The Top 10 Chess Puzzles of All Time
The ten greatest chess puzzles ever set are, in order: Ortueta–Sanz 1933, Rotlewi–Rubinstein 1907, the Immortal Game 1851, the Evergreen Game 1852, Levitsky–Marshall 1912, Torre–Lasker 1925, Lasker–Thomas 1912, Byrne–Fischer 1956, Kasparov–Topalov 1999 and Sam Loyd's Excelsior 1861. Nine came from real games. Every position below is shown as a puzzle first — try it before you scroll.
- What makes a chess puzzle great
- 1. Ortueta–Sanz, Madrid 1933
- 2. Rotlewi–Rubinstein, Lodz 1907
- 3. The Immortal Game, London 1851
- 4. The Evergreen Game, Berlin 1852
- 5. Levitsky–Marshall, Breslau 1912
- 6. Torre–Lasker, Moscow 1925
- 7. Lasker–Thomas, London 1912
- 8. Byrne–Fischer, New York 1956
- 9. Kasparov–Topalov, Wijk aan Zee 1999
- 10. Sam Loyd, Excelsior 1861
- How to solve puzzles like these
- Frequently asked questions
What Makes a Chess Puzzle Great
Every serious player has solved thousands of tactics. Almost none of them stay in the memory. The handful that do share three qualities, and it is worth naming them before we start, because they are what separates the ten positions below from the next ten thousand.
The first move looks wrong. Not clever — wrong. It hangs a queen, or it gives away a rook for a pawn, or it pushes the one pawn on the board that nobody was looking at. A puzzle whose first move is the move you wanted to play anyway is an exercise, not a puzzle.
The point arrives late. In a great puzzle the justification for the sacrifice is three or four moves further on, in a position you have to see before you commit. That gap between the move and its reason is the whole experience.
It is forced. The opponent is never allowed a real choice. A beautiful idea that only works if the defender cooperates is a swindle. Every solution on this page holds up against best defence, and we checked all of them against Stockfish move by move before publishing — see the verification note at the end.
How to use this page. Each puzzle is shown as a diagram with the side to move and nothing else. Cover the text under it, set a five-minute timer, and write down the whole line before you read on. Solving from a diagram you have already read the answer to is the single most common way players waste study time.
1. Ortueta–Sanz, Madrid 1933 — Two Pawns Against the World
We start with the position that professional players nominate more often than any other when asked for the most beautiful sequence ever played. It is not from a world championship. It is from a club-level event in Madrid in 1933, and for decades people argued about whether it had even happened.
Count the material honestly. White has a rook and a knight; Black has a rook and a bishop and two connected passed pawns on c4 and c5. White's rook on b7 is active, the knight covers the queening squares, and any normal move — moving the bishop, pushing a pawn, defending the b-pawn — lets White consolidate. Black found something else.
The solution
1...Rxb2!! The rook takes a pawn and is simply hanging. 2.Nxb2 c3! and suddenly the knight cannot move: it is attacked by a pawn that is one square from becoming unstoppable. 3.Rxb6 — White grabs the bishop with tempo, hitting nothing but hoping the extra piece counts. 3...c4!! is the move that made the position famous. Black ignores the rook and pushes the other pawn. 4.Rb4 a5! attacks the rook a second time, and after 5.Nxc4 c2 the pawn cannot be stopped by anything on the board.
Why it is so hard. Every move of the combination gives away material, and the payoff is a pawn on the second rank five moves later. There is no check anywhere in the line. Solvers who look for checks and captures — which is the correct habit almost every other time — will not find it.
And neither does the engine, at first. We put the diagram to Stockfish at search depth 26 and asked for its four best moves. 1...Rxb2 was not among them; it preferred the quiet 1...Kg6 at +2.28. Play 1...Rxb2 on the board, though, and the evaluation immediately jumps to +4.83 for Black — and from there every single move White chose in the game turned out to be the engine's own first choice, with Black's advantage growing to +5.17. The combination is completely sound. It is simply over the horizon of a normal search, which is a fair description of what it did to the players too.
2. Rotlewi–Rubinstein, Lodz 1907 — Every Piece at Once
Akiba Rubinstein was the best player in the world never to play a world championship match, and this is the game his name is attached to. It is known simply as Rubinstein's Immortal.
