Good knight against bad bishop · For club players · 14 min read
Good Knight vs Bad Bishop: Karpov v Kasparov 1984, Checked
Karpov'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.
- plan
- target
- Light-square pawns
- pawn chain or line of force
In short
- A bishop is bad when most of its own pawns stand on its colour: they block it and it can only defend. A knight, which does not care about colour, can then dominate it.
- In game 9 of the 1984 World Championship match, Karpov steered a level middlegame into a knight against a bishop hemmed in by Kasparov's light-square pawns, and won after 70 moves.
- Stockfish 19 disagrees with the tidy story: the bad bishop held an equal position for 45 moves. The ending was decided by a pawn move that loosened Black's own kingside, a slip by Karpov that nearly gave the win back, and one last bishop move to the wrong square.
- The lesson for club players is practical: a bad bishop is a long-term burden, not a lost position, so the knight's side must make the weaknesses real before cashing in.
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 1Four-nil down: Moscow, October 1984
Anatoly Karpov defended the world title against the 21-year-old Garry Kasparov in a match played to six wins. Before game 9 the champion led 3-0, with five draws, and the challenger had yet to score. Game 9 made it 4-0. It is one of the most studied minor-piece endings in the books. [4] Verified
| White | Black | |
|---|---|---|
| Player | Anatoly Karpov | Garry Kasparov |
| Rating then | 2700 | 2710 |
| Role | Champion | Challenger |
| Score before game 9 | 3 | 0 |
The game is a textbook for one idea: a good knight against a bad bishop. Fishbein, writing in American Chess Magazine, uses it as the 'classic example' of a bishop 'biting on its own pawns'. [3] Verified
Part 2What is a bad bishop, and why does a knight love it?
A bad bishop is a bishop whose own pawns stand mostly on squares of its colour. The pawns block its diagonals and leave it little to attack, so it becomes a defender. A good knight is one that has an outpost: a square, usually on the fifth or sixth rank, that no enemy pawn can attack, from where it hits targets and cannot be driven off.
A knight cannot be hemmed in by colour, because it hops over pawns and visits every square. That is the whole asymmetry: pawns on one colour hurt a bishop and hardly matter to a knight.
- outpost
- target
- Light-square pawns
- pawn chain or line of force
- plan
- Dark-square pawns
- pawn chain or line of force
Same material, different pawns. Left: Black's pawns sit on light squares, so the c8 bishop is walled in while White's knight enjoys e5, a square no pawn can attack. Right: Black's pawns stand on dark squares and the bishop eyes the whole c8-h3 diagonal.
Two caveats come first. A bad bishop is also a fine defender of the pawns it stands in front of, so the knight's side has to create a second target. And pieces matter less than the whole position: see the companion piece on bishop against knight for the opposite case.
Part 3How the Tarrasch Defence leaves a pawn on d5
Kasparov chose the Tarrasch Defence, which accepts an isolated queen's pawn on d5 (a pawn with no friendly pawns on the neighbouring files) in return for active pieces. The pawn is a source of energy in the middlegame. In an ending it is a liability and, crucially, it sits on a light square. The mechanics of that pawn are explained in the isolated queen's pawn article.
1.d4 d5 2.c4 e6 3.Nf3 c5 4.cxd5 exd5 5.g3 Nf6 6.Bg2 Be7 7.O-O O-O 8.Nc3 Nc6 9.Bg5 cxd4 10.Nxd4 h6 +0.43White slightly better
The Tarrasch line to the isolated pawn on d5. White fianchettoes the bishop and keeps a small pull. Verified
- target
- key square
- The d-file fight
The game itself is a slow manoeuvring battle in which Stockfish sees nothing but a small edge for White. Karpov trades queens on move 26 and, after the rooks come off on moves 33 and 35, the structure that matters is on the board. Inferred
Part 4Two plans from the same position
After 36...Ke7 the heavy pieces are gone. White has a knight on c3 and a bishop on e2, Black a knight on f6 and a bishop on e6. Black's pawns on a6, b5, d5 and h5 stand on light squares, so the bishop on e6 guards without attacking.
- plan
- target
- Knight's targets
- plan
- key square
- Black's squares
White aims the knight at the two light-square pawns the bishop and king must guard. Black wants active pieces so that the bishop is not the only defender.
The defender's counterplay depends on activity. As long as a Black piece can reach e4 and the king stands on d6, the d5 pawn is safe and the bishop has a job. Karpov's task is to take those squares away without hurrying. Inferred
Part 5The decisions that mattered
The graph shows how little happens for 45 moves. The first two decisions are quiet and strategic; the swings come later.
