Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 8 — the core of a convex game is its unique von Neumann–Morgenstern stable set

Proved
CoresConvexGames.Stability.core_is_unique_stable_set

by mikedeng1 · Sep 27, 2026 · Mathlib 0df444a (Lean v4.33.1)

convex-gamecooperative-gamecorep2o-batch-p100ap2o-gran-per-chapterp2o-plan-paperp2o-v1stable-set

Let N={1,…,n}N=\{1,\dots,n\}N={1,…,n} be a finite set of players and v:2N→Rv:2^N\to\mathbb Rv:2N→R a game, i.e. a set function with v(∅)=0v(\emptyset)=0v(∅)=0. For a payoff vector a∈RNa\in\mathbb R^Na∈RN and a coalition S⊆NS\subseteq NS⊆N write a(S)=∑i∈Saia(S)=\sum_{i\in S}a_ia(S)=∑i∈S​ai​. Suppose vvv is convex: v(S)+v(T)≤v(S∪T)+v(S∩T)v(S)+v(T)\le v(S\cup T)+v(S\cap T)v(S)+v(T)≤v(S∪T)+v(S∩T) for all S,T⊆NS,T\subseteq NS,T⊆N. A payoff vector aaa is feasible if a(N)≤v(N)a(N)\le v(N)a(N)≤v(N). The core CCC is the set of feasible aaa with a(S)≥v(S)a(S)\ge v(S)a(S)≥v(S) for all S⊆NS\subseteq NS⊆N. A vector aaa dominates bbb if there is a nonempty coalition SSS with a(S)≤v(S)a(S)\le v(S)a(S)≤v(S) and ai>bia_i>b_iai​>bi​ for all i∈Si\in Si∈S. A set VVV of feasible vectors is stable if every feasible vector is either a member of VVV or dominated by a member of VVV, but not both.

Then the core of vvv is stable, and it is the only stable set:

C is stable,andV stable  ⟹  V=C.C \text{ is stable},\qquad\text{and}\qquad V \text{ stable} \implies V=C .C is stable,andV stable⟹V=C.

That is, the core of a convex game is its unique von Neumann–Morgenstern solution. For general games stable sets need not exist and need not be unique, so convexity singles out a class where the two classical solution concepts coincide.

Formalization Note Players are Fin n (a relabelling of Shapley's arbitrary finite NNN), a game is f : Finset (Fin n) → ℝ, and the core is the published Supermodularity.Cooperative.Core Finset.univ f. Convexity is the published IsConvexGame f, which includes v(∅)=0v(\emptyset)=0v(∅)=0. Stable sets are defined over feasible vectors, as on the page (a footnote notes the classical imputation-based variant gives the same answer for the core).

Preamble
import Mathlib
import Definitions.Def_Supermodularity_Cooperative_IsConvexGame
import Definitions.Def_Supermodularity_Cooperative_Core
import Definitions.Def_CoresConvexGames_Stability_IsStableSet
Formal statement
namespace CoresConvexGames.Stability

open Supermodularity.Cooperative

/-- Shapley (1971), p. 24, Theorem 8: the core of a convex game is stable, and it is the
unique stable set (von Neumann–Morgenstern solution). -/
theorem core_is_unique_stable_set {n : ℕ} (f : Finset (Fin n) → ℝ) (hf : IsConvexGame f) :
    IsStableSet f (Core Finset.univ f) ∧
      ∀ V : Set (Fin n → ℝ), IsStableSet f V → V = Core Finset.univ f := by sorry

end CoresConvexGames.Stability
Source
Shapley, Cores of Convex Games, Int. J. Game Theory 1, 1971, https://doi.org/10.1007/BF01753431, p. 24, §4.3, Theorem 8
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

Throughout, nnn is a natural number and N={0,1,…,n−1}N = \{0, 1, \dots, n-1\}N={0,1,…,n−1} is a set of nnn players. Three names in the statement come from the accompanying definition bundle, and their Lean bodies are not part of the code under review:

  • "convex game" (IsConvexGame)
  • "core relative to a player set" (Core)
  • "stable set" (IsStableSet)

This read-back can only describe them by name and by type. Any further meaning they carry, such as supermodularity, the normalisation f(∅)=0f(\emptyset)=0f(∅)=0, efficiency, individual rationality, or internal and external stability, depends entirely on those definitions. The statement itself does not show it.

Data. The statement takes any function

f:2N→R,f : 2^{N} \to \mathbb{R},f:2N→R,

which assigns a real number f(S)f(S)f(S) to every subset S⊆NS \subseteq NS⊆N, including S=∅S = \emptysetS=∅ and S=NS = NS=N. The statement does not itself require f(∅)=0f(\emptyset) = 0f(∅)=0 or monotonicity.

Hypothesis. fff is a convex game in the sense of the bundle's definition IsConvexGame. This is the only hypothesis.

Notation. Write

C(f)⊆RNC(f) \subseteq \mathbb{R}^{N}C(f)⊆RN

for the set Core gives when its player-set argument is the full set NNN and the game is fff. C(f)C(f)C(f) is a set of payoff vectors x=(xi)i∈Nx = (x_i)_{i \in N}x=(xi​)i∈N​.

Conclusion. The theorem asserts two things together:

  1. C(f)C(f)C(f) is a stable set for fff, in the sense of IsStableSet.
  2. For every subset V⊆RNV \subseteq \mathbb{R}^{N}V⊆RN, if VVV is a stable set for fff, then
V=C(f).V = C(f).V=C(f).

This is exact equality of sets. The quantifier ranges over all subsets of RN\mathbb{R}^NRN. The only restriction on VVV is whatever IsStableSet itself imposes; the statement adds none, such as requiring VVV to lie among the imputations.

So under the hypothesis, C(f)C(f)C(f) is stable and is the only stable set. The statement does not separately say that C(f)C(f)C(f) is nonempty. It follows only if IsStableSet rules out the empty set, or if its conditions force the empty set to fail them.

Degenerate cases.

  • n=0n = 0n=0. There are no players. The only coalition is ∅\emptyset∅, and RN\mathbb{R}^{N}RN has exactly one point, the empty vector. fff is then a single real number f(∅)f(\emptyset)f(∅). If that fff is a convex game, the theorem says C(f)C(f)C(f) is stable and is the only stable set. Since there is one point, C(f)C(f)C(f) is either that point or the empty set; which one depends on how Core treats f(∅)f(\emptyset)f(∅).
  • n=1n = 1n=1. There are two coalitions, ∅\emptyset∅ and {0}\{0\}{0}, and payoff vectors are single real numbers.
  • Empty core. Nothing in the statement stops C(f)C(f)C(f) from being empty unless the definitions do. If it is empty, the theorem says the empty set is stable and that every stable set is empty.
  • Vacuous hypothesis. If IsConvexGame can never be satisfied by some fff, the theorem says nothing about that fff.

The statement uses no division, subtraction of natural numbers, integrals, suprema or extended-real arithmetic. Any junk-value behaviour of that kind could only come from inside the three bundle definitions.

Human review
  • Endorsed by Shuze Chen · Sep 27, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Sep 27, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me