Daily Hitori

Hitori starts full and asks you what to remove

Hitori is a Japanese number-elimination logic puzzle. Unlike Sudoku or Hitori, the grid begins with every cell already filled. Your task is to shade some cells black so the remaining visible numbers obey three rules at the same time. No unshaded number may repeat in a row or column. Two black cells may never touch horizontally or vertically. Finally, every unshaded cell must belong to one connected orthogonal region. The tension comes from balancing local duplicate cleanup against the global connectivity rule.

Duplicate pairs are the first source of forced choices

If the same number appears twice in one row, at least one of those copies must eventually be shaded. The same is true for repeated values in a column. A pair does not always tell you immediately which copy is black, but it creates a binary choice that can interact with nearby repeats. Three equal numbers in a line are especially useful: because adjacent black cells are forbidden, the middle copy often becomes white while the outer choices become tightly constrained.

A black square proves its neighbors stay white

The no-touch rule is one of Hitori's strongest local deductions. Once a cell is definitely black, the cells directly above, below, left and right are definitely unshaded. Marking those neighbors as Keep can make later duplicate decisions much easier because a repeated number that is already known white forces its other copy to be shaded. Diagonal contact is allowed, so only the four orthogonal neighbors receive this protection.

Connectivity turns distant cells into one problem

The white area must remain connected by up, down, left and right movement. This rule prevents solutions that look locally clean but cut the board into separate islands. Narrow corridors matter: if one unshaded cell is the only bridge between two regions, that bridge cannot be shaded. Near corners and edges, connectivity often decides between two duplicate candidates that otherwise look equivalent.

Keep marks are reasoning notes, not extra rules

The game provides a Keep tool that circles a cell you believe must stay white. A Keep mark does not alter the underlying Hitori rules; it records your deduction and makes Check able to flag a duplicate pair that you have explicitly committed to keeping. Shade, Keep and Clear are separate tools so mobile play does not depend on awkward multi-click cycles. Right-click also places a Keep mark on desktop.

Four board sizes change the depth without changing the rules

V0.98 includes 4×4, 5×5, 6×6 and 7×7 boards. Four by four is compact enough to learn the interaction between duplicates and adjacency. Five by five is the default Practice size. Six and seven by seven create longer chains of duplicate choices and more meaningful connectivity bottlenecks while remaining comfortable on a phone-sized board.

Easy, Medium and Hard use separate frozen source pools

Every size has four Easy, four Medium and four Hard source puzzles. That produces forty-eight frozen sources before geometric symmetry is applied. Practice never invents an unchecked board in the browser. It selects a source from the requested pool using a deterministic URL seed, then applies one of eight rotations or reflections to the grid and its stored solution together.

Every shipped source is counted to exactly one solution

The Python generator constructs candidate number grids and runs an exact solver over the real Hitori constraints. Duplicate pairs become logical requirements that at least one matching cell is black. Orthogonal neighbors cannot both be black. A completed assignment is accepted only if all remaining white cells form one connected component. Candidates are shipped only when the solver finds exactly one complete assignment.

An independent verifier repeats the proof

Release verification does not simply trust the generator's intended answer. A separate JavaScript verifier rebuilds the Hitori constraints, counts solutions independently, and checks that the single solution matches the frozen mask stored with each source. It then rotates and reflects every source through all eight square symmetries and repeats the same count. Forty-eight sources therefore become three hundred eighty-four verified playable variants.

The calendar date replaces the Practice seed

Daily mode hashes the local calendar date to select a size, difficulty, source and symmetry variant. The same date therefore resolves to the same board on the same release. There is no server-side daily database to drift, and future dates cannot be opened from the built-in day navigation.

Hint, Check, Undo and Redo have explicit trust costs

Hint reveals one correct state from the verified solution. Check evaluates your current markings for adjacency conflicts, explicit Keep duplicates, remaining duplicate pairs and connectivity. Undo and Redo restore exact earlier player states. Because these tools can reveal information or reverse decisions, using them marks the current run Assisted. The puzzle can still be completed, but its record stays separate from a clean unaided completion.

Autosave preserves temporary mistakes instead of hiding them

A player's intermediate board can contain a wrong shade, an unresolved duplicate or a disconnected region. Autosave stores that real state together with move count and assistance status. Reopening the same puzzle restores it rather than silently discarding it because the board is not yet valid. The restore contract checks data shape and puzzle identity, while rule validation remains the player's job.

Daily Hitori is stable by calendar date

Daily Hitori derives its size, difficulty, source and symmetry from the date. Everyone running the same release on the same date receives the same specification without a server-hosted puzzle feed. The specialist keeps its own local completion history and streak. It does not enter the platform's frozen Today’s 6 rotation and does not change the six-game Daily Score.

There are no fabricated rankings or paid puzzle calls

The entire bank, solver and selection logic live in the project. There is no paid API, server opponent, global percentile or invented IQ score. Local statistics describe only what this browser actually knows: solved puzzles, clean solves, move-based bests and specialist Daily history. The product can therefore make strong uniqueness claims because those claims are backed by frozen data and repeatable verification instead of marketing language.

A useful solving loop is repeat, adjacency, connectivity

Start by scanning rows and columns for repeated values. For each duplicate, ask whether one candidate is adjacent to a cell already known black; if so, that candidate must stay white and its matching copy may become forced black. Next look for white cells that form a narrow bridge between regions. Finally rescan duplicate pairs after every forced Keep. Hitori rewards this repeated interaction between local number logic and global board shape.