How to Solve Jigsaw Sudoku - Counting the Leftovers of Irregular Boxes

/10 min read

Jigsaw Sudoku is solved by reading how its irregular boxes spread across rows and columns, and by using the fact that, for a set of adjacent rows and the same number of boxes, the parts left over on each side hold the same digits. This article shows what you can read from the number of cells a box has in one row, the steps for counting leftovers, and how to compare the two sides to remove candidates, with tables.

Where do you start a Jigsaw Sudoku?

Jigsaw Sudoku replaces the 3x3 square boxes of standard Sudoku with nine-cell boxes of different shapes. It is also known as Irregular Block Sudoku or Irregular Sudoku. The short answer to where you start is that there are two things to look at first. One is the shape itself: how many cells each box has in each row and column. Boxes with a lopsided spread readily give eliminations of the pointing pair and box/line reduction kind. The other is a way of counting known as the Law of Leftovers: compare a set of adjacent rows with a set of boxes, and the parts left over on each side hold the same digits. This article starts with reading box shapes, then goes through the steps for counting leftovers and how to use the result to remove candidates. Note that the puzzles you can play on this site follow the standard rules; Jigsaw Sudoku is covered here as something to read about.

The rule of irregular boxes

Apart from the shape of the boxes, the rules of Jigsaw Sudoku are the same as standard Sudoku. You place 1 to 9 once in every row and every column, and once in each of the nine boxes outlined in bold. Every box has nine cells, but the shapes change from puzzle to puzzle: some are L-shaped or stepped like a staircase, others stretch out along a row. SudokuWiki, a site that explains solving strategies one by one, notes that Jigsaw Sudoku has not been studied as much as the standard puzzle, and that the different box shapes open up new strategies. Strategies that rely on a unique solution are harder to use: a valid rectangle is hard to identify among the jumble of shapes, and the SudokuWiki solver leaves these tests out. The main new strategy is the leftover count this article is about. Rows and columns keep the standard rule, so techniques that use only rows and columns work exactly as before.

Reading the shape of a box

Before you start, read each box: which rows and columns it touches, and how many cells it has in each. A standard box has three cells in each of three rows. A jigsaw box may have five cells in one row and one in the next, and this spread decides when the intersection techniques come into play. If, inside a box, all the candidates for a digit lie in the part of the box that sits in one row, you can remove that digit from the cells of the row outside the box. This is the same idea as the standard pointing pair, and it holds whether that part is two cells or five. The reverse also works: if, in a row, all the candidates for a digit lie in the part that belongs to one box, you can remove it from the cells of that box outside the row, which is box/line reduction. SudokuWiki notes that the box shapes give these two techniques more chances than in standard Sudoku. Long, thin boxes with more than three cells in one row are especially useful, because they leave few cells in the rest of the row. If a box has seven cells in a row, only two cells of the row lie outside the box. Any digit that cannot go in those two cells must go in the seven, so it can be removed from the two cells of the box outside the row. If a box has eight cells in a row, the one cell of the box outside the row and the one cell of the row outside the box must hold the same digit. That is the smallest case of the leftover count in the next section.

Cells a box has in one row and what they tell you
Cells of the box in the rowCells of the row outside the boxCells of the box outside the rowWhat you can read
8 cells1 cell1 cellThe two cells hold the same digit
7 cells2 cells2 cellsThe two pairs of cells hold the same two digits
6 cells3 cells3 cellsThe two groups of three cells hold the same three digits
5 cells4 cells4 cellsThe two groups of four cells hold the same four digits
4 cells5 cells5 cellsThe two groups of five cells hold the same five digits
3 cells6 cells6 cellsThe same count as a standard box in one row; both sides hold the same six digits, but narrowing them down needs other clues

Columns can be read the same way. With one row and one box, the cells on each side all sit in the same row or the same box, so no digit repeats within a side.

The idea behind the leftover count

The leftover count comes from the fact that a set of rows and a set of boxes each contain 1 to 9 the same number of times. Two adjacent rows contain each digit from 1 to 9 twice. Any two boxes also contain each digit twice. So pick two boxes that cover most of the two rows. The cells where the rows and the boxes overlap belong to both, and once you take them away, what remains on each side must hold the same digits. On the row side, what remains is the cells of the two rows that are not in the chosen boxes: cells poking in from other boxes. On the box side, what remains is the cells of the chosen boxes that stick out of the two rows. The two groups have the same number of cells and hold the same digits. SudokuWiki explains this with the totals: one row adds up to 45, two rows to 90 and three rows to 135, so any shortfall inside the region must be made up elsewhere with an identical set of digits. Knowing the digits match does not tell you which cell holds which. The order comes from comparing the candidates on both sides or combining them with other clues.

