Deciding the Winner of an Arbitrary Finite Poset Game is PSPACE-Complete

A poset game is a two-player game played over a partially ordered set (poset) in which the players alternate choosing an element of the poset, removing it and all elements greater than it. The first player unable to select an element of the poset loses. Polynomial time algorithms exist for certain restricted classes of poset games, such as the game of Nim. However, until recently the complexity of arbitrary finite poset games was only known to exist somewhere between NC^1 and PSPACE. We resolve this discrepancy by showing that deciding the winner of an arbitrary finite poset game is PSPACE-complete. To this end, we give an explicit reduction from Node Kayles, a PSPACE-complete game in which players vie to chose an independent set in a graph.

💡 Research Summary

The paper resolves a long‑standing open problem in combinatorial game theory by proving that determining the winner of an arbitrary finite poset game is PSPACE‑complete. A poset game is defined on a partially ordered set: players alternately pick an element, remove it together with all elements that are greater (i.e., lie above it in the order), and the player who cannot move loses. While special cases such as Nim admit polynomial‑time solutions, the complexity of the general problem was previously known only to lie between NC¹ and PSPACE.

The authors achieve the completeness result through a careful polynomial‑time reduction from Node Kayles, a well‑known PSPACE‑complete impartial game. In Node Kayles, players select vertices of a graph; each chosen vertex and all its neighbors are deleted, and the last player to move wins. The reduction constructs, for every vertex of the input graph, a small “gadget” within a poset. Each gadget consists of two layers: a primary element representing the vertex itself and a secondary element used only for enforcing removal of adjacent vertices. If two vertices u and v are adjacent in the original graph, the reduction adds order relations u₁ > v₂ and v₁ > u₂, where the subscript denotes the layer. Consequently, selecting u₁ removes both u₁ and its secondary copy u₂, and simultaneously forces the removal of v₁ and v₂, exactly mirroring the effect of choosing u in Node Kayles.

Key technical points of the construction include:

1. Isolation of selectable elements – only primary layer elements are ever chosen, ensuring a direct correspondence between a move in the poset game and a move in Node Kayles.
2. Automatic removal via secondary elements – secondary elements are never chosen directly but are positioned so that they are eliminated whenever a neighboring primary element is selected.
3. Polynomial size – the total number of poset elements is O(|V| + |E|), preserving polynomial bounds on the reduction.
4. Preservation of game dynamics – the reduction guarantees a one‑to‑one mapping between winning strategies: a first‑player win in the original graph translates to a first‑player win in the constructed poset, and vice versa.

Having established the PSPACE‑hardness, the authors then argue that the poset game decision problem lies in PSPACE. The game length is bounded by the number of elements, and each game state can be described using a polynomial amount of space. A depth‑first search of the game tree, with memoisation of visited states, fits within polynomial space, satisfying the definition of PSPACE.

The paper’s contributions are twofold. First, it closes the complexity gap for arbitrary finite poset games, showing that they are as hard as any problem solvable with polynomial space. Second, it introduces a versatile gadget‑based reduction technique that may be adapted to prove PSPACE‑completeness for other impartial games with structural constraints.

In the discussion, the authors note that while the general problem is intractable, many natural subclasses (e.g., posets of bounded width, tree‑like posets, or those arising from specific combinatorial structures) might admit more efficient algorithms. They suggest future work on identifying such tractable subclasses, developing approximation schemes, or exploring parameterised complexity with respect to poset height, width, or other structural measures.

Overall, the paper provides a rigorous and elegant proof that the winner‑determination problem for arbitrary finite poset games is PSPACE‑complete, thereby deepening our understanding of the computational boundaries of impartial combinatorial games.