Tetravex is NP-complete

Tetravex is a widely played one person computer game in which you are given $n^2$ unit tiles, each edge of which is labelled with a number. The objective is to place each tile within a $n$ by $n$ square such that all neighbouring edges are labelled with an identical number. Unfortunately, playing Tetravex is computationally hard. More precisely, we prove that deciding if there is a tiling of the Tetravex board is NP-complete. Deciding where to place the tiles is therefore NP-hard. This may help to explain why Tetravex is a good puzzle. This result compliments a number of similar results for one person games involving tiling. For example, NP-completeness results have been shown for: the offline version of Tetris, KPlumber (which involves rotating tiles containing drawings of pipes to make a connected network), and shortest sliding puzzle problems. It raises a number of open questions. For example, is the infinite version Turing-complete? How do we generate Tetravex problems which are truly puzzling as random NP-complete problems are often surprising easy to solve? Can we observe phase transition behaviour? What about the complexity of the problem when it is guaranteed to have an unique solution? How do we generate puzzles with unique solutions?

💡 Research Summary

The paper investigates the computational complexity of the popular one‑player puzzle game Tetravex. In Tetravex the player is given n² unit squares (tiles); each tile has an integer label on each of its four edges. The goal is to place every tile inside an n × n board so that any two adjacent tiles share the same label on the touching edge. The authors formalize this as the decision problem TETRAVEX‑DECIDE: given a set of tiles, does there exist a placement that satisfies all adjacency constraints?

First, they show that the problem belongs to NP. A candidate placement can be verified in O(n²) time by scanning all neighboring pairs and checking label equality, which is clearly polynomial. Hence TETRAVEX‑DECIDE ∈ NP.

The core contribution is a polynomial‑time many‑one reduction from 3‑SAT, a classic NP‑complete problem, to TETRAVEX‑DECIDE, establishing NP‑hardness. The reduction proceeds in three conceptual layers:

1. Variable tiles – For each Boolean variable x_i the construction creates two distinct tiles, one representing the assignment x_i = true and the other x_i = false. These tiles are placed in a dedicated row of the board; their orientation (or rotation) encodes the truth value of the variable.

2. Clause gadgets – For each clause C_j = (ℓ₁ ∨ ℓ₂ ∨ ℓ₃) a small “gate” composed of several tiles is built. The gate’s edge labels are chosen so that it can be satisfied (i.e., the tiles can be placed without conflict) if and only if at least one of its three input literals is true. The gate essentially simulates the logical OR operation using matching numbers on tile edges.

3. Frame and wiring – Additional “frame” tiles surround the variable rows and clause gadgets, ensuring that the only way to fill the whole board is to respect the intended wiring between variables and clauses. The frame prevents spurious placements and guarantees that any complete tiling corresponds to a consistent truth assignment for all variables.

If the original 3‑SAT instance is satisfiable, one can set the orientation of each variable tile according to a satisfying assignment, and then each clause gadget will fit perfectly, allowing the entire board to be tiled. Conversely, any valid tiling of the constructed Tetravex board reveals the orientation of each variable tile, which can be read off as a satisfying assignment for the original formula. This bidirectional correspondence is computable in polynomial time, establishing the reduction 3‑SAT ≤_p TETRAVEX‑DECIDE.

Since TETRAVEX‑DECIDE is both in NP and NP‑hard, the authors conclude that it is NP‑complete. The result explains why Tetravex puzzles can feel “hard” to human solvers: the underlying decision problem encodes arbitrary Boolean formulas.

Beyond the core proof, the paper outlines several avenues for future work. It asks whether the infinite‑board version of Tetravex can simulate a Turing machine, i.e., be Turing‑complete. It raises the practical question of generating “truly puzzling” instances: random NP‑complete instances are often easy for humans, so one might need to control structural properties to achieve a challenging puzzle. The authors suggest studying phase‑transition phenomena (the sudden change in solvability probability as instance parameters vary) in Tetravex, analogous to what has been observed for SAT and other combinatorial problems. Finally, they discuss the special case where a puzzle is guaranteed to have a unique solution, asking how to generate such instances efficiently and what impact uniqueness has on computational difficulty.

In summary, the paper provides a rigorous proof that deciding the existence of a valid Tetravex tiling is NP‑complete, situating the game within the broader landscape of tiling‑based puzzles whose decision versions are computationally intractable. It also opens a rich set of research directions concerning infinite versions, instance generation, difficulty measurement, and uniqueness constraints, thereby linking recreational puzzle design with deep questions in theoretical computer science.