Correlated optimin

Mehmet Mars Seven

arxiv: 2605.19129 · v1 · pith:7TJBDQE3new · submitted 2026-05-18 · 💰 econ.TH · cs.GT

Correlated optimin

Mehmet Mars Seven This is my paper

Pith reviewed 2026-05-20 07:16 UTC · model grok-4.3

classification 💰 econ.TH cs.GT

keywords correlated equilibriumoptiminfinite gamesPareto optimalityguaranteed payoffszero-sum gamescorrelated distributions

0 comments

The pith

Every finite game has a correlated optimin that improves players' guaranteed payoffs over correlated equilibria.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper extends the optimin notion from mixed strategy profiles to correlated distributions over action profiles in finite games. A correlated distribution is assessed by the worst expected payoff each player can secure when opponents may obey their recommendations or switch to a strictly profitable unilateral deviation given the posterior beliefs induced by the distribution. The authors establish that correlated optimins, defined as the Pareto optimal distributions under this vector of guaranteed payoffs, exist in every finite game. They further show that for any correlated equilibrium one can always find a correlated optimin delivering weakly higher guaranteed payoffs to every player. A sympathetic reader would care because this supplies a selection device among distributions that protects each participant against certain forms of exploitation at least as well as equilibrium play, and sometimes strictly better.

Core claim

What carries the argument

Correlated optimin: a Pareto-optimal correlated distribution with respect to the vector of each player's guaranteed payoff, where the guarantee is the worst-case expected payoff against unilateral recommendation-contingent deviations profitable under the induced posterior.

If this is right

In two-player zero-sum games, correlated optimin coincides with correlated equilibrium and yields the maximin value.
Outside zero-sum games, correlated optimin may strictly improve upon all correlated equilibria.
In a simple 2x2 game possessing a unique correlated equilibrium that is also the unique coarse correlated equilibrium, a correlated optimin strictly Pareto dominates the equilibrium payoffs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The existence result suggests that recommendation devices or mediators could be designed to target correlated optimins rather than equilibria to improve participants' security levels.
The strict improvement outside zero-sum games indicates that equilibrium selection problems in applied economic models might be addressed by searching within the set of correlated optimins.
Because the concept recovers the standard value in zero-sum settings, it provides a conservative generalization that preserves classical results while allowing gains in other environments.

Load-bearing premise

The results rest on evaluating distributions via the specific notion of unilateral recommendation-contingent deviations that are strictly profitable under the posterior induced by the distribution.

What would settle it

A finite game in which no correlated optimin exists, or in which for some correlated equilibrium every correlated optimin fails to deliver weakly higher guaranteed payoffs to all players.

read the original abstract

We extend the optimin notion of Ismail (2025) from mixed strategy profiles to correlated distributions. A correlated distribution is evaluated by the worst expected payoff each player can receive when opponents may either obey their private recommendations or make unilateral recommendation-contingent deviations that are strictly profitable under the posterior induced by the distribution. Correlated optimins are Pareto optimal with respect to this vector of guaranteed payoffs. We show that a correlated optimin exists in every finite game. In addition, for every correlated equilibrium, there exists a correlated optimin such that every player's guaranteed payoff is weakly higher than his or her correlated equilibrium payoff. In two-player zero-sum games, correlated optimin coincides with correlated equilibrium and yields the maximin value. Outside zero-sum games, correlated optimin may strictly improve upon all correlated equilibria. We illustrate this with a simple 2x2 game with a unique correlated and coarse correlated equilibrium, in which there exists a correlated optimin that strictly Pareto dominates the equilibrium payoff.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Correlated optimin extends optimin to correlated distributions with an existence claim and a strict improvement over correlated equilibrium in a 2x2 non-zero-sum example.

read the letter

The key takeaway is that this paper introduces correlated optimin, proves it exists for all finite games, and shows it can give players weakly higher guaranteed payoffs than any correlated equilibrium, with a strict Pareto improvement in a simple non-zero-sum example. What is new is the move from mixed strategies in Ismail (2025) to correlated distributions. The evaluation uses the worst expected payoff for each player when opponents can stick to the recommendation or make a unilateral deviation that is strictly profitable under the posterior. They define correlated optimins as the Pareto optimal vectors under this criterion. The existence result and the improvement property are the main claims, backed by the 2x2 game where the unique correlated equilibrium is dominated. The paper does a good job making the improvement visible with the example and noting the reduction to standard correlated equilibrium in zero-sum games. This keeps the contribution focused. The soft spot is the potential problem with upper semi-continuity of the guaranteed payoff map. The strict profitability threshold can cause jumps in the worst-case payoffs as the distribution varies, which might mean the set of attainable vectors is not closed and therefore lacks maximal elements. The existence proof must address this directly, perhaps by showing the map is usc after all or by some other compactness argument. If it does not, the central claim would need revision. This work is for specialists in game theory who look at refinements and correlated solution concepts. Readers who want alternatives to correlated equilibrium that raise security levels will find it relevant. It is worth a serious referee report to check the proofs and the handling of the continuity issue. I would send it to peer review.

Referee Report

1 major / 2 minor

Summary. The manuscript extends the optimin notion of Ismail (2025) from mixed-strategy profiles to correlated distributions over finite normal-form games. A correlated distribution μ is evaluated via the vector v(μ) of guaranteed payoffs, where each player i receives the worst-case expected payoff when opponents may follow their recommendations or select any unilateral recommendation-contingent deviation that is strictly profitable under the posterior induced by μ. