Most people who feel stuck at Sudoku are not missing a technique. They are running a technique they already know, badly, or in the wrong order, or with a habit that quietly destroys the information they just generated. The errors below account for the large majority of stalls, and every one of them has a specific fix.
1. Guessing, and calling it intuition
This is the expensive one. You narrow a cell to two candidates, one of them “feels right”, you write it in, and you carry on. Sometimes it is right. That is the problem — a guess that happens to be correct is indistinguishable from a deduction until much later, so you never learn which of your moves were sound.
When a guess is wrong, the contradiction usually surfaces fifteen or twenty placements downstream, at which point you cannot tell where the error entered. The puzzle is now unrecoverable without erasing back to a point you cannot identify, and the experience gets filed as “this puzzle was too hard” rather than “I guessed at move thirty”.
The fix: adopt a hard rule that you write nothing you cannot justify in a sentence. If the sentence is “because 4 and 7 are the only options and 7 looks better”, do not write it. This site's board enforces the rule structurally — you remove candidates you can defend and cells resolve themselves — which is uncomfortable for about two puzzles and then becomes the only way you want to solve.
2. Restarting the scan after every placement
You place a digit, then return to the top-left corner and begin sweeping the grid again. This is the most common efficiency error in Sudoku, and it is self-inflicted: the digit you just placed eliminated a candidate from up to twenty cells, and the most likely place for the next deduction is precisely those cells. By restarting elsewhere you throw away the information you just created and pay full price to rediscover it.
The fix: after every placement, immediately check that digit's row, column and box. Follow the cascade until it genuinely dries up, and only then resume systematic scanning. On easy and medium puzzles this one change can halve solve time.
3. Mistaking an un-followed cascade for a hard puzzle
Closely related, and worth separating because it changes how people rate their own ability. A large share of “I am stuck” moments on medium puzzles are not stalls at all — there is an available Naked Single or Hidden Single somewhere on the board that simply has not been looked at. Solvers conclude they need an advanced technique, go looking for an X-Wing, fail to find one, and give up on a puzzle that never required anything beyond singles.
The fix: before deciding a puzzle needs something harder, run one complete, disciplined pass — cross-hatch all nine digits through all nine boxes, then all nine rows, then all nine columns, without shortcuts. If that pass is genuinely empty, the puzzle really does need more. Usually it is not empty.
4. Cross-hatching only the digits you like
Scanning tends to drift toward whichever digits are visually salient, and away from the ones that are sparse on the grid. But the productive digits are the ones already placed six or seven times, because each placement projects a line, and a digit with seven copies on the board has carved the grid up thoroughly.
The fix: scan in numerical order, 1 through 9, every time. It feels mechanical, and that is the point — it removes the choice, and the choice was the source of the bias. Count each digit's occurrences first and start with the most common if you want to bias the order deliberately rather than accidentally.
5. Scanning rows but not columns
Reading habits are horizontal, and it shows in solving. Ask a stalled solver to check for Omissions and they will reliably examine rows and skip columns, then conclude the technique is not available. Roughly half of all instances of every line-based technique run vertically.
The fix: make the vertical pass an explicit, separate step rather than something you intend to do. Some solvers physically rotate the page or tilt the phone ninety degrees, which is a crude trick that works surprisingly well because it recruits the horizontal habit for the vertical direction.
6. Pencil-marking everything, immediately
Filling every empty cell with its full candidate list before making a single deduction is a standard beginner ritual, and on easy and medium puzzles it is close to pure waste. You will spend several minutes annotating a grid that would have fallen to cross-hatching in less time than the annotation took, and you now have a dense, cluttered board that is harder to read.
Worse, full marks must be maintained. Every placement invalidates marks in twenty cells, and a stale pencil mark is not neutral — it is misinformation that will eventually produce a wrong deduction you trust completely.
The fix: solve as far as singles and cross-hatching will take you, then mark up only when you genuinely need candidate sets to see subsets and fish. If you do mark, mark completely and keep it current; half-maintained marks are worse than none. The pencil-marks guide covers the trade-off properly.
7. Applying subset eliminations in the wrong direction
This is the most common logical error, as opposed to a habit error, and it produces confident wrong answers. The naked and hidden subsets act in opposite directions, and under time pressure people reverse them:
- A Naked Pair — two cells holding the same two candidates — eliminates those two digits from every other cell in the unit. The pair itself is untouched.
- A Hidden Pair — two digits that can only live in the same two cells — eliminates every other candidate from inside those two cells. The rest of the unit is untouched.
The same inversion applies at triplet and quad size. Reversing either one deletes a digit from the cell it actually belongs in, and the puzzle becomes unsolvable in a way that looks like a contradiction much later.
The fix: before acting, say the deduction out loud in the form “these digits are locked into these cells, therefore…”. Whichever side of the sentence is the locked pair keeps its digits; the other side loses them. The subsets guide derives both directions from first principles, which is a more durable fix than memorising which is which.
8. Requiring patterns to look tidy
Textbook diagrams show a Naked Triplet as three cells each displaying the same three candidates, and a Swordfish as a clean three-by-three arrangement. Real boards almost never look like that. A perfectly valid Naked Triplet can be {1,4}, {4,9} and {1,9} — no cell showing all three digits. A genuine Swordfish usually has two candidates in some rows and three in others, so it looks ragged and asymmetric.
The fix: verify by counting, not by shape. For subsets, take the union of the candidates across the cells and check that it has the same size as the number of cells. For fish, count how many distinct columns the rows use. Both are arithmetic checks that do not care what the pattern looks like.
9. Practising by volume instead of by pattern
The default improvement strategy is to solve lots of puzzles and hope. It works, slowly, because exposure is exposure — but it is badly targeted. If you cannot see Hidden Pairs, solving fifty random puzzles might present you with six, spaced far enough apart that no recognition ever consolidates, and unlabelled, so you may not notice when you miss one.
The fix: drill the specific pattern in isolation, with feedback, until it is boring — then return to mixed puzzles so the recognition has to compete with everything else on the board. That is the entire premise of the technique practice pages: every puzzle in a set is solver-verified to require the named technique, so you are guaranteed the exposure. When a rated puzzle elsewhere defeats you, identify which step you missed, drill that one, and move on. Difficulty stops being a property of puzzles and becomes a curriculum.
The pattern behind the pattern
Read back over the nine and a theme emerges: almost all of them are failures of systematic process, not failures of knowledge. Scanning in a fixed order, following cascades to exhaustion, checking both axes, verifying subsets by counting — none of this requires learning anything new. It requires doing the thing you already know how to do, in the same order, every time. That is unglamorous advice, and it is where nearly all of the available improvement actually is.