Critical moment 1 of 5 · Move 37, White to play
Where should the knight go?
- target
- key square
- Light-square pawns
The queens and rooks are off. White's knight on c3 has little to do. Which square makes the most of Black's light-square pawns?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 37.h4 37...Nd7 | +0.05about equal | Level, a twentieth of a pawn better than the move played. Stockfish sees no difference that matters. Verified |
| 37.Bd3 37...Kd6 | +0.12about equal | Also level, about a tenth of a pawn for White. Verified |
| 37.Kf3 37...Kd6 | −0.38Black slightly better | Loses a pawn: with the king on f3 the h3 pawn falls to ...Bxh3 and White is about a pawn down. Verified |
What Karpov played
37.Na2+0.02about equal
Karpov plays 37.Na2, heading for b4 to pressure d5 and a6 from a square where the bishop cannot reach it.
Stockfish preferred 37.Bd3 (+0.14); the game move gave up about 0.1 pawns. Depth 31.
What it revealed
The move shows the plan for the whole ending: the knight goes where Black's bishop cannot follow, and the pieces that guard d5 and a6 are tied down.
- Inferred From b4 the knight hits d5 and a6, the two pawns the bishop on e6 and the king must guard. This is the first sign of a second target.
- Verified Stockfish rates 37.Na2 and the sound alternatives h4 and Bd3 within a tenth of a pawn of each other: this is a plan move, not a winning move. Only 37.Kf3 is a mistake, since it loses the h3 pawn.
Critical moment 2 of 5 · Move 44, White to play
Which pieces come off?
- threat
- target
Black's knight on f5 is the one active piece in the position. White has a bishop on d3 and a knight on c2. What should White trade?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 44.Ne3 44...Nxe3 45.Kxe3 | 0.00about equal | Level. The knights come off and the bishops stay, which is what the bad bishop wants. Verified |
| 44.hxg5 44...fxg5 45.Ne3 | +0.06about equal | Also level, a few hundredths of a pawn for White; it opens the king side before the knight is in place. Verified |
What Karpov played
44.Bxf5+0.02about equal
Karpov plays 44.Bxf5. White gives up the bishop for Black's only good piece and leaves a knight against the c8 bishop.
Stockfish's first choice too. Depth 37.
What it revealed
Trading a bishop for a knight sounds like a concession. Here it is the whole strategy: the piece traded is the one that could have defended, and the piece left is the one that can win.
- Verified Stockfish also chooses 44.Bxf5, so the structural decision is supported by the engine, not just by the story.
- Inferred After 44...Bxf5 the remaining Black bishop still sits behind d5 and h5 while White's knight can head for e3 and g2.
Part 6The turning point: 46...gxh4?
Everything so far is level. The decision that changes the evaluation by more than two pawns is one pawn capture.
Critical moment 3 of 5 · Move 46, Black to play
Capture the pawn on h4, or wait?
- threat
- target
- key square
Black's g5 pawn can take on h4. White threatens to take on g5 and open the king side. What should Black do with the move?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 46...Ke6 47.Kg2 Kd6 | +0.04about equal | Level: the king keeps guard of d5 and the position holds, as Fishbein says. Verified |
| 46...Bg6 47.hxg5 fxg5 | 0.00about equal | Also level: Stockfish plays it out with ...fxg5 and equal chances. Verified |
| 46...Ke7 47.Kg2 Kd6 | +6.76White completely winning | Loses a pawn at once: the king leaves d6 and Nxd5+ wins the pawn, which leaves White better by a large margin. Verified |
What Kasparov played
46...gxh4+3.55White winning
Kasparov plays 46...gxh4, trying to simplify. After 47.Ng2 the h-pawns are weak and White's king and knight can pick them off.
Stockfish preferred 46...Ke6 (+0.02); the game move gave up about 3.5 pawns. Depth 24.
What it revealed
The capture removes the pawn that was tying White down. The h5 pawn is now alone, the king side is open to White's king and knight, and the bishop cannot defend both wings.
What Stockfish adds The bad bishop was not the problem: Stockfish finds two level moves and a swing of about three pawns for the capture, which opens the king side for White's king and knight.
- Verified Fishbein calls 46...gxh4 'a fatal mistake' and notes that 'a waiting move would hold'.[3]
- Verified Stockfish agrees: a waiting move keeps the evaluation level, and the capture swings it to about three pawns for White.
