The List

Character Clues

Character clues are the practice of narrowing a cell's candidates by asking what kind of digit can legally go there, not just which digits are already visible in its row, column and box. In a standard 9×9 Sudoku — 81 cells, nine rows, nine columns, nine 3×3 boxes, digits 1 through 9 — the obvious move is to scan for a digit that appears eight times in a house and drop it into the ninth cell. Character clues sit one level underneath that: they describe the shape a digit's position must take within a house, so you can eliminate candidates before any single cell resolves. The technique pays off in the middle game, when naked and hidden singles have dried up and you are staring at a grid where every empty cell has two or three pencil marks and none of them look wrong.

1 article

Character Clues

Character Clues

Newest first
Identifying a scoundrel using crossword clues
Character Clues

Identifying a scoundrel using crossword clues

The most reliable approach to solving any highly structured puzzle, whether it's tracking down a miscreant in a fictional mystery or filling an entire Sudoku grid, is pure deductive logic.

1 SepOpen →

Character clues are the practice of narrowing a cell's candidates by asking what kind of digit can legally go there, not just which digits are already visible in its row, column and box. In a standard 9×9 Sudoku — 81 cells, nine rows, nine columns, nine 3×3 boxes, digits 1 through 9 — the obvious move is to scan for a digit that appears eight times in a house and drop it into the ninth cell. Character clues sit one level underneath that: they describe the shape a digit's position must take within a house, so you can eliminate candidates before any single cell resolves. The technique pays off in the middle game, when naked and hidden singles have dried up and you are staring at a grid where every empty cell has two or three pencil marks and none of them look wrong.

What a character clue actually is, and why the name fits

A character clue is a constraint on where a digit can sit inside a unit, expressed as a property of the digit's placement rather than a property of a single cell. The most familiar form is the locked candidate: in a box, all the remaining spots for a 7 might fall on one row. That gives you a character clue about the 7 — within this box it behaves as a row-dweller — and you can strike 7 from the rest of that row outside the box. The digit has a character (row-bound, column-bound, cornered, paired) and the character does the eliminating.

The reason the framing matters is that it forces you to think in units rather than cells. Most solvers learn Sudoku by filling cells. Character clues invert that: you look at one digit across one house, ask what its remaining positions have in common, and only then go back to individual cells. On a 9×9 grid with nine boxes, nine rows and nine columns, each digit has 27 possible homes per row-column-box triple, and the intersections between those houses are where the clues live. A digit can be legal in a cell by row and column and still be impossible because of what its box-mates are doing elsewhere.

The cost is attention. Scanning for character clues means holding a single digit in mind and sweeping the whole grid for it, which is slower than the cell-by-cell pass most people default to. On easy puzzles this is wasted effort — singles finish the job. On hard puzzles it is the difference between progress and a grid you abandon. The honest trade-off: character-clue hunting is the right investment from roughly the point where your pencil marks stop producing new placements, and the wrong investment on anything you can solve in a few minutes.

The three characters worth knowing first

Start with locked candidates, then conjugate pairs, then the box-line inversion of the same idea. These three cover the majority of situations where a digit's position is constrained but not yet determined, and they share a single mental motion: find the digit, find the house, find what the remaining spots have in common.

Notice what these have in common: none of them places a digit. They only remove candidates. That is the whole point. Character clues are eliminative tools, and their output is a thinner pencil-mark grid, not a filled cell. A solver who expects an immediate placement will conclude the technique failed when in fact it worked perfectly and simply handed the placement to a later, simpler step.

The limit is real. Locked candidates require the digit's remaining positions in a house to be genuinely confined; if you have missed a candidate, you will "find" a lock that does not exist and eliminate a correct candidate. This is the single most common way character-clue solving goes wrong, and it is why the discipline of keeping pencil marks complete matters more here than anywhere else in Sudoku.

Reading a grid for character before you read it for digits

Do the character pass digit by digit, not cell by cell. Pick a digit, sweep all nine boxes for its remaining positions, note any box where those positions collapse onto a single row or column, then sweep the rows and columns for the claiming version. Nine digits, two directions, one pass. It takes a couple of minutes on paper and less than that once the motion is automatic.

The order in which you check digits matters less than the discipline of checking all of them. A grid where the 2s are locked and the 8s are not will reward a solver who happens to start with 2 and punish one who starts with 8 and gives up. Sweeping the full digit range is the fix. Some solvers work high-to-low, some follow the digit that appears most often as a given; either is fine as long as you do not stop at the first digit that yields nothing.

What you are looking for has a visual signature once you have done it a few hundred times. A box where three empty cells hold the same digit's candidates, all on one line, looks different from a box where they scatter. The scattered case is not a failure — it is a digit with no character clue in that house, and you move on. The trap is confirmation bias: after finding one lock, solvers tend to see locks everywhere, including in houses where a candidate was missed. Slow down on the eliminations that feel too good.

There is a practical sequencing point here. Character clues are cheapest to find immediately after you have refreshed your pencil marks, because the marks are what you are reading. If you have been solving for twenty minutes without a full candidate update, your character scan is operating on stale data and will produce either nothing or a false lock. Refresh, then scan.