Steps for counting leftovers

Here is the count step by step. First, pick the rows that make up the region. According to SudokuWiki, a region can be a single row or column, or two or three adjacent ones, and going to four or more gains little. Two is the most common choice. Next, choose the boxes that cover most of the region, as many boxes as the region has rows. Then mark the cells of those boxes that stick out of the region, and mark, in a different way, the cells of the region that are not in the chosen boxes. Once you have checked that the two sets of marks are equal in number, write out the candidates on each side and compare them. A digit missing from one side can be removed from the other. If the numbers of marks do not match, suspect the number of boxes you chose or your reading of the box borders. The table below sets out these steps and what to check at each stage.

Steps for counting leftovers
StepWhat to doWhat to check
1Pick one to three rows or columns as the regionGroup adjacent rows together, or adjacent columns together
2Choose as many boxes as the region has lines, covering most of the regionThe chosen boxes cover as many cells of the region as possible
3Mark the cells of the chosen boxes that stick out of the regionCount them by following the bold borders
4Mark, differently, the cells of the region not in the chosen boxesThe numbers of marks in steps 3 and 4 are equal
5Write out and compare the candidates on both sidesRemove from one side any digit found in no cell of the other
6Place the digits in cells that are now narrowed downCarry each placement into its row, column and box

If the numbers of marks are not equal, either the number of boxes differs from the number of lines in the region or a box border has been misread.

Matching the two sides

Once you have counted, compare the candidates on the two sides. If each side is a single cell, the two cells hold the same digit, so any candidate missing from one can be removed from the other. If the two cells share only one candidate, both are placed at once. With two or more cells per side the idea is the same: both sides hold the same digits, so a digit that is a candidate in no cell on one side cannot go in any cell on the other. In the SudokuWiki example, a two-row region had candidates 3, 6, 8 and 9 on both sides. One side had no 3 and the other had no 9, so both sides were narrowed to 6 and 8, which then placed two cells. Take care when the region has two or more rows. The cells sticking out may be spread across different boxes, and the cells poking in may lie in different rows. Two cells that share no row, column or box can hold the same digit. In that case, read the rule as each digit appearing the same number of times on both sides, rather than as the same set: if a digit appears twice on one side, it appears twice on the other. With one row and one box, each side sits within a single row or box, so this cannot happen.

Widening the region

If a single row gives nothing, widen the region to two rows, then three. As the region grows, you choose that many more boxes, and the number of cells you mark changes too. Whatever the size, each digit still appears the same number of times on both sides. Columns work the same way: group adjacent columns and choose the boxes that cover most of them. When choosing a region, prefer one that leaves few cells sticking out. The fewer cells you mark, the less work it is to write out and compare the candidates. SudokuWiki cautions that, depending on the box shapes, some puzzles suit the leftover count well and others hardly at all. If rows give nothing, try columns; if those give nothing either, go back to standard techniques to thin out the candidates, then count again. Each time candidates drop, comparing the same region can give a different result.

Region size and how to count
RegionTimes each digit appearsBoxes to chooseDigits within one side
One row or columnOnce1No repeats
Two adjacent linesTwice2Two cells sharing no row, column or box can hold the same digit
Three adjacent linesThree times3Two cells sharing no row, column or box can hold the same digit
Four or moreAs many times as linesAs many as linesMore cells to count for little gain

Whatever the size, choose as many boxes as the region has lines. With one box too many, the side sticking out has nine more cells than the side poking in, and the two cannot be compared.

Common mistakes and how to spot them

Mistakes in Jigsaw Sudoku come from misreading shapes and miscounting. The first is reading the boxes as if they were standard 3x3 squares. Miss a bold border and you take a cell of the next box for part of this one, then remove candidates by mistake. Before removing a candidate, trace the bold border again to confirm which box the cell belongs to. The second is choosing a different number of boxes from the number of lines in the region. Choose three boxes for two rows and the side sticking out has nine more cells, so the comparison no longer holds. You can catch this because the marks on the two sides do not match. The third is confusing matching digits with a known order. Even when both sides are narrowed to 6 and 8, telling which cell is the 6 takes another clue. The fourth is overlooking, when the region has two or more rows, that a digit can appear twice on one side. Two cells that share no row, column or box can hold the same digit. The fifth is applying unique-solution techniques without checking that the four corners fit in two rows, two columns and two boxes. Valid shapes are hard to identify among irregular boxes, and the SudokuWiki solver leaves those tests out. Techniques that remove candidates at box intersections, and the idea of rows, columns and boxes, are covered in the related articles and glossary entries.