47.Ng2 hxg3+ 48.Kxg3 Ke6 49.Nf4+ Kf5 50.Nxh5 Ke6 51.Nf4+ Kd6 52.Kg4 +3.30White winning
White's king and knight win the h5 pawn. The knight on f4 hits d5 and the king walks toward h5 and g6. Verified
- plan
- threat
- target
- White's grip
Part 7What could Black have done? Two chances
Black had a second chance, and it came from the winner. Karpov spent the ending doing what a knight should do, but at move 55 the way the d5 pawn was taken gave the whole advantage back.
Critical moment 4 of 5 · Move 55, White to play
Take the d5 pawn now?
- target
White's king stands on g6 and the d5 pawn hangs to the knight. Is there a reason to pause before capturing it?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 55.Nh5 55...Bxf3 56.Nxf6 | +3.76White winning | Keeps the win: about four pawns for White. This is Fishbein's 55.Nh5. Verified |
| 55.Nd3 55...Bxf3 | −12.24Black completely winning | Loses a piece: after ...Bc2 the knight is pinned to the king on g6 and falls, and Stockfish has Black winning. Verified |
What Karpov played
55.Nxd5+0.00about equal
Karpov plays 55.Nxd5+, winning the pawn at once.
Stockfish preferred 55.Nh5 (+3.85); the game move gave up about 3.9 pawns. Depth 31.
What it revealed
The pawn was poisoned in the sense that taking it lets Black's king become active. Fishbein writes that the key for White is to take d5 under conditions in which Black's king cannot become active.
What Stockfish adds Stockfish and Fishbein agree: the move played gives away the whole advantage. Black's saving line is a draw.
- Verified Fishbein calls the capture an error, gives 55.Nh5 as the winning move, and refers readers to Dvoretsky's Endgame Manual for the full analysis of this ending.[3]
- Verified Stockfish moves from almost four pawns for White to a level evaluation after 55.Nxd5+.
55.Nh5 Bxf3 56.Nxf6 Ke6 57.Ne8 Kd7 58.Ng7 Kd6 59.Kf6 Bd1 60.Nf5+ Kc6 61.Ke5 Bb3 62.Ne7+ Kd7 63.Nxd5 +7.20White completely winning
The improvement Fishbein gives. White takes f6 first, and only later d5, when Black's king is cut off. Verified
Critical moment 5 of 5 · Move 66, Black to play
Where does the bishop belong?
- key square
Black is two pawns down, with a bishop against a knight. Which square for the bishop keeps the draw within reach?
Think it through on the boardWeigh the candidates
| Move | Stockfish | Verdict |
|---|---|---|
| 66...Bh1 67.Nf5 Kd5 | +0.57White slightly better | Holds: about half a pawn for White. This is the move Fishbein gives to save the game. Verified |
| 66...Bf1 67.Nf5 | +8.46White completely winning | Loses: Stockfish has White more than eight pawns up, since the bishop abandons the long diagonal. Verified |
What Kasparov played
66...Bb7+6.40White completely winning
Kasparov plays 66...Bb7, putting the bishop on a diagonal where the knight can harass it.
Stockfish preferred 66...Bh1 (+0.22); the game move gave up about 6.2 pawns. Depth 39.
What it revealed
A single bishop move turned an ending that Stockfish rated at about a fifth of a pawn into a lost one. With so few pieces left, one wrong square is enough.
What Stockfish adds The decisive error is neither the pawn capture nor the king walk, but a bishop move to the wrong square.
- Verified Fishbein calls 66...Bb7 'the decisive error' and writes that after 66...Bh1 67.Nf5 Kd5 'White cannot win, even two pawns up'.[3]
- Verified Stockfish rates 66...Bh1 at about half a pawn for White, and the move played at more than six pawns.
Part 8A fresh angle: the bad bishop held for 45 moves
| Move | Who | Before | After |
|---|---|---|---|
| 46...gxh4 | Kasparov | level | about three pawns for White |
| 55.Nxd5+ | Karpov | almost four pawns for White | level |
| 66...Bb7 | Kasparov | about a fifth of a pawn for White | more than six pawns for White |
That does not make the strategy wrong. It makes it honest about its scale: a bad bishop is a standing handicap that raises the cost of every mistake, and the knight's side wins by making the opponent play almost perfectly for 70 moves. Kasparov defended an equal ending until one pawn capture; Karpov converted a second chance, and the bishop's scope in the final phase is the part the engine finds hardest to explain in a rule. Inferred
Part 9How to use a good knight against a bad bishop
The recipe
- Count the pawnsCount the opponent's pawns on the bishop's colour. If they outnumber the others, the bishop is bad.
