Solving extreme classic sudoku

At extreme, single-cell techniques have stopped being enough. Naked singles run out within the first three or four placements; hidden singles dry up not long after. The rest of the solve depends on patterns that span multiple cells across multiple regions, and the route from one third of the puzzle solved to the other two thirds runs through unique rectangles, X-wings, swordfish, and forcing chains. These aren't decorative — they're load-bearing. You won't finish an extreme puzzle without at least three of them, and the larger ones can take 30 minutes or longer.

Pencil marks first

Pencil marks are not optional at extreme. The full candidate set in every empty cell — every digit that hasn't been ruled out by the row, column, or box — is the working memory you need to spot anything more involved than a basic pair. Most solvers fill in the candidate notes for the entire grid before placing more than a handful of digits. The notation looks busy, but it's the surface the advanced techniques operate on; without it you're flying blind. Our piece on reading pencil marks like a shape covers the visual habit that turns the busy grid back into something readable.

The four techniques you'll see every puzzle

Unique rectangles exploit the puzzle's uniqueness — the guarantee that a valid sudoku has exactly one solution. When two digits could form a "deadly pattern" across four cells in two boxes, accepting both candidates in those four cells would imply two valid solutions, which the puzzle's uniqueness rules out. The technique works backward from that contradiction — one of the candidates has to be eliminated. Unique rectangles are the workhorse advanced technique at extreme; spotting them gets faster with practice and they fire often enough to save real time on most puzzles.

X-wings and swordfish are the same idea applied to single digits across rows or columns. An X-wing is when a digit can only appear in two cells of two rows (or two columns) — that constraint forces the digit into the four corners of the rectangle and eliminates it from any other cell in those columns or rows. Swordfish is the same pattern at three rows or columns instead of two. Our deep-dive piece on why the X-wing keeps tripping people up is worth reading if the pattern hasn't clicked yet — it's the technique most solvers miss most often, and the article walks through where the recognition pattern actually lives.

Forcing chains are the general technique behind unique rectangles and many of the more advanced patterns. Pick a candidate that can only be placed in two cells of a region, and trace what happens to the rest of the grid in each of the two cases. If both cases lead to the same conclusion at some other cell — that cell must hold a particular digit, or that candidate must be eliminated from a third cell — the conclusion stands regardless of which case is actually true. Chains are slower than pattern-spotting, but they're available when the patterns aren't, and following one to a forced placement is the most reliable way to break a stalled extreme puzzle.

Stalling is the default state

Extreme is the tier where stalling is normal. Most solvers hit at least one wall on every puzzle, and the wall is rarely "I haven't seen the right pattern yet" — it's "I haven't looked at the right configuration yet." The fix is structural rather than tactical.

When you stall, the first thing to do is re-pencil-mark the grid. Half-an-hour into a solve, the candidate notes drift; cells that should have been narrowed haven't been, and cells that have been narrowed are missing the elimination that would have triggered the next move. A clean re-pencil pass takes two minutes and frequently surfaces the move that was already there.

The second move is to systematically check for each technique in turn. Walk every row for X-wings, every column for X-wings, every box pair for unique rectangles. The technique you're missing is usually one of the easier patterns rather than something exotic — the conscious technique-by-technique scan finds it.

The third move, if those don't work, is to abandon the puzzle for half an hour and come back. Extreme rewards fresh attention more than persistence. The patterns are often visible the moment you see the grid again.

Two habits to drop

Half-pencil-marking. Filling in candidate notes for some cells but not others is the failure mode unique to extreme. At lower tiers it's reasonable to skip notation for cells that look "about to resolve." At extreme, every empty cell needs full candidate marks, because the techniques that fire here often span cells that look unrelated. A bilocation strong link between two cells in opposite corners of the grid only becomes visible when both cells have their full candidate sets visible.

Bashing on a stall. When a puzzle has stalled for fifteen minutes, the move that's there isn't going to come from staring harder. Either re-pencil-mark, technique-by-technique scan, or take a break. Extreme puzzles aren't a battle of attention; they're a search problem, and search rewards method over endurance.

There isn't really a higher tier than extreme. Most generators top out here, and the few that publish "evil" or "inhuman" tiers above mostly bump the difficulty by reducing givens, not by introducing new techniques. The graduation, if there is one, is from solving extreme in 30 minutes to solving it in 15. Speed at this tier is its own discipline, and most solvers stay at extreme for years without ever feeling they've exhausted it. The puzzle keeps having something new to show.