Black's queen is attacked. The natural instinct is to move it, and every queen move is met by a defence. Rubinstein left it where it was.
The solution
1...Rxc3! — a rook for a knight, with the black queen still hanging. 2.gxh4 Rd2!! The second rook walks into the position where it can be taken by the queen, and Black is now a full queen down with two rooks en prise. 3.Qxd2 Bxe4+ 4.Qg2 Rh3! and White resigned: the rook on h3 cannot be captured because of mate on h2, and mate on g2 cannot be prevented.
What makes this the connoisseur's choice is that there is not a single wasted move. Every black piece — both bishops, both rooks and the queen that was sacrificed to clear the way — is doing work in the final position. It is the model example of what coordination actually means.
3. The Immortal Game, London 1851 — Anderssen's Three Pieces
Adolf Anderssen against Lionel Kieseritzky, played casually in a London coffee house during the world's first international tournament. By the diagram position Anderssen has given away a bishop and both rooks and is about to give away his queen.
The solution
1.Nxg7+ Kd8 2.Qf6+!! — the queen is offered to the knight on g8, and it must be taken because the alternative is mate immediately. 2...Nxf6 3.Be7#. The mate is delivered by two knights and a bishop while Black still owns a queen, two rooks and a bishop that have never left their starting area.
Honest footnote. Modern engines do not think much of the whole game — Black had several defences earlier and 1851 was not an era of accurate defence. But the final position is genuine and forced, and as a puzzle it is perfect: the only way in is to give away the last valuable piece you own.
4. The Evergreen Game, Berlin 1852 — Anderssen Again
The following year Anderssen produced a finish that Wilhelm Steinitz called an evergreen in the laurel crown of the departed chess master. This one is harder to find than the Immortal, because the first move is quiet.
The solution
1.Rxe7+! Nxe7 (1...Kd8 avoids the immediate mate but loses just as surely after 2.Rxd7+) 2.Qxd7+!! Kxd7 3.Bf5+ Ke8 4.Bd7+ Kf8 5.Bxe7#. Black's own threatened mate on g2 never gets to happen, because from move one White never gives him a free move again.
Anderssen's 19.Rad1, the move immediately before the diagram, is the one annotators praise most: it develops the last piece and does nothing forcing at all, one move before an eight-move forced finish. That is the kind of move engines find instantly and humans almost never do.
5. Levitsky–Marshall, Breslau 1912 — The Gold Coins Move
If a single move has ever become more famous than the game it came from, it is this one. Frank Marshall, the American champion, was playing Stefan Levitsky at the German chess federation congress in Breslau.
The solution
1...Qg3!! and White resigned on the spot. Look at the square: the queen can be taken by the h-pawn, by the f-pawn, and by the white queen on g5. All three lose.
- 2.hxg3 — 2...Ne2 is mate.
- 2.fxg3 — 2...Ne2+ 3.Kh1 Rxh2 is mate.
- 2.Qxg3 — 2...Ne2+ picks up the queen with a fork and Black is winning easily.
The story says the spectators showered the board with gold coins. Chess historians have looked hard and found no contemporary evidence for it, and Marshall himself told the tale in different ways at different times. The move needs no legend.
6. Torre–Lasker, Moscow 1925 — The Windmill
Carlos Torre was a 20-year-old from Mexico. Emanuel Lasker had been world champion for 27 years. What Torre did to him in Moscow is now the textbook illustration of a tactic called the windmill, or see-saw.
The solution
1.Bf6!! offers the queen and ignores the attack on it. 1...Qxh5 2.Rxg7+ Kh8 3.Rxf7+ — and here is the mechanism. The rook steps off g7 with discovered check from the bishop on f6, helps itself to whatever is on the new square, and then returns to g7 with check. The king can never do anything but step between h8 and g8.
The rook harvests f7, then b7, and finally after 3...Kg8 4.Rg7+ Kh8 5.Rxb7+ Kg8 6.Rg7+ Kh8 7.Rg5+ Kh7 8.Rxh5 it collects the queen as well.
The windmill is worth knowing as a pattern rather than as a puzzle, because once you have seen it you will spot the next one in seconds: a bishop on a long diagonal, a rook that can check on the seventh, and an enemy king locked on two squares.
7. Lasker–Thomas, London 1912 — The Longest Walk
A different Lasker — Edward, a distant relative of the world champion — playing a casual game at the City of London Chess Club. Sir George Thomas was one of England's strongest amateurs. The game lasted eighteen moves.