Where character clues stop being enough

Past locked candidates and conjugate pairs, character reasoning escalates into pattern techniques — X-Wing, Swordfish, XY-Wing, and the chain methods beyond them — and each step costs more working memory for a smaller return. The honest assessment: most solvers who plateau at "hard" Sudoku plateau because they never learned locked candidates thoroughly, not because they lack the exotic patterns. Get the two locked-candidate forms and conjugate pairs solid before touching anything named after a fish.

The escalation has a real cost in error rate. A missed candidate in a single cell can invalidate an entire chain of eliminations three techniques deep, and the failure is silent — the grid simply stops resolving and you cannot tell whether the puzzle is hard or you have corrupted it. This is why experienced solvers treat character-clue eliminations as provisional and re-verify marks when a puzzle stalls. The technique set is powerful; the dependency on clean data is absolute.

There is also a category of puzzle where character clues are the wrong tool entirely. Some grids are built to fall to a single well-hidden forcing chain, and hunting locks and wings will burn time without touching the critical path. You notice this when every character scan comes back empty across all nine digits — that is the signal to switch from eliminative techniques to something that reasons about a cell's two-way choice directly. Knowing when to abandon the scan is as much a skill as running it.

Character clues outside the grid, and what the research does and does not say

The same reasoning — infer a thing's character from the constraints around it rather than from the thing itself — shows up in deduction puzzles well beyond Sudoku, and in logic teaching generally. That transfer is the strongest argument for learning the technique properly rather than memorising a handful of patterns. A solver who understands why a pointing pair eliminates will recognise the structure in a Nurikabe or a Slitherlink without being taught a separate named rule for each.

On the cognitive side, the evidence is narrower than the marketing suggests. A large study on number puzzles including Sudoku and brain function in older adults assessed 14 cognitive measures in participants aged 50 to 93 years old. That is a study of association between puzzle practice and cognitive measures in a specific older population — it is not a claim that learning locked candidates does anything in particular, and it is certainly not a claim that character-clue practice staves off decline. Treat any stronger claim you encounter with suspicion. The defensible statement is that number-puzzle practice has been studied in relation to cognitive function in older adults across a defined set of measures; the mechanism and the causal direction are separate questions the study design does not settle.

What that means for how you spend your solving time: the transferable benefit, if there is one, almost certainly comes from the reasoning rather than the filling. Grinding easy puzzles with singles builds speed. Working hard puzzles with character clues builds the habit of holding a hypothesis and testing it against constraints, which is the part that resembles anything useful outside the grid. If you are choosing how to practise, choose the harder mode and accept the slower session.

Building the habit: what to practise and in what order

Learn locked candidates in both directions before anything else, and drill them on puzzles just above your comfort level. Pointing and claiming are the same insight viewed from two houses, and solvers who learn only one spend months missing half the available eliminations. Once those are automatic, add conjugate pairs and the X-Wing they generate. Stop there until your error rate on hard puzzles drops — adding techniques faster than you can verify them is how solvers end up with corrupted grids and no diagnosis.

Practise on paper with full pencil marks at least some of the time. Software that auto-fills candidates hides exactly the skill character-clue solving depends on: knowing your marks are complete and correct. When you solve on paper and a lock fails to produce a placement, you can go back and find the missed candidate. When software does it for you, the error is invisible and the lesson is lost.

The realistic trajectory: locked candidates become a reflex within a few dozen puzzles, conjugate pairs take longer because chaining them requires holding more in mind, and the named patterns beyond that are optional. Many strong solvers never go past X-Wing and simply choose puzzles that do not require it. That is a legitimate ceiling. The techniques above it cost disproportionately more practice for puzzles most people never encounter, and the character-clue habit — scan the digit, find the constraint, eliminate — is doing the real work at every level.

The one thing worth guarding against is treating character clues as a checklist to run mechanically. They are a way of looking, not a sequence of steps. The solver who scans digit by digit because that is the procedure will find fewer locks than the one who has learned to notice, while filling in a box, that three cells on a line are all holding the same digit. The procedure gets you started; the noticing is what you are actually building.

Character clues when a digit has only two homes in a box

A digit confined to two cells in a box is not yet a conjugate pair — that term belongs to rows and columns — but it behaves the same way and is worth flagging separately in your marks. Where those two cells share a row or column, the digit is locked to that line and can be struck from the rest of the line outside the box; that is the pointing form again. Where they sit on different lines, the pair still tells you something: neither cell can take any other digit that is exclusive to one of them, and if one of the two cells is later forced to another value, the digit lands in the other by default. The practical value is bookkeeping. Marking two-cell boxes as you sweep means a later placement resolves the digit without a second scan.

Why character reasoning survives a switch between puzzle types

The transferable skill is asking what a set of remaining positions has in common, and that question is not Sudoku-specific. Nurikabe and Slitherlink both punish solvers who reason from a single cell outward; both reward solvers who look at where a wall or a loop segment must run and eliminate from there. The cost of learning the Sudoku version properly is a few dozen puzzles of deliberate practice. The return is a habit you carry into any grid with a constraint structure — and one that, unlike memorised named patterns, does not need re-learning when the puzzle type changes.