Every respectable published Sudoku comes with an unwritten guarantee: the puzzle has exactly one solution. Most solvers absorb this convention without ever examining it — yet it shapes what counts as a fair puzzle, it can be broken in instructive ways, and it powers a whole family of advanced techniques whose legitimacy solvers still argue about. This guide examines the promise itself.
Proper puzzles
A puzzle is called proper (or valid) if it admits exactly one completed grid. Too few or badly placed givens and multiple solutions exist; contradictory givens and none do. Uniqueness is not an aesthetic nicety — it is what makes Sudoku a logic puzzle. If two solutions existed, then at some point no amount of deduction could decide between them: the remaining cells would be genuinely undetermined, and “solving” would require an arbitrary choice. Deduction works precisely because exactly one grid is consistent with the clues, so every correct inference is a discovery about the answer rather than an answer.
Uniqueness is also expensive to provide. A generator cannot simply scatter clues; it must verify, essentially by exhaustive solving, that one solution exists — and re-verify after every clue it removes. The mathematics of how few clues can do the job is surprisingly deep: seventeen is the proven minimum, a story told in The Mathematics of Sudoku.
Deadly patterns: how uniqueness breaks
The simplest way to destroy uniqueness is the unavoidable set. Take a finished grid and find four cells at the intersections of two rows and two columns, all four inside just two boxes, holding two digits in an A-B / B-A arrangement. Swap the two digits around the rectangle and every row, column, and box involved still contains both digits exactly once — you have produced a second valid grid that differs only in those four cells.
Now the key inversion: if a puzzle left all four of those cells unclued, nothing could ever distinguish the two arrangements, and the puzzle would have (at least) two solutions. Such a configuration is called a deadly pattern. A proper puzzle must place at least one given inside every unavoidable set — this “hitting set” idea is exactly the machinery behind the 16-clue impossibility proof. The two-digit rectangle is merely the smallest deadly pattern; larger ones abound, and every one of them is a place where the setter was forced to spend a clue.
Techniques that exploit the promise
Here is where it gets philosophically interesting. If you trust the setter's guarantee, you can use it as a premise in deductions. The classic is the Unique Rectangle: suppose four cells form the rectangle geometry above, three of them are marked exactly {2, 5}, and the fourth is marked {2, 5, 8}. If that fourth cell were 2 or 5, all four cells would resolve into the deadly A-B / B-A swap — two solutions, contradicting uniqueness. Since the puzzle is proper, the fourth cell must be 8. A striking amount of information from a single assumption.
A family of related moves follows the same template — Unique Rectangle variants, the Bivalue Universal Grave (BUG), avoidable rectangles: assume this candidate, derive a deadly pattern, conclude the opposite. All of them convert the setter's promise into eliminations.
The controversy
Some solvers refuse to use uniqueness techniques, and their objection is worth understanding because it clarifies what “solving” means.
- The purist position: a solution should be provable from the rules and the givens alone. “The setter promised uniqueness” is a fact about the publisher, not the grid. A uniqueness deduction applied to an improper puzzle can produce a false “solution” while ordinary techniques would simply stall — the technique silently assumes what it should be checking. There is also an aesthetic version of the objection: such moves prove this cell must be 8 or the puzzle is broken, which is subtly different from this cell must be 8.
- The pragmatist position: the uniqueness guarantee is as much a rule of the game as “one digit per row.” Published puzzles state one solution exists; reasoning from stated premises is exactly what logic is. Computer raters and most competitive solvers happily count uniqueness moves among legitimate techniques.
Both positions are coherent — they simply disagree about whether the meta-rule is part of the puzzle. It is a rare example of a genuine foundational dispute inside a pastime, and where you land is a matter of taste. (Worth noting: every technique taught on this site's ladder, from Naked Singles through Swordfish, is uniqueness-free — the eliminations are locally provable and valid on any grid, proper or not.)
What uniqueness means for guessing
The promise also reframes the ethics of guessing. In a proper puzzle, bifurcation — pick a bivalue cell, try one candidate, backtrack on contradiction — always works eventually; it is how computers solve. Most human solvers avoid it not because it fails but because it abandons the pleasure of deduction for bookkeeping. Precision Sudoku's player is deliberately elimination-first for this reason: progress comes from removing candidates you can justify removing, and cells resolve when justification is complete. If you find yourself wanting to guess, the honest fix is a bigger toolbox — the technique ladder tells you which rung to train next, and the technique picker will serve puzzles that need exactly that rung.
The setter's side of the bargain
Uniqueness is half of an implicit contract. The setter warrants one solution reachable by logic; the solver, in exchange, gets to trust every deduction. The other half of the contract is difficulty honesty — that “hard” means something — and that half is far shakier in practice, as we explore in What Makes a Sudoku Hard? When a puzzle frustrates you, it is worth remembering which promises were actually made: one solution, always; a fair path to it, usually; a path matching the label on the tin, sometimes.