Nine common Sudoku mistakes, and how to stop making them

1 September 2026 · 8 min read

Most Sudoku failures are not failures of intelligence. They are a handful of specific, repeatable process errors, and each has a concrete fix. Here they are in rough order of how much time they cost.

All of them assume you already know the rules and the basic techniques. If not, the beginner’s guide is the place to start.

1. Guessing

The most expensive mistake, and the least obvious, because a guess often works out.

The damage is not the wrong digit. It is that from the guess onwards, every deduction you make rests on an assumption you did not record. Hit a contradiction thirty moves later and you cannot tell whether the guess was wrong or your reasoning since was — so the only sound response is to unwind all thirty.

The fix. Refuse to write a digit you cannot justify in a sentence. If you are genuinely out of ideas, the honest options are to look harder with a system, or to have the deduction shown to you. Both leave the grid trustworthy.

2. Working from stale pencil marks

You place a 7. You mean to clear the 7s from its row, column and box, but you spot another move first and chase it. Those 7s are now lies sitting in your grid, and they look exactly like verified candidates.

This is the most common way a solve ends in an unsolvable position — and the technique that breaks first is usually a naked pair, because a cell that looks like {5, 7} may really be {3, 5, 7}, which makes the pair and every elimination from it invalid.

The fix. Clear the peers before you look for the next move, every single time. If you lose confidence in a unit, rebuild its candidates from the placed digits — it takes twenty seconds.

3. Checking one line and stopping

“3 can’t be in these two cells because column 7 already has one, so it goes in the third.” Except you never checked whether the third cell was ruled out by its own row.

A hidden single requires every other cell in the unit to be eliminated. Most is not enough, and “most” feels identical to “all” at speed.

The fix. Before placing a hidden single, say out loud why each other cell is excluded. If you cannot give a reason for one of them, you have not found the pattern yet.

4. Marking up the whole grid before starting

It feels productive and usually is not. Filling in every candidate takes several minutes, needs constant upkeep, and on anything below Hard the puzzle is over before it pays for itself. Worse, it hands you a grid so dense that the patterns are harder to see, not easier.

The fix. Scan first. Start marking the first time you go all the way round the grid and place nothing — that is the signal you have exhausted what you can hold in your head. Then mark one unit at a time and scan it as you go, because the noticing happens while you write. More on notation.

5. Hunting advanced patterns too early

Learning the X-wing and then looking for one on move six. It is a real technique, but nearly every position containing an X-wing also contains something far cheaper — and rectangles take minutes to find.

People who learn advanced patterns early often get slower before they get faster, purely because of search order.

The fix. Fixed escalation: full single sweep, then box/line intersections (cheap, and they need no marks), then pairs and triples, and only then anything exotic.

6. Finding an elimination and not cashing it in

You find a naked pair, make the eliminations, and go looking for another pair. Meanwhile the eliminations you just made have created a single that is sitting there unnoticed.

Pairs, triples and fish never place a digit. The only reason to make an elimination is that it might create a single.

The fix. After every elimination, immediately re-sweep for singles. Solving a hard puzzle is that loop — eliminate, sweep, eliminate, sweep.

7. Scanning the grid in reading order

Starting top-left and working across spends equal effort on the emptiest and fullest parts of the board. Forced moves are overwhelmingly in the fullest parts.

The fix. Rank by density. Work the box with seven digits before the one with two, and take the digit already placed seven times before the one placed twice. A full scanning routine.

8. Restarting instead of debugging

You hit a contradiction — a cell with no candidates, or a unit needing two of the same digit — and wipe the grid to start over.

That throws away the most useful information you have. The contradiction tells you which unit is broken. The error is almost always a digit placed without proof, or an uncleared pencil mark, in that unit or one touching it.

The fix. Rebuild the candidates for the failing unit from the placed digits and check each of its digits can still be justified. This is faster than restarting and, more importantly, it teaches you which of your habits caused it. On screen, undo makes this nearly free — step back until the contradiction disappears and look at the move you just reversed.

9. Treating difficulty labels as clue counts

“24 clues, that looks doable” is not what difficulty means. Clue count and difficulty are only loosely related — the best-known Sudoku on the internet has 30 clues, which most sites would call Hard, and the simplest technique in the game finishes it unaided.

What makes a puzzle hard is which techniques it forces you to use. Jumping to Expert because the grid looks sparse means meeting patterns you have no method for, which teaches nothing.

The fix. Treat the label as a statement about technique — on this site it literally is one — and move up when a level’s hardest required step comes quickly rather than when you can grind through it. What difficulty actually measures.

The pattern behind all of them

Seven of these nine are the same error wearing different clothes: acting on something you have not verified. An unproven digit, an uncleared mark, a partially-checked unit, a difficulty judged from the wrong signal.

Which suggests one habit that covers most of it — before each placement, be able to say the reason in a sentence. Slower per move, much faster per puzzle, and it is the difference between practising and just playing. It is also why this site explains moves rather than revealing them: an answer gets you one cell, and a named reason gets you every future instance of the same pattern.

Every diagram on this page is a real position from a generated puzzle, and every elimination shown is checked against that puzzle’s unique solution as part of the test suite. Want a move explained on your own board? Play a puzzle.