The rules fit in three lines
- Cut the entire grid into rectangles.
- Each rectangle contains exactly one number, and that number equals how many cells it covers.
- Rectangles never overlap, and every cell ends up inside exactly one of them.
That is all there is to memorise. Unlike sudoku, there is no candidate bookkeeping to maintain. The only question you ever ask is "where can this number's rectangle go?"
There is a correct place to start
Where you begin changes how hard the grid feels. You are hunting for numbers with the fewest possible placements.
How to spot a low-choice number
- Primes (5, 7, 11…): only two shapes exist, 1×N and N×1. They often snap straight into a single line.
- Numbers in a corner or against an edge: the grid boundary removes half their options.
- Small numbers like 2 and 3: they simply cannot travel far.
- Large numbers (12, 16…): a big area bumps into edges and other numbers, so the option count is smaller than it looks.
The opposite case: a 4 or a 6 sitting in open space near the middle has many shapes and many positions. Staring at those first is how you lose ten minutes.
Worked example: clearing a grid with one technique
This is the first level of the beginner ladder in our app (5×5). The whole grid falls to a single idea — place the number that has only one placement left, repeat.
| 6 | 5 | |||
| 4 | 5 | |||
| 5 |
| 6 | 5 | |||
| 4 | 5 | |||
| 5 |
The order runs like this.
- The 5 in the fourth column: 5 is prime, so the shape is 1×5 or 5×1. Laid flat it would swallow other numbers on that row, so it is forced into a vertical column.
- The 5 in the fifth column: same argument, forced vertical.
- The 5 at the bottom left: this one started with two options, flat or vertical. With columns four and five now taken, it can no longer run flat — forced vertical.
- The 6: only one 2×3 placement still fits the remaining space.
- The 4: what is left is exactly a 2×2.
Step three is the interesting one. Something that had two options a moment ago drops to one because of what happened elsewhere. That chain reaction is the whole game.
Only five techniques actually matter
Our app ships with a solver that works the way a person does, one technique at a time. We ran all 486 shipped levels through it and counted every deduction: 10,651 in total.
| Technique | What it does | Times fired | Share |
|---|---|---|---|
| 1. Only placement | A number has just one placement left | 6,025 | 56.6% |
| 2. Only cover | An empty cell can be covered by exactly one candidate rectangle on the whole grid | 2,145 | 20.1% |
| 3. Owned cell | Every candidate covering a cell belongs to the same number, so that number must pass through it | 1,325 | 12.4% |
| 4. Shared cell | A number passes through a cell in every one of its placements, so no other number may use it | 967 | 9.1% |
| 5. Starvation | A placement would leave some cell with no possible cover at all, so it is impossible | 189 | 1.8% |
Techniques 1 and 2 — keep harvesting what is forced
More than three quarters of all progress comes from these two. Either you look from the number's side and find only one placement, or you look from an empty cell's side and find only one rectangle that can reach it. The worked example above is nothing but this.
Technique 2 is the one people forget. Ask it out loud: "this corner cell — who is ever going to take it?" Grids give up a surprising number of forced moves to that question.
Techniques 3 and 4 — eliminate instead of place
These do not put a rectangle down. They remove possibilities, which is what unlocks the next forced move.
Technique 4 (shared cell) is the one that feels best to use. Suppose a 4 could sit vertically or horizontally, but either way it covers one particular cell. Then no other number is allowed to touch that cell — even though you still have no idea where the 4 finally goes. Getting hard information out of an undecided number is the most satisfying moment shikaku has.
That is also why the hints in our app return the reason a cell is forced rather than the answer. The reasoning is the part worth taking away.
Technique 5 — place it, then find the wreckage
"If I put it here, that cell over there becomes impossible for anyone to cover." So the placement is out. It fired 189 times (1.8%), and the rarity is honest: this one only shows up when you are grinding out the hardest grids.
The technique that never fired
An honest note. The solver also implements region area counting: when the empty space splits into islands, you check things like "this island is 7 cells; the only numbers that can live there are 4 and 3; 4 + 3 = 7, exact fit." It sounds powerful.
Across the 486 shipped levels plus generated ones, it fired zero times. The grid only fragments into islands late, and by then the other techniques have already settled everything.
It does no harm sitting there, but counting it as a technique that "works" would be false. The load is carried by the five above, and mostly by the first four. Technique lists tend to grow because a longer list looks more authoritative — you do not know which entries earn their place until you count.
How to spot a puzzle that forces you to guess
This is the part that matters most as a player. Puzzles really do exist that stall halfway, and whose published solution turns out to require brute force at the end. There are two causes.
| Cause | What you experience | How to detect it |
|---|---|---|
| More than one valid solution | Two different cuts both satisfy every rule, and being told yours is "wrong" feels arbitrary | Swap a region around and nothing contradicts |
| Logic runs out before the end | The solution is unique, but the only route to it is assume-and-contradict | You stall completely, then a random placement turns out to be accepted |
The second is backtracking: assume a placement, push it forward, hit a contradiction, conclude it was wrong. Perfectly valid logic — and essentially unplayable unless you can hold several moves of state in your head and unwind them. A puzzle that depends on it is not deep, it is tedious.
What we do about it
All 486 levels in SHIKAcookie are verified by machine before release: (1) an exact solver confirms there is exactly one solution, and (2) the human-technique solver must finish the grid on its own and land on that same solution.
The result: zero levels needed backtracking. Puzzles generated on the fly in endless mode go through the same two checks before you ever see them.
Difficulty comes from technique depth, not grid size
A bigger grid is not a harder grid. What decides difficulty is how far down the technique list you have to go. Here is the breakdown of the 486 shipped levels.
| Deepest technique needed | Levels | What it feels like |
|---|---|---|
| Technique 1 | 76 | Harvest forced placements, nothing else |
| Technique 2 | 173 | Start asking from the empty cell's side too |
| Technique 3 | 77 | Work out who a cell belongs to |
| Technique 4 | 62 | Extract information from undecided numbers |
| Technique 5 | 98 | Place it and look for the wreckage |
Grids run from 5×5 up to 14×14, and a 14×14 that needs nothing but technique 1 is a perfectly ordinary puzzle. Wide does not mean hard. A 5×5 that needs technique 4 will stop most people on a first encounter.
Three ways back in when you are stuck
- Flip to the empty cell's point of view. Pick the loneliest-looking empty cell and ask who can reach it (technique 2).
- Find a number with exactly two placements. Two placements very often share a cell, and that cell is then off limits to everyone else (technique 4).
- Go back to the edges. When the middle is stuck, there is usually a forced move still sitting untouched along the border.
If none of that moves the grid, it is fair to start suspecting the puzzle. In a properly built shikaku there is always a "this can only go one way" somewhere on the board.
Summary
- One rule: cut the grid into rectangles whose areas match their numbers.
- Open with primes, edges and small numbers. Leave the mid-board 4s and 6s for later.
- Five techniques do all the work, and 76.7% of it is just harvesting what is already forced (measured).
- When stuck, ask the question from the empty cell's side.
- Needing to guess is a defect in the puzzle, not in you. A good one goes the whole way on logic.