One of Sudoku's great ironies: the puzzle requires no mathematics to solve — no arithmetic, no formulas, just deduction — yet it sits on top of some genuinely deep mathematics. The digits could be replaced with nine colours or nine animals and the puzzle would play identically. But ask innocent-sounding questions like "how many Sudoku grids exist?" and you fall straight down a combinatorial rabbit hole that took serious computing power to escape.

How many completed grids are there?

A completed Sudoku grid is a 9×9 array where every row, column, and 3×3 box contains 1–9 exactly once. For years the exact count was an open question. In 2005, Bertram Felgenhauer and Frazer Jarvis settled it by a carefully organised computer enumeration:

6,670,903,752,021,072,936,960 — about 6.67 sextillion (6.67 × 10²¹) valid grids.

For scale: if every human on Earth solved one unique grid per second, it would take roughly 26,000 years to exhaust them. Interestingly, that gigantic number is only a tiny fraction of the ~5.5 × 10²⁷ Latin squares of order 9 — the 3×3 box constraint eliminates about 99.9999% of them.

…but many are secretly the same

Two grids that differ only by cosmetic transformations — rotating the board, mirroring it, swapping all the 3s and 7s, exchanging two rows within the same band — are, to a mathematician, the same puzzle wearing different clothes. Accounting for every such symmetry, Ed Russell and Frazer Jarvis showed in 2006 that the number of essentially different Sudoku grids is

5,472,730,538 — about 5.5 billion genuinely distinct solutions.

Still more than you'll solve in a lifetime, but a striking illustration of how much apparent variety is really symmetry.

The magic number 17

A proper Sudoku must have exactly one solution. That raises the constructor's key question: how few clues can a valid puzzle have?

Puzzle makers had found many 17-clue puzzles over the years, but never a 16-clue one. In 2012, Gary McGuire's team at University College Dublin closed the question with one of the great brute-force proofs of the computing era: an exhaustive search — roughly 800 processor-years of computation — showing that no 16-clue Sudoku puzzle with a unique solution exists. Seventeen is the floor.

Two important footnotes. First, 17-clue puzzles are rare treasures: only tens of thousands are known, and they're prized by collectors. Second — and this surprises most players — fewer clues does not mean harder. Difficulty depends on which solving techniques the puzzle forces you to use, not on the clue count. Many 17-clue puzzles solve with basic scanning, while some 25-clue monsters demand X-Wings and chains. (Our own Hard mode uses 23 clues, tuned for challenge rather than minimality.)

Why computer scientists adore Sudoku

Generalised Sudoku — the n²×n² version — is NP-complete, placing it in the same complexity family as the most famous hard problems in computer science. The everyday 9×9 grid is trivial for machines, but it's become a beloved teaching example because it showcases the field's classic ideas in miniature:

• Backtracking search — try a digit, recurse, undo on contradiction. A dozen lines of code will solve any newspaper puzzle.
• Constraint propagation — encode "no repeats" as constraints and let deductions cascade, exactly the way a human solver's pencil marks resolve each other.
• Elegant algorithms — Donald Knuth's "dancing links" technique for exact-cover problems turns Sudoku solving into a thing of genuine beauty, beloved of algorithm courses everywhere.

Writing a solver is a rite of passage for programming students; writing a fast one teaches optimisation better than most textbooks. And generating good puzzles is harder than solving them — a story we tell in How Sudoku Puzzles Are Made.

The labyrinth under the newspaper page

There is something quietly wonderful about all this: cosmic-scale numbers, an 800-processor-year proof, and NP-completeness, all hiding beneath a puzzle printed next to the horoscopes. Every time you pencil in a digit, you're threading a path through a labyrinth with billions of dead ends — and finding, with nothing but patience and logic, the single route that works.