Wataridori Puzzle: NP-Complete Grid Challenge

• Wataridori Puzzle is a grid-based challenge featuring polyomino regions with circles that are either labelled or wildcards, requiring exact pairings via orthogonal, vertex-disjoint paths.
• The puzzle’s NP-completeness is demonstrated through a reduction from the Numberlink puzzle, using intricate block-gadget constructions to mirror region-count constraints.
• Practical solving approaches include depth-first backtracking, ILP/SAT encodings, and dynamic programming on restricted instances, highlighting both algorithmic challenges and heuristic strategies.

Wataridori is a pencil puzzle defined on a rectangular $m\times n$ grid. Each cell belongs to one of several connected regions (polyominoes), and some cells contain circles which may be unlabelled (wildcards) or carry positive integers. The objective is to connect every circle to exactly one other circle via vertex-disjoint paths that satisfy specific labeling and region-traversal constraints. Paths must be orthogonal and may not cross or share any cell. In each path, the total number of distinct regions traversed must equal the integer label (if present) on the endpoints; unlabelled (wildcard) circles are compatible with any label. The union of paths is not required to cover all grid cells—only that every circle is paired.

1. Formal Structure and Problem Definition

A formal definition (Ruangwises, 14 Jan 2026) specifies that a Wataridori instance comprises:

• an $m\times n$ grid,
• a partition into connected regions (polyominoes),
• circles placed on distinct cells, each either unlabelled or carrying a positive integer.

A solution consists of a collection of vertex-disjoint paths, each subject to:

1. Pairing two circles with matching labels (wildcards may match any label),
2. Orthogonality—paths only move horizontally or vertically,
3. Path disjointness,
4. Enforced region-count—each path traverses exactly $k$ distinct regions if the endpoint label is $k$ (entry/exit into each region at most once per path).

The canonical decision problem, Wataridori–Solvability, asks whether such a pairing exists for a given puzzle instance. Membership in $\mathrm{NP}$ arises since a nondeterministic algorithm can guess all paths and verify constraints in polynomial time.

2. NP-Completeness and Computational Complexity

Wataridori–Solvability is proven to be NP-complete (Ruangwises, 14 Jan 2026). The proof proceeds via a polynomial-time reduction from the Numberlink puzzle, another grid-based pairing problem where vertices labeled with symbols must be connected via vertex-disjoint orthogonal paths. Numberlink–Solvability is known to be NP-complete, even without the requirement that all grid cells be covered.

The reduction maps a Numberlink instance $G$ (with $p$ symbol pairs) to a Wataridori instance $H$ constructed on a $(4k+5)m \times (4k+5)n$ grid, where $k = \lceil (p-1)/2 \rceil$. Each cell in $G$ becomes a $(4k+5)\times(4k+5)$ block in $H$:

• Number blocks: For symbol $i$, center circle is labeled $4k + 2i + 1$ and block structure enables choosing among $k+1$ distinct zig-zag path extensions with region-counts matched to the center label.
• Empty blocks: Admits only straight path traversal without branching.

Paths in $H$ correspond one-to-one with those in $G$, and construction details ensure polynomial size and time. The reduction is correct in both directions: solutions for $G$ translate to solutions for $H$ and vice versa.

3. Gadget Construction and Reduction Details

The block-gadget methodology is central to the computational hardness result. Each block mimics the behavior of a Numberlink cell:

• Paths can enter and exit through four sides, corresponding to grid adjacency.
• Number blocks have walls and zig-zag corridors engineered to admit $(k+1)$ region traversal choices, all with odd region counts in $\{2k+2, 2k+4, \dots, 4k+2\}$; paired endpoints yield total counts in $\{4k+3, 4k+5, \dots, 8k+3\}$, exactly matching assigned labels.
• Empty blocks restrict path movement to a unique straight path, disallowing additional region traversal options.

The block arrangement ensures that global constraints on vertex-disjointness and local region-counts are preserved under the reduction. Each path in $G$ transforms to a path in $H$ whose endpoints and region-count correspond precisely to the original connection.

4. Algorithmic Approaches and Practical Solving

Given NP-completeness, polynomial-time solutions are infeasible unless P=NP. Exact solvers thus rely on exponential-time strategies:

• Depth-first backtracking with constraint propagation (especially region-count and connectivity checks),
• Integer linear programming (ILP) or SAT encodings for use with combinatorial solvers,
• Dynamic programming on instances with bounded width or tree-width (fixed-parameter tractable in certain cases).

Traditional approximation algorithms are not directly relevant since Wataridori is a decision problem; optimization versions (e.g., maximize number of correct pairings) represent potential future investigation but are not covered by existing results (Ruangwises, 14 Jan 2026).

In practice, human solvers employ heuristics combining local deductions (especially from the region-count constraint), search with look-ahead and pruning, and pattern-based rules adapted from Nikoli-style puzzles (e.g., avoidance of 2$\times$2 “wiggle” traps).

5. Connections to Related Puzzles and Theoretical Implications

Wataridori generalizes and extends complexity boundaries established by Numberlink and related grid–pairing puzzles. The proof leverages established NP-completeness for Numberlink by embedding its logic in a region-count-constrained environment. This connection demonstrates that the addition of region constraints does not reduce computational hardness but inherits it via construction.

Table: Comparison of Key Features

Puzzle	Pairing Constraint	Region Traversal	Known Complexity
Wataridori	Same label or wildcard	Region-count matches label	NP-complete
Numberlink	Same symbol	None	NP-complete

The region-count property and wildcard mechanics in Wataridori open further directions for constraint satisfaction problem research, especially in the design of heuristic solvers and the study of fixed-parameter tractability under restricted region structures.

6. Concluding Perspective

Deciding the solvability of Wataridori is computationally intractable in the general case, as established via polynomial-time reduction from Numberlink (Ruangwises, 14 Jan 2026). The puzzle encapsulates a rich blend of local connectivity, label-pairing, and region cardinality constraints, providing a challenging benchmark for both human and algorithmic solver design. The result situates Wataridori among the class of NP-complete combinatorial puzzles and clarifies the computational limits facing both brute-force and modern SAT/ILP-based approaches. The presence of additional constraints—particularly the region-count parity and wildcards—suggests potential for novel algorithmic developments, though such enhancements await further detailed study.