[Paper Review] On the complexity of strong Nash equilibrium: Hard-to-solve instances and smoothed complexity

Nicola Gatti, Marco Rocco+1 more|arXiv (Cornell University)|Apr 4, 2013

Game Theory and Applications15 references3 citations

TL;DR

This paper establishes that mixed-strategy strong Nash equilibria (SNE) in strategic-form games require payoff vectors over supports to lie on a specific geometric structure—collinear in 2-player games, lying on an (n−1)-dimensional hyperplane in n-player games. Leveraging this, the authors construct worst-case instances for support-enumeration algorithms and prove that SNE computation is in smoothed polynomial time, as generic games admit only pure-strategy SNEs, making the problem tractable under small random perturbations.

ABSTRACT

The computational characterization of game-theoretic solution concepts is a central topic in artificial intelligence, with the aim of developing computationally efficient tools for finding optimal ways to behave in strategic interactions. The central solution concept in game theory is Nash equilibrium (NE). However, it fails to capture the possibility that agents can form coalitions (even in the 2-agent case). Strong Nash equilibrium (SNE) refines NE to this setting. It is known that finding an SNE is NP-complete when the number of agents is constant. This hardness is solely due to the existence of mixed-strategy SNEs, given that the problem of enumerating all pure-strategy SNEs is trivially in P. Our central result is that, in order for a game to have at least one non-pure-strategy SNE, the agents' payoffs restricted to the agents' supports must, in the case of 2 agents, lie on the same line, and, in the case of n agents, lie on an (n - 1)-dimensional hyperplane. Leveraging this result, we provide two contributions. First, we develop worst-case instances for support-enumeration algorithms. These instances have only one SNE and the support size can be chosen to be of any size-in particular, arbitrarily large. Second, we prove that, unlike NE, finding an SNE is in smoothed polynomial time: generic game instances (i.e., all instances except knife-edge cases) have only pure-strategy SNEs.

Motivation & Objective

• To understand the structural conditions under which mixed-strategy strong Nash equilibria (SNE) exist in strategic-form games.
• To address the lack of efficient algorithms for finding SNEs, especially in general games beyond pure strategies.
• To develop worst-case benchmark instances for support-enumeration algorithms used in SNE computation.
• To analyze the smoothed complexity of SNE computation and show that generic games have only pure-strategy SNEs.

Proposed method

• Identify geometric constraints: for mixed-strategy SNE to exist, payoff vectors restricted to agents' supports must lie on a line (2-player) or an (n−1)-dimensional hyperplane (n-player).
• Construct worst-case game instances with exactly one SNE and arbitrary support size, ensuring exponential time for support-enumeration algorithms.
• Apply smoothed analysis to show that, under small random perturbations, the probability of having mixed-strategy SNEs is zero.
• Use a two-phase algorithm: first check all pure-strategy supports; if no SNE is found, verify whether any mixed-strategy support profiles satisfy the geometric condition.
• Leverage the fact that perturbations almost surely break the alignment required for mixed SNEs, ensuring only pure SNEs exist generically.
• Prove that the algorithm runs in smoothed polynomial time by showing that exponential support enumeration is avoided with high probability under perturbations.

Experimental results

Research questions

• RQ1 Under what geometric conditions on payoff vectors over supports can a mixed-strategy SNE exist in n-player games?
• RQ2 Can we construct game instances that are worst-case for support-enumeration algorithms for SNE computation?
• RQ3 Is the problem of finding an SNE in generic games computationally tractable under small random perturbations?
• RQ4 How does the smoothed complexity of SNE computation compare to that of Nash equilibrium?
• RQ5 Are there structural properties that make SNE computation robust or fragile under perturbations?

Key findings

• Mixed-strategy SNEs exist only if the payoff vectors over the supports of agents lie on a line in 2-player games or on an (n−1)-dimensional hyperplane in n-player games.
• Worst-case game instances can be constructed with exactly one SNE and support sizes up to m/2 per agent, requiring O(4^{m/2}) time for support-enumeration algorithms.
• The problem of finding an SNE is in smoothed polynomial time, as generic games (except for knife-edge cases) have only pure-strategy SNEs.
• The probability that a small random perturbation preserves a mixed-strategy SNE is zero, due to the high codimension of the required geometric alignment.
• Support-enumeration algorithms for SNE are guaranteed to run in polynomial time under smoothed analysis, as exponential enumeration occurs only with zero-probability events.
• Unlike Nash equilibrium, SNE computation is robust to small perturbations, as perturbed games generically admit only pure-strategy SNEs.