How to Beat Expert Sudoku: Where Scanning Stops Working

Most players arrive at Expert with the habits that beat Hard - scan faster, look harder, stare longer - and get nowhere. The reason is not effort. An Expert grid is built so that the scanning method runs out of moves early, on purpose, and no amount of looking harder brings them back.

This page is one real Expert board, worked from the first digit to the point where it opens. Everything below was measured on that board rather than asserted: how far scanning gets, where it dies, and exactly which move restarts it.

The board

Watch the scanning die

Start the way you always start: hunt for cells and digits that are already forced. On this grid that method places three digits and then has nothing left. Three, out of fifty-six empty cells. Fifty-three cells sit there with no forced move anywhere on the board.

That is the single most useful thing to know about Expert. When the board goes quiet after a handful of placements, nothing has gone wrong and you have not missed an easy one. The puzzle was built to arrive here.

What it means in practice is that the first phase of an Expert solve is not solving at all. It is writing the board out: a full pass of pencil marks, every candidate in every empty cell, before you try to place anything else. The moves that break Expert grids are properties of the candidate layout, so they literally do not exist until the layout is on the paper. Snyder notation is the lighter version if a full grid of marks feels like too much.

The move that restarts it

With the marks down, stop asking what goes in a cell. Ask where a digit can go in a unit. Here is the question that opens this board, and it is about as simple as this level gets.

Read that again as a method, because the method is the point. The 5 in the bottom-middle box is confined to two cells, and both of them lie in the same row. Wherever it ends up, that row's 5 is inside the box - so every other cell in the row loses its 5. One of those cells was down to 1 or 5, and with the 5 gone it is a 1.

This is a pointing pair: a digit trapped on one line inside a box, pointing out along that line. Its mirror image, box-line reduction, works the same way in the other direction.

That one digit is worth sixteen more cells. Placing the 1 lets the plain scanning method run again, and it does not stop until sixteen further cells are filled. The grid went from dead to two thirds solved on the strength of a single eliminated candidate.

Not every elimination is worth making

This is the part guides usually skip. At that same stalled position the board offers several perfectly valid eliminations, and most of them are worthless. We measured every one of them:

• A pointing pair on the 9 in the top-left box - eliminates candidates correctly, places nothing.
• A pointing pair on the 5 in the top-middle box - correct, and again places nothing.
• A pointing pair on the 8 in the bottom-middle box - correct, places nothing.
• The pointing pair above, on the 5 - sixteen cells.

Finding a valid pattern is not the skill. Finding the one that lands is. And there is a reliable way to aim: work backwards from the cells that are nearly decided. A cell holding two candidates only needs one of them removed to fall, so scan your marks for the two-candidate cells first, ask what would kill one of their digits, and go hunting for that specific elimination. On this board the amber cell was the target and the pointing pair was simply the tool that reached it.

The same conclusion, the expensive way

There is also an X-Wing on this board at the same moment: the 5 in the fourth column is down to two cells, the 5 in the sixth column is down to two cells, and all four sit in the same pair of rows. That pattern eliminates the 5 from four cells - and two of those four are the same cells the pointing pair struck. It reaches the identical breakthrough.

Two things follow, and both are worth carrying to every Expert grid:

• The cheapest technique that reaches the elimination is the right one. A pointing pair takes a second to see; an X-Wing takes a scan of the whole grid. Look for the cheap ones first and you will rarely need the expensive ones.
• Half of a technique's eliminations are usually dead weight. The X-Wing here strikes four cells and only two of them do anything. Do not judge a pattern by how impressive it looks.

What to do, in order

1. Scan first anyway. It is cheap, and on this board it did place three digits. Just stop the moment it dries up instead of grinding.
2. Full pencil marks. Every empty cell, every candidate. This is not optional at Expert - it is the phase that makes the rest possible.
3. List your two-candidate cells. These are the pressure points. One elimination each and they fall.
4. Interrogate units, not cells. "Where can this digit go in this box, this row, this column?" is the question that finds pointing pairs, box-line reductions and pairs.
5. Only then go looking for a fish. X-Wings and the rest are real and sometimes necessary, but they are the last resort, not the first move.
6. Re-scan after every breakthrough. One elimination bought sixteen cells here. Cash it in before hunting for the next pattern.

How much of this does one grid need?

After the cascade this board still had thirty-seven cells empty, and the plain method had stopped again. That is normal: an Expert grid usually needs a handful of these breakthroughs, not one. What separates it from Evil is mostly how often you have to find one.

What you never need is a guess. Every board here is checked at generation time to be solvable by the techniques the School teaches, so a position that looks impossible is a position with a move you have not found - never a coin to flip.

Where to practise it

Reading a breakthrough and finding one are different skills, and only the second one counts. Play an Expert board and give yourself permission to spend the first few minutes writing marks rather than placing digits - that is what the level is asking for. About Expert Sudoku covers what the level is made of, and techniques by difficulty maps which patterns first pay off where.