The solution
1.Qxh7+!! Kxh7 2.Nxf6+ Kh6 3.Neg4+ Kg5 4.h4+ Kf4 5.g3+ Kf3 6.Be2+ Kg2 7.Rh2+ Kg1 8.Kd2# — eight white moves, every one of them forced. The black king is dragged from h7 to g1 — the whole length of the board, into White's own back rank — and mated by a quiet king move because White has run out of pieces that can check.
Solve this one. Of all ten puzzles here, this is the one a beginner can genuinely find, because every single black move is forced and every white move is a check. If you check every check, you will get there. Lasker himself said he could have mated a move earlier with 8.O-O-O, and castling to deliver mate would have been even better — but he did not see it.
8. Byrne–Fischer, New York 1956 — The Game of the Century
Bobby Fischer was thirteen years old. Donald Byrne was one of the strongest players in the United States. The position below arrived on move seventeen, after Byrne had attacked Fischer's queen with 17.Kf1.
The solution
1...Be6!! Fischer does not move the queen. He develops a bishop, offers the queen, and relies on what happens if White takes it.
If 2.Bxb6 then 2...Bxc4+ 3.Kg1 Ne2+ 4.Kf1 Nxd4+ 5.Kg1 Ne2+ 6.Kf1 Nc3+ 7.Kg1 axb6 and Black emerges with a rook, two bishops and a knight for the queen — more than enough, with White's position in ruins. And if White declines with 2.Qxc3 instead, 2...Qxc5 leaves Black simply winning.
The detail that makes the move immortal is the line Byrne avoided: 2.Bxe6 Qb5+ 3.Kg1 Ne2+ 4.Kf1 Ng3+ 5.Kg1 Qf1+!! 6.Rxf1 Ne2# — a smothered mate, found over the board by a thirteen-year-old.
9. Kasparov–Topalov, Wijk aan Zee 1999 — The Modern Immortal
The most recent position on this list, and the only one whose combination runs for more than twenty moves. Garry Kasparov was world champion; Veselin Topalov would become one. This is the game most often named as the best of the computer era.
The solution
1.Rxd4!! cxd4 2.Re7+ Kb6 3.Qxd4+ Kxa5 — the black king is pulled out to a5, four ranks from home, and now White must keep checking or lose. 4.b4+ Ka4 5.Qc3 Qxd5 6.Ra7 Bb7 7.Rxb7 Qc4 8.Qxf6 Kxa3 9.Qxa6+ Kxb4 10.c3+! Kxc3 11.Qa1+ Kd2 12.Qb2+ Kd1 13.Bf1! and the king, having walked from b8 to d1, is finally trapped. Kasparov won on move 44.
The honest part. Stockfish agrees that 24.Rxd4 is the best move in the position — it is the engine's own first choice at depth 30 — but it rates the resulting position at only about 0.00. In other words, the most celebrated combination of the modern era is, objectively, a way of keeping the game level rather than a forced win. Topalov had defences later, and Kasparov had to see fifteen moves of forced king hunt at the board without any of them. That gap between beauty and correctness is exactly what makes it worth showing.
10. Sam Loyd, Excelsior 1861 — The Least Likely Piece
The one composed problem on the list, and the one with the best story. Sam Loyd bet a friend dinner that he could not pick the piece that would not give mate. The friend pointed at the pawn on b2, the most harmless-looking unit on the board.
The solution
1.b4! threatening 2.Rf5 or 2.Rd5 followed by mate on the back rank. Black's best try is 1...Rc5+, blocking the rook's path with check, and White answers 2.bxc5! — the b-pawn, now on c5, threatens 3.Rb1 mate. After 2...a2 3.c6 Bc7 4.cxb7 the pawn has crossed the board, and whatever Black plays, 5.bxa8=Q# ends it.
A note on engines and problems. Ask Stockfish what to play here and it answers 1.bxa3, winning a pawn, at +6.03. It is not wrong — that move does win the game. But a problem asks a different question from the one an engine answers: not "what is best?" but "what mates in exactly five?" That is why composed problems remain a genuinely human corner of chess, and why solving software has to be told to look for mate before it finds the key.
