Most chess puzzles are designed to be solved. You spot the fork, you find the back-rank mate, you feel clever, and you move on. But a tiny handful of positions are different. They are so deep, so counter-intuitive, and so beautiful that they have humbled World Champions, embarrassed the strongest computers, and taken composers decades to perfect. These are not tactics from a weekend tournament - they are works of art.
In this guide we count down five of the most difficult chess puzzles ever created. Every one is a genuine, documented composition with a verifiable history, and for four of them we have included a full board diagram and the solution. If you are still learning the ropes, you may want to start with our complete beginner's guide to chess first - but if you enjoy a real mental challenge, read on.
Difficulty in chess composition is not about having lots of pieces or a flashy sacrifice. The puzzles that earn legendary status almost always share a few specific qualities:
If you want to understand why strong players see these ideas faster, our article on core chess strategy for beginners explains the pattern-recognition habits that separate club players from masters.
| Puzzle | Composer | Year | Goal | What makes it brutal |
|---|---|---|---|---|
| Plaskett's Puzzle | Gijs van Breukelen | c.1970 | White to play and win | Two knight under-promotions; only Tal solved it |
| The Babson Task | Leonid Yarosh | 1983 | White mates in 4 | Every Black promotion matched by White |
| Saavedra Position | F. Saavedra / Barbier | 1895 | White to play and win | Under-promote to a rook to dodge stalemate |
| Reti's Study | Richard Reti | 1921 | White to play and draw | King defies the "rule of the square" |
| Longest tablebase mate | Computer-discovered | 2012+ | Forced mate | Over 500 moves - beyond human reach |
This study was composed by the Dutch endgame artist Gijs van Breukelen around 1970. It earned its name in 1987, when English grandmaster James Plaskett showed it to the elite field at a tournament in Brussels. According to contemporary accounts collected by ChessBase, several world-class grandmasters stared at the position and gave up - all except former World Champion Mikhail Tal, who reportedly went for a walk in the park and came back with the answer.
White wins with a stunning sequence: 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Bc2+ Kxh5 4.d8=Q Nf7+ 5.Ke6 Nxd8+ 6.Kf5 e2 7.Be4 e1=N! Here Black is forced to under-promote to a knight just to survive. White continues 8.Bd5 c2 9.Bc4 c1=N! - a second knight under-promotion - 10.Bb5 Nc6 11.Bxc6 Nc7 12.Ba4, and the lone white bishop somehow weaves a mating net through Black's crowd of knights.
For nearly sixty years, the Babson Task was the impossible dream of problem composers. Named after the American Joseph Ney Babson, who described it around 1925, the challenge was breathtaking: compose a position where Black can promote a pawn to a queen, rook, bishop, or knight - and in each case White is forced to reply with the exact same promotion to force mate. In 1934 the great endgame theorist Andre Cheron declared it would never be done.
Then, in March 1983, an unknown 26-year-old football coach from Kazan named Leonid Yarosh published a clean, legal solution in the Soviet magazine Shakhmaty v SSSR. The author Tim Krabbe famously wrote that learning the Babson had finally been solved felt like opening a newspaper to the headline "Purpose Of Life Discovered." You can read the full story on Wikipedia's Babson Task page.
It is a mate in four. After the key move, the four thematic lines run: 1...cxb1=Q 2.axb8=Q, 1...cxb1=R 2.axb8=R, 1...cxb1=B 2.axb8=B, and 1...cxb1=N 2.axb8=N - each leading to a forced mate, and crucially, no other White promotion works in each line. That perfect mirror symmetry is what makes it the single most celebrated construction task in chess.
This is arguably the most famous endgame study of all time, and its backstory is wonderful. It grew out of an 1875 game between Richard Fenton and William Potter. Twenty years later the position was reprinted as a study marked "White to play and draw." Then a weak amateur - a Spanish priest named Reverend Fernando Saavedra, living in Glasgow - spotted something everyone else had missed. White was not drawing. White was winning.
The win runs 1.c7 Rd6+ 2.Kb5 Rd5+ 3.Kb4 Rd4+ 4.Kb3 Rd3+ 5.Kc2 Rd4, and now the magic: 6.c8=R! White under-promotes to a rook. Why not a queen? Because after 6.c8=Q? Rc4+! 7.Qxc4 Black is stalemated and the game is drawn. The humble rook keeps the win alive, threatening mate. After 6...Ra4 7.Kb3, Black must either lose the rook or be mated. You can explore the full history on the Saavedra Position Wikipedia article.
Published in 1921 by the hypermodern master Richard Reti, this tiny study looks utterly hopeless for White. The black h-pawn is racing to promote with its king alongside, while White's king sits in the far corner, seemingly miles outside the "square" of the pawn, and White's own c-pawn looks too slow to matter. Beginners and masters alike conclude it must be lost.
And yet White draws with 1.Kg7! h4 2.Kf6 Kb6 3.Ke5! The secret is that the king moves on a diagonal that does two jobs at once: it edges toward Black's runaway pawn while simultaneously rushing to support White's own pawn. If Black captures on c6, White's king is suddenly back inside the square of the h-pawn; if Black pushes the h-pawn, White promotes the c-pawn. The composer Abram Gurvich nicknamed this theme "the hunt of two hares." It is covered in depth on the Reti Endgame Study Wikipedia page.
The Reti idea reshaped how players think about king activity in the endgame - a concept you will see again and again if you study the classics in our list of famous chess games every player should study.
The final entry is a different kind of "hardest." It is not a human composition at all - it is a discovery made by computers. Using endgame tablebases (databases that have perfectly solved every position with a small number of pieces), researchers analysing all seven-piece endings uncovered forced checkmates of astonishing length. The longest known examples run to over 500 moves - one celebrated position is a forced mate in 549 moves.
No human being could ever calculate such a sequence, and to a watching grandmaster the moves look completely random for hundreds of moves before the mating idea appears. These positions are also a reminder of how far engines have come; if you are curious how that technology works in practice, see our explainer on how to use Stockfish and our roundup of the best chess engines of 2026. The science behind tablebases is summarised on the Endgame Tablebase Wikipedia page.
| Technique | What it means | Seen in |
|---|---|---|
| Under-promotion | Promoting to a rook, bishop or knight instead of a queen | Saavedra, Plaskett, Babson |
| Stalemate defence | The losing side escapes by forcing stalemate | Saavedra, Babson lines |
| Dual-purpose king march | One king move that achieves two goals at once | Reti's Study |
| Allumwandlung | All four promotion pieces appear in the solution | The Babson Task |
| Extreme depth | Solutions too long for human calculation | Tablebase mates |
You will not solve a Babson Task tomorrow - but you can build the calculation and pattern-recognition skills that make hard puzzles approachable. Here is a practical plan:
Free live multiplayer chess with Elo ratings, skill-based rooms, and real human opponents. No signup needed to start.
The puzzles on this list are more than brain-teasers - they are proof that a board with 64 squares still hides ideas deep enough to defeat champions and computers alike. You do not need to solve them to enjoy them; understanding why the quiet move, the under-promotion, or the impossible king walk works is its own reward. The best next step is simple: start calculating. Open a live game on ChessDada, look for the move nobody expects, and keep training. For more guides and stories, browse the ChessDada Chess Blog.