EconCSLib.GameTheory.StrategicGame.ZeroSum.OptimalStrategySetPolytope #
For a finite matrix game A, the sets of optimal mixed row and column
strategies are nonempty polytopes in the ambient mixed-strategy space.
We supply the four structural properties that together characterise an
H-polytope (Mathlib does not yet ship a packaged IsPolytope):
- Nonempty, via the minimax theorem.
- Convex, via Mathlib's
Convex.interand the half-space form. - Closed, via continuity of the defining linear functionals.
- Compact, as a closed subset of the compact standard simplex.
The deliverable is stated on the image of MatrixGame.optimalRowStrategies
in I → ℝ (under the subtype value map), which is the natural ambient
vector space for polytope vocabulary.
Main results #
MatrixGame.image_optimalRowStrategies_eq— H-representation:Subtype.val '' A.optimalRowStrategies = stdSimplex ℝ I ∩ ⋂ j, {f | A.value ≤ ∑ i, f i * A.g i j}.MatrixGame.optimalRowStrategies_image_isPolytope— bundles the convex / closed / compact / nonempty conclusion.- Symmetric results for
optimalColumnStrategies.
References #
- [LRS] Laraki, Renault, Sorin, Mathematical Foundations of Game Theory, Proposition 2.4.1(a).
H-representation of the row-optimal set #
H-representation of the row-optimal strategy set in the ambient
space I → ℝ: the standard simplex intersected with the finitely
many half-spaces A.value ≤ ∑ i, f i * A.g i j, one per pure
column j.
Equations
Instances For
H-representation of the column-optimal strategy set in the ambient
space J → ℝ.
Equations
Instances For
The image of the row-optimal subtype set equals the H-representation
polytope A.optimalRowSet.
Convex / closed / compact / nonempty #
The row-optimal strategy set is a nonempty polytope: convex, closed, compact, and nonempty.