THEOREM 3 — a closed convex set is approachable iff it meets every T(q); otherwise it is excludable with gₙ ≡ q₀
ProvedVectorPayoffs.Convex.convex_approachable_iffapproachabilityexcludabilityminimaxp2o-batch-p200ap2o-gran-per-chapterp2o-plan-paperp2o-v1vector-payoffs
Let be an matrix () of probability distributions on a closed bounded convex , and for let be the convex hull of the points . Let be closed and convex. Then
- is approachable in if and only if for every ;
- if for some , then is excludable in with the stationary strategy .
In particular every closed convex set is either approachable or excludable. This is the paper's vector analogue of von Neumann's minimax theorem.
Formalization Note may be unbounded or empty (the empty set meets no , is not approachable because its distance is , and is excludable). is added so that both players have strategies; the approachability notion quantifies over all plays of the strategies, on every probability space in Type.
Preamble
import Mathlib import Definitions.Def_VectorPayoffs_Convex_Game open MeasureTheory
Formal statement
namespace VectorPayoffs.Convex
/-- Blackwell (1956), §3, p. 6, THEOREM 3: a closed convex set `S` is approachable if and only
if it intersects every set `T(q)`; if it fails to intersect `T(q₀)`, it is excludable with
`gₙ ≡ q₀`. -/
theorem convex_approachable_iff {N r s : ℕ} (G : Game N r s) (hr : 1 ≤ r) (hs : 1 ≤ s)
(S : Set (E N)) (hS : IsClosed S) (hSc : Convex ℝ S) :
(G.ApproachableIn S ↔ ∀ q ∈ stdSimplex ℝ (Fin s), (S ∩ G.T q).Nonempty) ∧
∀ (q₀ : Fin s → ℝ) (hq₀ : q₀ ∈ stdSimplex ℝ (Fin s)), S ∩ G.T q₀ = ∅ →
G.ExcludableWith S (Strategy.const q₀ hq₀) := by sorry
end VectorPayoffs.Convex
Source
Blackwell, An analog of the minimax theorem for vector payoffs, Pacific J. Math. 6(1), 1956, p. 6, THEOREM 3
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.