Every Sudoku puzzle can be solved with pure logic — no guessing required. The techniques below are organized by skill level, from the scanning habits you'll use on every easy puzzle to the pattern-recognition tricks reserved for evil-rated grids. Work through them in order, or jump straight to the level that matches your current puzzle.
These four techniques cover almost everything you need for easy and most medium puzzles. They rely on scanning the grid methodically rather than deep deduction, so they're the fastest skills to build.
Pick one digit and scan every row and column it already appears in. Since that digit can't repeat anywhere in those lines, you can mentally cross out the cells they pass through. Do this for a single 3×3 box and you'll often find only one open cell remains for that digit — an instant placement. This is usually the first move on any fresh puzzle.
Rather than solving cell by cell, write every digit that could still legally go in each empty cell. This turns the puzzle into a visible map of possibilities instead of something you have to hold in your head. Most of the techniques further down this list depend on having accurate pencil marks in place first — it's less a "technique" and more the foundation everything else is built on.
Once pencil marks are filled in, look for any cell that has exactly one candidate left. There's nothing to deduce — if every other digit has been ruled out by the row, column, and box, the remaining one has to be correct. Naked singles are the easiest wins in the entire puzzle and worth sweeping for after every few placements.
A hidden single is also the only valid home for a digit — but unlike a naked single, the cell itself still shows several candidates. The trick is to stop scanning cell-by-cell and instead scan digit-by-digit: pick a number and check whether it can only fit in one cell within a given row, column, or box, even if that cell has other pencil marks sitting alongside it.
Once singles dry up, these pairing and grouping techniques let you eliminate candidates in bulk — essential for breaking through medium and hard puzzles.
When two cells in the same unit share the exact same two candidates and nothing else, those two digits are locked into those two cells — you just don't know the order yet. Since no other cell in that unit can hold either digit, you can safely erase both candidates from every other cell sharing that row, column, or box.
The mirror image of a naked pair: two digits that only appear as candidates in the same two cells of a unit, even though those cells are cluttered with other pencil marks too. Once you spot the pair, every other candidate in those two cells can be stripped away, leaving just the hidden pair behind.
This isn't about a cell having one candidate — it's about a unit having one open cell left for a given digit, after accounting for every digit already placed nearby. It often surfaces almost by accident while you're applying cross-hatching, but it's worth deliberately checking for once a row, column, or box is mostly filled.
Three cells in one unit, between them, contain only three distinct candidates — though no single cell needs to show all three. Because those three digits must distribute across exactly those three cells, every other cell in the unit can have those three candidates removed, even cells that only shared one of the three digits.
Three specific digits appear as candidates only within three cells of a unit — but those cells may also carry unrelated candidates. Spotting which three cells "own" the trio lets you clear out every other candidate in those cells, narrowing them down toward a hidden triple.
These techniques look for geometric patterns across the whole grid rather than within a single unit — the jump in difficulty is less about the logic itself and more about training your eye to spot the shape.
Find a digit that appears as a candidate in exactly two cells in each of two different rows, and those four cells line up into the same two columns — forming a rectangle. Since the digit must land in one of the two cells per row, it's also locked into those two columns overall, so it can be cleared from every other cell in those columns.
The same idea as an X-Wing, scaled up to three rows and three columns instead of two. The candidate doesn't need to appear in all three cells of every row — two or three is enough — as long as every occurrence across the three rows stays confined to the same three columns. Once confirmed, the candidate can be eliminated from those columns everywhere outside the pattern.
Look for a candidate appearing in exactly two cells across two different rows, where one pair shares a column (the "floor") and the other pair sits in two separate columns (the slanted "roof"). Any cell that can see both roof cells has that candidate eliminated, since at least one of the two roof cells must hold it.
Every well-formed Sudoku has exactly one solution, which means certain candidate patterns simply can't be allowed to exist. If four cells form a rectangle spanning two rows, two columns, and two boxes, and all four share the identical pair of candidates, the puzzle would have two valid solutions — which is impossible. That contradiction is the lever: it tells you one of those candidates has to be wrong, letting you eliminate it and collapse the rectangle.
Sometimes a candidate within a single box only appears along one row or column of that box. Since the digit must end up somewhere inside the box, it has to land in that row or column — so it can be removed from the rest of that row or column outside the box (a "pointing" lock). The reverse also holds: if a candidate in a row or column only appears inside one box, it can be cleared from the rest of that box (a "claiming" lock).
At this level you're chaining logical implications across multiple cells rather than reading a single static pattern. These techniques are reserved for evil-rated puzzles where every easier method has been exhausted.
Start with two cells that each hold the exact same pair of candidates but don't share a row, column, or box. If those two cells are connected by a strong link — a chain where one of the shared candidates appears in only two cells of some row, column, or box, with each end touching one of the two starting cells — then the other shared candidate can be eliminated from any cell that sees both starting cells.
Three two-candidate cells form this pattern: a pivot holding candidates x and y, and two pincers — one sharing x with the pivot, the other sharing y — where each pincer also shares a unit with the pivot but not with each other. Both pincers carry a third candidate z in common. Since one of the two pincers must eventually hold z, any cell that sees both pincers can have z eliminated.
A close cousin of the XY-Wing, except the pivot cell carries three candidates (x, y, z) instead of two. The two pincers still hold x/z and y/z respectively and each shares a unit with the pivot. With all three cells now sharing candidate z, it can be eliminated from any cell that sees all three — pivot included.
Pick a cell with two candidates and follow the logical chain of forced placements that would result from each one being true, one assumption at a time. If both chains independently force the same outcome somewhere else in the grid, that outcome must be true regardless of which original candidate was correct. If a chain instead loops back into a contradiction, the assumption that started it can be ruled out entirely.
An almost locked set (ALS) is a group of cells within a unit holding exactly one more candidate than cells — for example, three cells sharing four candidates total. When two separate almost locked sets share one "restricted" candidate that can only be true in one set or the other, plus a second shared candidate that isn't restricted, that second candidate can be eliminated from any outside cell that sees every instance of it across both sets.