Computational Properties of Slime Trail

We investigate the combinatorial game Slime Trail.This game is played on a graph with a starting piece in a node. Each player’s objective is to reach one of their own goal nodes. Every turn the current player moves the piece and deletes the node they came from. We show that the game is PSPACE-complete when played on a planar graph.

💡 Research Summary

The paper investigates the computational complexity of the combinatorial game Slime Trail, a two‑player impartial game played on a connected graph. A token starts on a designated vertex; on each turn a player moves the token to an adjacent vertex and permanently “slimes” (deletes) the vertex it just left, making it unavailable for future moves. Each player has a set of goal vertices colored uniquely (Blue for the first player, Red for the second). A player wins by moving the token onto one of their own goal vertices. An additional rule requires that a move must always leave at least one reachable goal vertex for at least one player; if a move isolates all of a player’s goals, the opponent wins automatically.

The authors first establish that Slime Trail belongs to PSPACE. They observe that the game tree depth is bounded by the number of vertices m minus one, because each move permanently removes a vertex. Consequently, a depth‑first search that explores each branch sequentially uses only O(m²) space (to store the current branch and visited status), which is polynomial. Hence the decision problem “does the first player have a winning strategy?” is in PSPACE (Lemma 2.1).

The core contribution is a PSPACE‑hardness proof via a reduction from Quantified Boolean Formulas (QBF), a canonical PSPACE‑complete problem. A QBF has the form ∃x₁ ∀x₂ ∃x₃ … Qₙxₙ φ(x₁,…,xₙ) where φ is a CNF formula. The reduction constructs a planar Slime Trail board that simulates the alternating quantifier game: the first player (Blue) chooses truth values for odd‑indexed variables, the second player (Red) chooses values for even‑indexed variables. The construction uses several reusable “gadgets”:

1. Odd‑Variable Gadget (Fig. 2) – When the token enters this gadget on Blue’s turn, Blue can move left (assign false) or right (assign true) to a pair of vertices a₁/b₁. After this choice, Red is forced along a deterministic path that eventually returns the token to an “Out” vertex, from which a wire gadget leads to the next component. The gadget is designed so that any deviation (e.g., moving directly to the choice gadget) results in an immediate loss for the deviating player, enforcing the intended variable assignment.

2. Even‑Variable Gadget (Fig. 3) – Symmetric to the odd gadget but with Red making the binary choice. The token then proceeds through a deterministic sequence that again ends at an “Out” vertex, ready for the next wire.

3. Wire Gadget (Fig. 4) – Connects the “Out” vertex of a variable gadget to the start of the next gadget (or to the final choice gadget). It guarantees that the player whose turn it is when leaving a variable gadget will be the same player who starts the next gadget, preserving the alternation of quantifiers.

4. Choice Gadget (Fig. 5) – After all variables have been set, Red moves from the wire into this gadget and selects a clause of the CNF formula. Blue then selects a literal within that clause. The structure of the gadget ensures that a literal is traversable only if the corresponding variable assignment (made earlier) makes the literal true; otherwise the path is blocked (the vertex has been “slimed”). If all literals in the chosen clause are blocked, Blue cannot reach any of its goal vertices and loses, giving Red the win. Conversely, if at least one literal is available, Blue can move to its goal vertex and win. This mirrors the evaluation of the CNF formula under the chosen assignments.

5. Crossover Gadget (Fig. 6) – To keep the overall graph planar, the construction must sometimes let two wires cross. The crossover gadget allows two logical “wires” to intersect without sharing vertices, while preserving the forced move order and preventing any player from gaining an unintended shortcut. It is carefully designed so that the token must follow a prescribed path through the crossover, and any alternative leads to an immediate loss.

By chaining these gadgets according to the quantifier order of the given QBF, the authors obtain a planar Slime Trail instance that is winning for the first player if and only if the original QBF evaluates to true. Since QBF ≤ₚ Slime Trail, the game is PSPACE‑hard on planar graphs. Combined with the earlier PSPACE membership result, Slime Trail is PSPACE‑complete on planar graphs (Theorem 3.1).

The paper concludes with two open problems: (1) whether Slime Trail remains PSPACE‑complete when restricted to the usual square or hexagonal grids used in practice, and (2) the complexity when each player has only a single goal vertex (the current reduction relies on multiple goal vertices). These questions point toward a richer understanding of the game’s difficulty under more realistic or constrained settings.

Overall, the work adapts the classic QBF‑to‑Geography reduction technique to a novel game, introduces a planar‑preserving crossover gadget, and thereby extends the landscape of PSPACE‑complete combinatorial games. The presentation suffers from typographical errors and occasionally vague gadget descriptions, but the logical structure of the reduction is sound and the result is a valuable contribution to algorithmic game theory.