c2-c3 and heads for d2 and f3. Moves such as 24.Kh2 or 24.Re2 score under a pawn.
Black to play (after 28.Nd2). The white knight lands on f3 next. Which black piece is worst, and where should it go?
Hint
The knight on a6 guards nothing important. Which square would let it guard e5?
A good answer
28...Nb8, heading for d7, is Stockfish's choice. With ...Bf6 to follow, the knight on d7 guards e5 and c5. The game's 28...Bd8 lost nearly two pawns by Stockfish's count.
White to play (after 29...f6). The knight is on f3 and the queen on c4. Which white piece is doing least, and where does it belong?
Hint
Look at the rook on e2 and the open file next to it.
A good answer
30.Rd2, taking the open d-file while Black's bishop on d8 blocks the black rooks, is Stockfish's first choice and the game move. Kapnisis's 30.Ng5 also wins; Stockfish rates both at more than five pawns for White.
Part 10Takeaways
- When there is no plan, find your worst piece and give it a job.
- Backward can be the most active direction: 24.Nb1 was Stockfish's first choice, and nothing else kept the edge.
- Spoil the opponent's best piece while you improve yours: c2-c3 sent the b4 knight to a6.
- The defender needs the rule most: 28...Nb8 was Black's saving idea, and it was Karpov's idea in reverse.
- Check the cost of the trip: a long manoeuvre is only free when the opponent has nothing to do.
Questions readers ask
What does improve your worst piece mean in chess?
It means that when you have no concrete plan, you find the piece doing the least for your position and move it to a square where it has a job. Because a position is only as strong as its weakest-placed piece, improving that one piece usually improves the whole position. Coaches often credit the idea to the trainer Mark Dvoretsky.
How do I find my worst piece?
Count, for each piece, the useful squares it controls and the jobs it does. The piece with the fewest is your worst piece: often a knight with no forward squares, a bishop blocked by its own pawns, or a rook on a closed file.
Why did Karpov play 24.Nb1 against Spassky?
To bring the knight from c3, where the c6 pawn and White's own a4 and e4 pawns left it no squares, round to f3 by way of d2. The move also guarded the rook on d2 and freed the c-pawn to play c2-c3, driving away Black's best piece, the knight on b4. Stockfish 19 rates 24.Nb1 as the best move in the position.
Who won the 1974 Karpov v Spassky Candidates match?
Anatoly Karpov won the semi-final in Leningrad with four wins, one loss and six draws; game 9 was the third of the four wins. Karpov then beat Viktor Korchnoi in the Candidates final and became world champion in 1975.
Sources and method
The game score was checked against two independent sources. Every position, line and candidate in this article was replayed for legality, and every engine number was produced by Stockfish 19 WASM Multithreaded for this article (depth 12 to 32, deepest at the critical moments), not copied from elsewhere. What the players thought is reconstructed from the moves, their published notes and later analysis; it is labelled as verified, inferred or speculative, and nothing is presented as a quotation unless a source is cited for it.
- PGN Mentor: Anatoly Karpov games, PGN archive (game score)
- American Chess Magazine #44 (2025): GM Spyridon Kapnisis, The positional genius of Anatoly Karpov (game score and notes)
- Sam Collins, Karpov: Move by Move (Everyman Chess), sample chapter with this game
- Raymond Keene, Arachnid, The Spectator, July 2015 (notes to the game)
- Mark Weeks: 1974 Candidates matches, Karpov v Spassky
Read next
Both rooks reached the seventh rank. Caruana's on b7 bit; Wei Yi's on b2 found nothing to bite. Stockfish shows that rook was not yet losing: one bishop move threw the draw away.
The Open File: Alekhine's Gun Against Nimzowitsch, CheckedSan Remo 1930: Alekhine tripled on the c-file and Nimzowitsch ran out of moves. Stockfish 19 says the game was decided by move 17, by a humble pawn on a6, and that the famous zugzwang was not needed.
Good Knight vs Bad Bishop: Karpov v Kasparov 1984, CheckedKarpov's knight beat Kasparov's bishop in game 9 of the 1984 match. Stockfish 19 shows the bad bishop held for 45 moves, then a loosening pawn move decided it.
Turn reading into skill
Strategy sticks when you use it. Practise real positions with Stockfish's verdict at once, or play the plans out from any diagram above.