Locked Candidate Sudoku: Pointing and Claiming, Worked Through

A locked candidate sudoku deduction happens when every remaining spot for one digit inside a unit is squeezed into the intersection that unit shares with a second unit. Because the digit has to land somewhere inside that intersection, it can be erased from the rest of the second unit. There are exactly two flavors: pointing, where a box locks a digit onto a line, and claiming, where a line locks a digit into a box.

This is the first technique most solvers learn that removes candidates rather than places digits, and it is the highest-value one per minute spent. It needs no rectangles, no chains, and no bi-value cells — just a box and a line that overlap in three cells.

Below: precise definitions, three fully worked examples with before-and-after candidate grids, a search routine, and the near-misses that trip people up.

What a locked candidate actually is

A locked candidate is a digit whose surviving candidate cells inside one unit all sit within a single intersection — the three cells where a 3×3 box crosses a row or a column.

Every box on a 9×9 grid overlaps three rows and three columns, and each of those overlaps is exactly three cells wide. Box 1 (rows 1–3, columns 1–3) shares R1C1–R1C3 with row 1, R2C1–R2C3 with row 2, and so on. There are 54 such intersections on the grid, and each one is a doorway between two units.

The logic runs in one sentence. If a digit must appear somewhere in unit A, and every place it could still go in unit A lies inside the A∩B intersection, then that digit is going to be placed inside B as well — so nowhere else in B can hold it.

Which unit is A and which is B gives you the two types.

Both are the same idea seen from opposite ends. Learning them as two separate techniques is a mistake that costs you half your finds, because if you only ever scan boxes you will walk straight past every claiming pattern on the grid.

Type 1: pointing, where a box aims a digit down a line

Pointing works when all the candidates for a digit inside one box fall in a single row or column — the digit then gets erased from the rest of that line, everywhere outside the box.

Here is box 1 on a stalled grid. R1C1 and R2C2 are givens; the other seven cells carry pencil marks.

Before — box 1:

Track the digit 4. It appears twice in this box: R2C1 and R2C3. Nowhere else. Both of those cells are in row 2.

Box 1 must contain a 4 somewhere — every box holds all nine digits. So the 4 for box 1 is at R2C1 or R2C3, and either way it lands in row 2. Row 2 therefore has its 4 inside box 1, and cannot have another one anywhere else along its length.

Now look at the rest of row 2:

Before — row 2 outside box 1:

Erase the 4 from R2C4, R2C5 and R2C8.

After:

R2C8 drops to a single candidate and becomes an 8. That placement was not available before: the 8 in row 2 had two homes, R2C3 and R2C8, so no hidden single existed. The locked candidate is what produced it.

And it keeps going. With R2C8 = 8, row 2's 8 is spoken for, so R2C3 loses its 8 and becomes {4, 6} — a fresh bi-value cell for the next deduction.

Pointing on a column works identically. If box 6 (rows 4–6, columns 7–9) has the digit 9 only at R4C7 and R6C7, then box 6's 9 lives in column 7, and you can erase 9 from R1C7, R2C7, R3C7, R7C7, R8C7 and R9C7 — six cells from one observation.

Note the vocabulary trap: people say "pointing pair," but three cells work exactly as well. If the digit sits at R4C7, R5C7 and R6C7 and nowhere else in the box, that is a pointing triple and the eliminations are identical. Do not skip a lock just because it has three cells instead of two.

Type 2: claiming, where a line pushes a digit into a box

Claiming works in the other direction: when all the candidates for a digit inside one row or column fall inside a single box, the digit gets erased from the rest of that box.

Take row 7. After scanning it, the digit 6 has exactly two possible homes: R7C4 and R7C6. Both of those sit in box 8 (rows 7–9, columns 4–6).

Row 7 must contain a 6. That 6 will be at R7C4 or R7C6, and either way it lands inside box 8. Box 8 has now had its 6 claimed by row 7, and the other cells of box 8 can give theirs up.

Before — box 8 (R8C5 and R9C5 are givens):

After — 6 removed from every cell of box 8 outside row 7:

Four candidates gone. That is a decent return on ten seconds of looking, but the second-order effect is better. Check the digit 7 in box 8 now. Before the elimination it lived at R8C4 and R8C6 already — but look at what the grid has become: 7 still appears at exactly R8C4 and R8C6, both in row 8.

That is a pointing lock, created inside the box you just cleaned. Row 8's 7 is inside box 8, so 7 can be erased from R8C1, R8C2, R8C3, R8C7, R8C8 and R8C9.

Claiming feeding pointing feeding a placement is the ordinary rhythm of a medium-to-hard grid. It is also the reason the advice "recheck the box you just changed" is worth more than any exotic technique.

Pointing versus claiming, side by side

The two types differ only in which unit does the constraining — and you find them with two different scans, which is why solvers who learn just one leave half the eliminations on the board.