The name comes from a Longfellow poem about a youth marching upward with a banner. Problem composers still call any problem in which a pawn marches from its starting square to promotion an Excelsior, which is a rare case of one puzzle giving its name to an entire theme.
The Ten at a Glance
| # | Position | Year | The key move | Difficulty |
|---|---|---|---|---|
| 1 | Ortueta–Sanz | 1933 | 1...Rxb2!! then 3...c4!! | Very hard |
| 2 | Rotlewi–Rubinstein | 1907 | 1...Rxc3! and 2...Rd2!! | Hard |
| 3 | Anderssen–Kieseritzky | 1851 | 2.Qf6+!! | Medium |
| 4 | Anderssen–Dufresne | 1852 | 2.Qxd7+!! | Medium |
| 5 | Levitsky–Marshall | 1912 | 1...Qg3!! | Hard |
| 6 | Torre–Lasker | 1925 | 1.Bf6!! | Medium |
| 7 | Ed. Lasker–Thomas | 1912 | 1.Qxh7+!! | Easy to spot, long to finish |
| 8 | Byrne–Fischer | 1956 | 1...Be6!! | Hard |
| 9 | Kasparov–Topalov | 1999 | 1.Rxd4!! | Very hard |
| 10 | Loyd, Excelsior | 1861 | 1.b4! | Hard |
How to Solve Puzzles Like These
Nine of the ten positions above turn on the same handful of ideas. If you train those ideas deliberately instead of clicking through random tactics, your solving rate on hard positions moves faster than almost anything else you can do with study time.
Look at forcing moves first, but look at all of them
Checks, captures and threats — in that order, and every one, including the ones that hang a piece. The reason strong players find moves like 1...Qg3 and 1.Bf6 is not deeper calculation. It is that their first pass over the position does not filter out moves that lose material. Beginners never even reach those moves, so no amount of calculation will help.
Ask what each defender is doing
Almost every brilliant move on this page works because one enemy piece is doing two jobs at once. Marshall's 23...Qg3 works because the h-pawn and the f-pawn are both needed where they stand. Fischer's 17...Be6 works because the bishop on c4 cannot both stay on the a2–g8 diagonal and keep guarding f1. Find the overworked piece and the move usually finds itself.
Calculate to a quiet position, not to the end of your checks
Ortueta–Sanz has no checks at all, and Kasparov–Topalov has almost nothing but checks for fifteen moves. What they share is that you cannot judge the line until you reach a position where nothing is forced any more and you can simply look at it. Stopping calculation one move early is the most common cause of a failed combination.
Write the line down before you move
On a board, on paper, in your head — but completely. Guessing the first move of a puzzle and then reading the answer feels like solving and teaches nothing, because the part you actually need to train is the four moves after the first one.
♔ Try your tactics in a real game
Reading about combinations is one thing. Finding one over the board against an opponent who is trying to stop you is another. Play free on ChessDada — no sign-up needed.
Was this article helpful?
Frequently Asked Questions
How these positions were checked
- Every position was rebuilt from the published game score with python-chess, so the diagrams are the positions that actually occurred, not redrawn from memory.
- Each one was then analysed with Stockfish to search depth 30, and every solution line was replayed move by move with the evaluation checked after each move. The figures quoted in the captions are that engine's output for the exact position shown.
- Where the engine disagrees with the historical verdict — the Immortal Game and the defence Topalov missed in 1999 — we say so in the text rather than quietly leaving it out.
- Board and piece graphics use the Cburnett piece set.
Final Thoughts
What the ten positions on this page have in common is not difficulty. Several of them are easier than the tactics you will be served by a puzzle trainer this evening. What they have is a moment where the ordinary rules of the game stop applying — where a queen is worth less than a tempo, where a pawn on b2 outranks every piece on the board.
That is the reason they survive. A hard puzzle is forgotten as soon as it is solved. A beautiful one gets replayed for a hundred and seventy years.
Set them up on a real board. Then go and try to make one of your own: start a game in the ChessDada lobby, work through the daily puzzles, or read more in the ChessDada Chess Blog.
Keep Reading
Plaskett's Puzzle, the Babson Task and the Saavedra Position — the hardest, rather than the most beautiful.
The complete games behind several of the positions above, from the first move to the last.
The building blocks every combination on this page is made of, explained from the beginning.
Smothered mate, Boden's mate, the Arabian mate and seven more patterns worth memorising.