MatrixGame.optimal_pairs_iff_saddle_point
theoremverifiedformalizationimported Lean
Optimal pairs ↔ saddle points. A pair of mixed strategies is in the product X(A) × Y(A) of optimal strategy sets iff it is a saddle point.
Statement
set_option autoImplicit true in set_option relaxedAutoImplicit true in open MatrixGame in open Finset BigOperators in set_option linter.unusedSectionVars false in universe u in variable {I J : Type u} [Fintype I] [Fintype J] [Nonempty I] [Nonempty J] in variable (A : MatrixGame I J ℝ) in theorem MatrixGame.optimal_pairs_iff_saddle_point (xx : stdSimplex ℝ I) (yy : stdSimplex ℝ J) : (xx ∈ A.optimalRowStrategies ∧ yy ∈ A.optimalColumnStrategies) ↔ A.IsSaddlePoint xx yy
- Formalizes
- EconCSLib, `MatrixGame.optimal_pairs_iff_saddle_point`
- Lean proof
- By xbei (from the file's git history), from github.com/gametheoryinlean/EconCSLib at
1a88f809b538(Copyright (c) 2026 EconCSLib contributors. All rights reserved, Apache-2.0). Compiled here against leanprover/lean4:v4.33.1, Mathlib v4.33.1; written for leanprover/lean4:v4.30.0.MatrixGame.optimal_pairs_iff_saddle_point← MatrixGame.optimal_pairs_iff_saddle_point (EconCSLib/GameTheory/StrategicGame/ZeroSum/MatrixGameNash.lean:343): verbatim
- Verified
- 08 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 24
- MatrixGame.value_eq_minimaxMatrixGame.value_eq_minimax
- MatrixGame.mem_optimalColumnStrategies_iff_E_leA column strategy is optimal iff against every mixed row it caps payoff at the value.
- MatrixGameA finite two-player zero-sum matrix game. Player I chooses row
i : I, Player II chooses columnj : J. Payoff to Player I isg i j… - MatrixGame.optimalRowStrategiesThe set of optimal row strategies: mixed strategies that achieve the maximin value.
- MatrixGame.optimalColumnStrategiesThe set of optimal column strategies: mixed strategies that achieve the minimax value.
- MatrixGame.IsSaddlePointA pair of mixed strategies is a saddle point when player I's payoff is maximised by the row strategy against the column strategy and…
- MatrixGame.valueThe value of a matrix game (
sSup-based; needs order completeness). Equal to both maximin and minimax by the minimax theorem… - MatrixGame.EExpected payoff when Player I uses mixed strategy
xand Player II usesy. - MatrixGame.mem_optimalRowStrategies_iff_E_geA row strategy is optimal iff against every mixed column it secures the value.
- MatrixGame.EjExpected payoff when Player I uses
xagainst pure columnj. - stdSimplex.purePoint-mass simplex element at
i₀. - wsumWeighted sum of
fwith weights from a simplex elementx. Thinabbrevover Mathlib's finite dot product⬝ᵥ: definitionally equal,… - wsum_pure_applyWeighted sum at a point mass evaluates the chosen coordinate.
- MatrixGame.EiExpected payoff when pure row
ifaces Player II's mixed strategyy. - MatrixGame.guarantee_IPlayer I's guaranteed payoff using mixed strategy
x: the minimum expected payoff over all of Player II's pure responses. A finite… - MatrixGame.guarantee_IIPlayer II's guaranteed loss using mixed strategy
y: the maximum expected payoff (for Player I) over all of Player I's pure responses. - MinimaxLoomis.lam.auxPlayer I's guaranteed payoff from mixed strategy
x: the minimum over pure columns of the expected payoff. - MinimaxLoomis.lam.aux.bddAbove
lam.aux Ais uniformly bounded above on the simplex. - MatrixGame.minimaxThe minimax value: the best guarantee Player II can achieve.
minimax = inf_y sup_i E(i, y) - MinimaxLoomis.mu.auxPlayer II's maximum loss against mixed strategy
y: the maximum over pure rows of the expected payoff. - MinimaxLoomis.mu.aux.bddBelow
mu.aux Ais uniformly bounded below on the simplex. - stdSimplex.pure_applystdSimplex.pure_apply
- Loomis.lamB.aux.bddAbove
lamB.aux A Bis bounded above on the simplex (continuous function on a compact set). - Loomis.muB.aux.bddBelow
muB.aux A Bis bounded below on the simplex.
