DO's and DON'T's for creating puzzles.

Content and Quality

Recognizable Picture

Puzzles need to have a picture that makes sense. Scattered squares or random lines make the puzzle boring to solve. View example.


If you want to deliver a message, do it through a picture and not just text.


Series of Alphabet are accepted if the letters are decorated or have a picture attached and aimed for the Studying English theme.


Simple pictures of houses, smilies, boats, flowers, etc are rarely accepted. Try to give life to the picture and make it unique in its own way.

Flags usually consist of 2 or 3 stripes of color and the result is not a real puzzle. Since we wish to maintain impartiality, no flags are accepted.


Use the triangles to smooth the image and make it nicer and not as a "color" to fill in a large area. Puzzles with many triangles are sometimes too confusing to solve. View example.


Totally symmetrical puzzles are very predictable to solve. They are accepted only if the image is designed for kiddlers. Try to do the puzzles on small grids and choose a suitable subject for kids.


An international community, including kids, solves the puzzles. Be kind to each other. Personally or generally offensive puzzles are not accepted, nor obscene, political or missionary images.


It's hard to know what images (puzzles) we already have online. The best way is to use the Themes. For example, if you want to create Pikachu, go to Cartoons and Comics: Animation: Anime: Pokemon. Browse through the titles, or search "pikachu".

Flipped images (like the Dogs) are considered as duplicates. The puzzles have identical clues, just in reversed order. In order to solve them we are required to use the same logic.

Technical Issues

Background Color

The background color defines the empty squares of the puzzle. Always select the color of the largest area. It makes the puzzle better and will require more logic to solve it.

In the Boat example the main color is black. When we select black for the background the workshop produces a picture that looks "hopeless" to validate.

When we select white the puzzle is solvable. However, one of the rules for solving black and white puzzles is "Between two clues there is at least one empty (background) square".

If we look at the image closely we can see that most of it is built of 1 pixel of white lines. Since the background is white - the puzzle becomes a fill-in and the solver doesn't need to use much logic to solve it. View example. Note that the grid size is 40x40 and the clues in the highlighted rows and columns sum up to 40.

We added lines and squares to make the Boat solvable, and the result is a good logic puzzle. View example.


Removing solid lines of the background around the image is important. If we leave those areas they turn into empty clue-lines on the puzzle.

Let's look at Winter Scene image (size - 50x50).
It has 12 solid lines of white at the bottom. We can trim those lines (size - 50x38).
We can also add details and make it a multiples-of-5 grid puzzle (size - 50x45).

Low Background after Trimming

Let's look at the Pears.
If we select black for the background - the straight lines at the top and the bottom will turn into empty clue-lines. View example.

So first we need to remove those lines. However, after trimming the workshop alerts: "too low background color". We can add black inside the image, or - select another color for the background.

The best background color for this image is the green because it covers the largest area of the grid. View example.


When taking a full color picture and using a graphic tool to reduce the number of colors to 8 - the result is usually a messy image with many 1 pixel colors. Each pixel turns into a clue and the puzzle consists of many clues of "1".

The original picture of the Bird had 256 colors. We reduced the colors to 8 and made it a puzzle. View example. Look at the 13th column from right. The number of clues is 21. The number of empty grid cells is 25. It is a boring line to solve.

We cleaned the image, used only 5 colors, and changed the random squares into solid lines and areas of colors. The result is an enjoyable puzzle to solve. View example.

In large grids, like the Cat, the need to clean the image is essential. If we don't do it, the puzzle will have many long lines of clues. Users may give up on solving that puzzle because it requires too much tedious work. View example: Left clue-lines | Top clue-lines


Image Size

The first step in creating a multi is finding the right size for the image. If the image is big and lacks details, the parts of the multi have large areas of fill-in. The solvers are required to count big numbers and draw long lines.

Tips for finding the right size:

  • Reduce the size of the image as much as possible until you see that it starts to lose details.
  • Zoom in and look at lines drawn in the image. If they are wide (2 pixels or more) - make the image smaller.

Image Quality

Solving a multi requires more time and effort. The solvers expect to get an impressive result according to their work.

A good way to test the image quality is to import it to the workshop and validate. If most of the parts have relatively few points the image lacks enough details to make it interesting.

Add details to the image and make each part of the multi an enjoyable puzzle to solve.

Balanced Parts

The most common mistake in creating multis is splitting the image into unbalanced parts.

The Dancers image is 60x60. After importing the image, the workshop split it automatically into 4 parts, as follows:

1 - 50x50
2 - 10x50
3 - 50x10
4 - 10x10

Only part 1 is a real puzzle. Parts 2, 3, 4 are unattractive to solve.

We changed the sizes to 30x30, got 4 equal parts and 4 good puzzles.

Trimming and Balancing

Removing surplus background around the image is essential. Note that after trimming, the workshop split the parts back to 50x50. You will need to change the sizes again in order to get balanced parts.

Multiples-of-5 Grids

Multis are better when the parts have multiples-of-5 grids, e.g.: parts of 30x35 are better than 29x34.
X-treme multis must have multiples-of-5 grids.

Self Duplicate

In multis, if the image is totally symmetrical, the result is duplicate puzzles.

Let's look at the Bear.
The image is completely identical on both sizes. Part X1 equals part Y1; Part X2 equals part Y2.

As puzzles, the top two have identical clues, just flipped. The same happens with the bottom two puzzles. So actually we duplicated ourselves two times.


Empty parts are a common problem in creating multis. Sometimes this problem is seemingly solved by drawing a frame around the image.

In the Nymph image there are 6 empty parts (marked in X). If we draw a black line on the left/top/right edges and the workshop enabled us to submit the multi. However, none of these parts are real puzzles because they contain only one horizontal and/or one vertical line.

Even if we make the frame nicer, for example:

the multi still will not be accepted because there are now 3 duplicates.

This is why frames are not allowed in multis. The only solution is to add real details.

Good Luck!