Skip to content
VEYRAPLAYPlay
TheoryAdvanced

Minimum Clues in Sudoku

Learn the minimum number of clues possible in a standard 9×9 Sudoku with a unique solution and why the answer is 17.

For a standard 9×9 Sudoku that must have exactly one solution, the proven minimum number of starting clues is:

17

Unique 17-clue puzzles exist.

A unique 16-clue classic Sudoku does not.

That second statement was established by an exhaustive computational search by Gary McGuire, Bastian Tugemann, and Gilles Civario.

The result answers a very precise mathematical question. It does not mean that 17 clues produce the best puzzles, the hardest puzzles, or the puzzles VeyraPlay should prefer generating.

What counts as a clue?

A clue is a digit present before the player starts.

Also called a given.

A 17-clue puzzle begins with:

  • 17 solved cells;
  • 64 empty cells.

The completed solution still contains 81 digits.

Clue count describes the starting state, not the size of the solution.

Why uniqueness is part of the question

If uniqueness did not matter, we could remove many clues and leave a starting state with enormous numbers of completions.

That would not answer the classic Sudoku minimum-clue problem.

The actual question is:

What is the smallest clue set that still determines one and only one valid completed classic 9×9 grid?

This couples two requirements:

  1. the starting clues are consistent with a solution;
  2. no second solution exists.

Why 17 was not obvious

Researchers and puzzle collectors had found valid 17-clue unique puzzles.

For a long time, no one had found a 16-clue example.

But:

“Nobody has found one”

is not the same as:

“None exists.”

The challenge was proving that an undiscovered 16-clue puzzle could not be hiding somewhere in the immense Sudoku search space.

The no-16-clue result

McGuire, Tugemann, and Civario converted the problem into an exhaustive computational search involving unavoidable sets and hitting sets.

At a high level:

  1. start from completed Sudoku solution grids;
  2. identify structures that a clue set must intersect if the resulting puzzle is to remain unique;
  3. treat the candidate clue positions as a hitting-set problem;
  4. enumerate possible small hitting sets efficiently;
  5. test the 16-clue possibilities exhaustively;
  6. find no unique 16-clue puzzle.

That establishes the lower bound:

Every unique classic 9×9 Sudoku requires at least 17 clues.

The proof is computational.

It is not a short hand proof based simply on counting cells or units.

What is an unavoidable set?

For a particular solution grid, an unavoidable set is a collection of cells with the property that a unique puzzle based on that solution must contain at least one clue from the set.

Why?

Because if all cells from that set were left unclued, an alternative completion could survive.

A valid clue set must therefore “hit” every relevant unavoidable set.

This is where the hitting-set formulation enters the proof.

For the public page, this level of intuition is enough; the paper contains the computational details.

Minimum vs minimal

This distinction matters.

Minimum-clue puzzle

Uses the global smallest possible clue count.

For classic unique 9×9 Sudoku:

17 clues.

Minimal puzzle

Every individual clue is necessary for that particular puzzle.

If you remove any clue, uniqueness is lost.

A minimal puzzle may contain:

  • 18 clues;
  • 24 clues;
  • more.

“Minimal” does not mean “contains 17.”

It means:

No clue is redundant for uniqueness.

Is every 17-clue puzzle automatically minimal?

A 17-clue unique puzzle cannot lose a clue and remain a unique classic Sudoku, because that would produce a 16-clue unique puzzle — which the minimum result rules out.

So every valid unique 17-clue puzzle is necessarily minimal.

The converse is false:

A minimal puzzle does not need to have 17 clues.

Can I choose any 17 cells from a solution grid?

No.

Most arbitrary 17-cell subsets will not produce a good unique puzzle.

They may:

  • allow multiple solutions;
  • create no solution if the starting digits are not taken consistently;
  • fail the intended human-solving model;
  • produce an undesirable or unsupported difficulty.

The global lower bound says that some 17-clue unique puzzles exist.

It does not say that 17 clues are generally sufficient in arbitrary positions.

Does 17 clues mean Expert?

No.

Clue count is a poor stand-alone measure of human difficulty.

Difficulty also depends on:

  • clue placement;
  • candidate structure;
  • required techniques;
  • number of steps;
  • dependency depth;
  • recognition burden.

A sparse grid can expose simple logic.

A denser grid can hide a difficult chain.

That is why Minimum Clues and Difficulty Rating are separate Theory pages.

Could a 16-clue puzzle have multiple solutions?

Of course.

The theorem rules out unique 16-clue classic Sudokus.

A 16-clue starting state can still:

  • have multiple solutions;
  • have no solution;
  • be a partial valid-looking grid.

It simply cannot determine exactly one classic solution.

Why this matters to a generator

It mostly tells the generator what not to optimize for.

Normal VeyraPlay generation should target:

unique
+ human-solvable
+ calibrated difficulty
+ varied
+ reproducible

not:

fewest clues possible

A generator chasing extreme clue reduction may spend significant compute producing puzzles that are mathematically sparse but not especially enjoyable.

Clue count remains useful metadata.

It is not the product goal.

Research vs gameplay

The 17-clue result is fascinating because it describes a hard boundary in the combinatorial structure of Sudoku.

For everyday play, the better question is:

Does this exact clue arrangement create a fair, interesting solving path?

Those are different standards of quality.

FAQ

What is the fewest clues in a unique classic 9×9 Sudoku?

17.

Are there unique 16-clue Sudokus?

No. The 16-clue case was ruled out by exhaustive computation.

Is every 17-clue Sudoku hard?

No.

Is every 17-clue unique puzzle minimal?

Yes, because removing one clue would create an impossible 16-clue unique puzzle.

Is every minimal Sudoku a 17-clue puzzle?

No.

Why can't clue count rate difficulty?

Because the positions and resulting deduction structure matter much more than the raw count alone.

What to learn next

Read Unique Sudoku Solutions for the global one-solution property.

Read How Sudoku Puzzles Are Generated for the practical clue-removal pipeline.

theory · advancedUnique Sudoku SolutionsUnderstand what it means for a Sudoku puzzle to have exactly one solution, how software verifies uniqueness and why the property matters.theory · advancedHow Sudoku Puzzles Are GeneratedLearn how Sudoku generators build solution grids, remove clues, verify uniqueness, analyze human solving paths and target puzzle difficulty.theory · advancedHow Sudoku Difficulty Is RatedLearn why Sudoku difficulty is not determined by clue count alone and how technique levels, solving effort and player data can produce better ratings.theory · advancedSudoku and MathematicsExplore the mathematics behind Sudoku, including Latin squares, constraint satisfaction, solution-grid counting, symmetry and search algorithms.