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
set_option autoImplicit true inset_option relaxedAutoImplicit true inopen MatrixGame inopen Finset BigOperators inset_option linter.unusedSectionVars false inuniverse u invariable {I J : Type u} [Fintype I] [Fintype J] [Nonempty I] [Nonempty J] invariable (A : MatrixGame I J ℝ) in/-- **Optimal pairs ↔ saddle points.** A pair of mixed strategies is in the product `X(A) × Y(A)` of optimal strategy sets iff it is a saddle point. -/theorem MatrixGame.optimal_pairs_iff_saddle_point (xx : stdSimplex ℝ I) (yy : stdSimplex ℝ J) : (xx ∈ A.optimalRowStrategies ∧ yy ∈ A.optimalColumnStrategies) ↔ A.IsSaddlePoint xx yy := by rw [A.mem_optimalRowStrategies_iff_E_ge, A.mem_optimalColumnStrategies_iff_E_le] refine ⟨?_, ?_⟩ · rintro ⟨hxx, hyy⟩ have hE : A.E xx yy = A.value := le_antisymm (hyy xx) (hxx yy) refine ⟨fun x' => ?_, fun y' => ?_⟩ · calc A.E x' yy ≤ A.value := hyy x' _ = A.E xx yy := hE.symm · calc A.E xx yy = A.value := hE _ ≤ A.E xx y' := hxx y' · rintro ⟨hsad_row, hsad_col⟩ classical have hjE : ∀ j, A.E xx yy ≤ A.Ej xx j := by intro j have := hsad_col (stdSimplex.pure j) have heq : A.E xx (stdSimplex.pure j) = A.Ej xx j := by show wsum xx (fun i => wsum (stdSimplex.pure j) (A.g i)) = wsum xx (fun i => A.g i j) apply Finset.sum_congr rfl intro i _ show xx.val i * wsum (stdSimplex.pure j) (A.g i) = xx.val i * A.g i j rw [wsum_pure_apply] linarith [heq] have hiE : ∀ i, A.Ei i yy ≤ A.E xx yy := by intro i have := hsad_row (stdSimplex.pure i) have heq : A.E (stdSimplex.pure i) yy = A.Ei i yy := by show wsum (stdSimplex.pure i) (fun i' => wsum yy (A.g i')) = wsum yy (A.g i) rw [wsum_pure_apply] linarith [heq] have hGI : A.E xx yy ≤ A.guarantee_I xx := by show A.E xx yy ≤ Finset.inf' Finset.univ Finset.univ_nonempty (fun j => A.Ej xx j) rw [Finset.le_inf'_iff]; intro j _; exact hjE j have hGII : A.guarantee_II yy ≤ A.E xx yy := by show Finset.sup' Finset.univ Finset.univ_nonempty (fun i => A.Ei i yy) ≤ A.E xx yy rw [Finset.sup'_le_iff]; intro i _; exact hiE i have hI_le : A.guarantee_I xx ≤ A.value := le_ciSup (bddAbove_def.2 (by obtain ⟨C, hC⟩ := MinimaxLoomis.lam.aux.bddAbove A.g exact ⟨C, by rintro r ⟨z, rfl⟩; exact hC z⟩)) xx have hII_ge : A.value ≤ A.guarantee_II yy := by have hv := A.value_eq_minimax have : A.minimax ≤ A.guarantee_II yy := ciInf_le (bddBelow_def.2 (by obtain ⟨C, hC⟩ := MinimaxLoomis.mu.aux.bddBelow A.g exact ⟨C, by rintro r ⟨z, rfl⟩; exact hC z⟩)) yy linarith refine ⟨fun y' => ?_, fun x' => ?_⟩ · have hEval : A.E xx yy = A.value := by linarith calc A.value = A.E xx yy := hEval.symm _ ≤ A.E xx y' := hsad_col y' · have hEval : A.E xx yy = A.value := by linarith calc A.E x' yy ≤ A.E xx yy := hsad_row x' _ = A.value := hEval