Everyone assumes fewer clues means harder. It is a reasonable guess and it is only loosely true. Two puzzles with the same clue count can differ enormously, and the reason is worth understanding because it changes how you approach a grid.
Clue count sets a floor, not a level
A grid with fifty clues will be easy; there is simply not enough left to be difficult. Below about thirty, difficulty becomes possible but is not guaranteed. The proven minimum for a unique solution is seventeen clues, and such puzzles are brutal — but a badly constructed twenty-five-clue grid can be easier than a well-constructed thirty-two.
Placement matters more
What actually drives difficulty is how the clues are distributed. Clues clustered in a few boxes leave other boxes almost empty, and empty boxes resist scanning. Clues spread evenly give every region something to work from, which keeps hidden singles available.
A box with no clues at all is a strong difficulty signal. So is a row or column with only one. When you open a grid, glance at the emptiest regions first — that is where the puzzle will fight you.
The real measure: required techniques
Puzzle setters grade difficulty by the hardest technique needed to finish without guessing.
| Grade | Hardest technique needed |
|---|---|
| Easy | Scanning and hidden singles only |
| Medium | Naked singles, occasional candidate work |
| Hard | Naked and hidden pairs, locked candidates |
| Expert | Triples and longer elimination sequences |
| Extreme | X-wing, colouring, chains |
This is why a puzzle can feel hard to one solver and easy to another: it depends entirely on which techniques you already have. Learning locked candidates turns a whole tier of puzzles from impossible to routine.
How the puzzles here are graded
Each grid is generated by filling a complete solution, then removing clues in symmetrical pairs and verifying after every removal that exactly one solution remains. Removal stops at a clue target: about 42 for easy, 34 for medium, 29 for hard, 25 for expert.
Because uniqueness is checked at every step, no puzzle here can ever require a guess. It also means the difficulty band is approximate — an individual hard puzzle might land closer to medium depending on where the clues happened to fall.
If a puzzle feels wrong for its label
Generated puzzles vary within a band. A hard grid that solves in six minutes is not a mistake, and one that takes forty is not broken. The guarantee is uniqueness and solvability, not a precise time.
Symmetry, and what it does and does not mean
Clues here are removed in mirrored pairs, which gives the familiar rotationally symmetrical look. This is an aesthetic convention from newspaper puzzles and it slightly reduces how low the clue count can go. It tells you nothing whatsoever about the solution.
Choosing the right level
Pick the level where you finish most puzzles but occasionally get stuck for a while. If you never get stuck, you are practising execution rather than learning. If you get stuck every time, you are missing a technique — and the hint button will name it, which is the fastest way to find out what to study next.
Moving up a level after learning one new technique is a reliable pattern. Locked candidates typically take a solver from medium to hard; hidden pairs from hard to expert.