- Trade the restExchange the pieces that could help it: rooks, the other minor pieces, the queens. Leave the knight and the bishop.
- Find an outpostPut the knight where no pawn can attack it, on the bishop's weak squares.
- Create a second targetA bad bishop defends well. Make a weakness on the other wing, and bring the king.
- Do not take the first pawnTake material only when the opponent's king cannot become active.
- Which of my opponent's pawns are on the bishop's colour?
- Is there a square for my knight that no pawn can attack?
- Where is the opponent's king, and how active?
- Do I have a second target, away from the pawns the bishop guards?
Your turnPractise it
Set each position up, decide your plan before you look, then play it out against Stockfish to see whether it holds.
Black to play. The g5 pawn can capture on h4. Is there something better?
Hint
Ask what the capture opens, and for whom.
A good answer
Play a waiting move such as 46...Ke6 (46...Bg6 also holds). Stockfish keeps the position level, while 46...gxh4 gives White about three pawns of advantage.
White to play. The d5 pawn can be taken with check. Should it be?
Hint
Look at the f6 pawn and what the Black king can do after the capture.
A good answer
No: 55.Nh5! first, aiming at f6. Stockfish keeps about four pawns for White; 55.Nxd5+ lets Black's king become active and the position becomes level.
Black to play, two pawns down. Where should the bishop go?
Hint
The bishop needs a long diagonal and must stay out of the knight's reach.
A good answer
66...Bh1!, which Fishbein gives as the move that holds. Stockfish rates it at about half a pawn for White; 66...Bb7 loses by more than six pawns.
White to play. Black's knight on f5 is strong. Which piece should be traded?
Hint
Which piece does Black need, and which does White want to keep?
A good answer
44.Bxf5 Bxf5 45.Ne3: the bishop goes for Black's only active piece, leaving a knight against the bad bishop. Stockfish rates every sensible move level; this is the one the game was built on.
Part 10Takeaways
- A bad bishop is a handicap, not a defeat. Stockfish rated game 9 level for 45 moves.
- Trade the pieces that could help the bishop. Karpov exchanged a bishop for the only active black piece.
- Put the knight where pawns cannot reach it. The route a2-b4 was the quiet plan.
- Do not rush the first pawn. 55.Nxd5+ threw the whole advantage away.
- A bishop needs the right square. Black's last mistake put it on the wrong one.
Questions readers ask
What is a bad bishop in chess?
A bad bishop is one whose own pawns mostly stand on squares of its colour. They block its diagonals, so it can defend but cannot attack. A knight is not affected by pawn colour, so a good knight often beats it.
When is a knight better than a bishop?
A knight is usually better in closed positions, when the pawns are locked and the bishop is hemmed in, and when the knight has an outpost that no pawn can attack. The bishop is better when the position is open and pawns are on both wings.
Is a bad bishop really bad?
Not always. It defends the pawns in front of it well, and the position can stay level for a long time. In Karpov v Kasparov, 1984 game 9, Stockfish rated the position level for 45 moves despite the bad bishop.
How do you play against a bad bishop?
Trade the pieces that could help the bishop, plant a knight on an outpost, create a second target on the other wing and bring the king. Take material only when the opponent's king cannot become active.
What happened in Karpov v Kasparov 1984 game 9?
Karpov won in 70 moves with a knight against Kasparov's bad bishop after a Tarrasch Defence, taking a 4-0 lead in the match. Kasparov's 46...gxh4 and 66...Bb7 were the decisive errors; Karpov's 55.Nxd5+ threw away the advantage in between.
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 17 to 40, 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.
Read next
In game 7 of the 1971 Candidates final, Fischer traded a dominant knight for a bad bishop and won a textbook bishop-against-knight ending. Stockfish 19 says the trade was not best. Here is why it worked anyway.
Bishop Pair Strategy: Kramnik v Tiviakov 2001, Checked by StockfishKramnik, the new world champion, turned the two bishops into a win at Wijk aan Zee 2001. Stockfish 19 shows what the pair was worth and where both players let the advantage slip.
Outposts: Kasparov's Octopus Knight v Karpov, Checked by StockfishKasparov's knight on d3 in game 16 of the 1985 match, what Stockfish says about it, and how to build an outpost no pawn can attack.
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.