Graph-based approach for enumerating the Nash equilibria of a two-player extensive-form game - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2023

Graph-based approach for enumerating the Nash equilibria of a two-player extensive-form game

Paolo Zappalà
Amal Benhamiche
  • Fonction : Auteur
  • PersonId : 1077187
Matthieu Chardy
  • Fonction : Auteur
  • PersonId : 964673
Francesco De Pellegrini
  • Fonction : Auteur
  • PersonId : 1044331
Rosa Figueiredo

Résumé

Enumerating the Nash equilibria of a game is a known problem of game theory. We take into account twoplayer extensive-form games with perfect information and perfect recall. Based on the notion of compatibility of strategies, a new graph representation permits to devise a new lightweight algorithm to enumerate all of its Nash equilibria. It is the first of its kind not to use brute force. The method provides also intermediate results, like upper (lower) bounds to the value of the utility of Nash equilibria or the enumeration of Nash equilibria that fulfill any given constraints (possibly, unconstrained) on their values. We compare our method to the only existing formulation providing an upper bound to the value of the utility of any Nash equilibrium. The experiments show that our method is faster by some orders of magnitude. We also test the method to enumerate the Nash equilibria on a new library, that we introduce as benchmark for representing all structures and properties of extensive-form games.
Fichier principal
Vignette du fichier
Graph_HAL.pdf (631.62 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04093334 , version 1 (10-05-2023)

Identifiants

  • HAL Id : hal-04093334 , version 1

Citer

Paolo Zappalà, Amal Benhamiche, Matthieu Chardy, Francesco De Pellegrini, Rosa Figueiredo. Graph-based approach for enumerating the Nash equilibria of a two-player extensive-form game. 2023. ⟨hal-04093334⟩
46 Consultations
34 Téléchargements

Partager

Gmail Facebook X LinkedIn More