A useful mental image: pointing takes a digit that is trapped in a box and aims it along a line. Claiming takes a digit that is trapped in a line and stakes a claim on one box. Same intersection, opposite arrows.

How to find locked candidate sudoku patterns quickly

Scan by digit, not by cell — pick one number, look at where it can still go, and the locks announce themselves.

The routine that works:

1. Fill in your candidate grid completely first. A locked candidate found against stale pencil marks is not a locked candidate, and acting on it will wreck the puzzle beyond repair.

1. Choose a digit — say 4 — and mark every cell that still holds it. Nine digits, nine sweeps. Most solvers do this far too slowly at first and then find it takes about twenty seconds per digit.

1. Box pass (pointing). For each of the nine boxes, ask whether that digit's marks all fall in one row, or all in one column. Two or three cells in a straight line inside a box is the whole signature.

1. Line pass (claiming). For each of the nine rows and nine columns, ask whether that digit's marks all fall inside one box. Two or three cells inside a single three-column span is the signature.

1. Erase, then restart at singles. Every locked candidate elimination can produce a naked single or a hidden single somewhere else. Go back to the top of your sequence before hunting for anything harder.

Two accelerators worth knowing. First, boxes that already hold five or six givens are where pointing concentrates — a nearly full box has few places left for anything. Second, if you have just placed a digit, re-scan that digit across the whole grid immediately, because a placement is the most common trigger for a new lock.

The three mistakes that break grids

Almost every failed locked candidate comes down to one of three errors, and all three are avoidable by stating the deduction out loud before erasing.

• Erasing from the wrong unit. Pointing removes the digit from the line, outside the box. It does not remove it from the rest of the box — the digit still has to go in the box, you just do not know which of the two or three cells yet. This is the single most destructive slip.

• Missing a cell in the intersection. If the digit also appears at a third cell in the box that is not on your line, there is no lock. All candidates in the unit must be inside the intersection, with no exceptions. Count them; do not eyeball them.

• Working from an incomplete candidate grid. If you have only pencil-marked the cells you found interesting, a digit will look locked when it is not. This technique demands full notation.

The check that prevents all three: say the sentence. "Box 1's only 4s are at R2C1 and R2C3, both in row 2, so row 2's 4 is inside box 1, so no other cell in row 2 can be 4." If the sentence does not come out cleanly, you do not have a lock.

Where locked candidate sudoku sits in the ladder

Locked candidates belong immediately after singles and before every subset and pattern technique, because they cost the least and fire the most often.

A sensible order of attack, easiest to hardest:

The next technique to learn is naked pairs. It is the natural follow-on: locked candidates work on one digit across many cells, naked pairs work on two digits inside two cells, and together they will finish most puzzles rated hard. After that the ladder moves into single-digit patterns, starting with the X-Wing.

If you want grids to practice on, our free puzzle maker generates unlimited classic 9×9 sudoku at easy, medium and hard, prints one to six per page with an answer key, and verifies that every puzzle has exactly one solution.

Frequently asked questions

What is a locked candidate in sudoku?

A locked candidate is a digit whose remaining candidate cells inside one unit are all confined to that unit's intersection with a second unit — the three cells where a box crosses a row or column. Because the digit must be placed somewhere in the first unit, and every option lies inside the shared intersection, it must also end up inside the second unit. Every other cell of that second unit can drop the digit from its candidate list.

What is the difference between pointing and claiming?

Direction. Pointing starts from a box: if all of a digit's candidates in that box sit in one row or column, you erase the digit from the rest of that line outside the box. Claiming starts from a line: if all of a digit's candidates in that row or column sit inside one box, you erase the digit from the rest of that box. Same intersection, opposite arrow, and you need two separate scans to catch both.

Is a pointing pair the same as a locked candidate?

A pointing pair is one case of a locked candidate — specifically the pointing type with exactly two candidate cells. The lock works identically with three cells, which is called a pointing triple, and the claiming type covers the reverse direction. "Locked candidate" is the umbrella term for all of them, which is why solving software usually reports it under that name.

Does the locked candidate sudoku technique ever place a digit directly?

Not by itself. It only removes candidates. What it does constantly is create placements elsewhere: a cell that drops to one candidate becomes a naked single, or a digit left with one home in a unit becomes a hidden single. That is why the routine after every locked candidate elimination is to return to singles before looking for anything more advanced.

How hard are puzzles that need locked candidates?

Most puzzles rated medium and a good number rated hard cannot be finished without them, while easy puzzles almost never require one. Locked candidates are the first genuine threshold in sudoku difficulty — the point where scanning stops being enough and you have to keep full pencil marks. Puzzles at expert or evil level assume you use them fluently and are built to require something harder on top.

Can a locked candidate exist without a full candidate grid?

You can occasionally spot one by scanning, especially in a box that is nearly full. But you cannot trust it. The technique's guarantee depends on knowing that a digit has no other home in the unit, and partial pencil marks cannot establish that. Complete the notation first; the ten minutes it costs is repaid by every technique above singles.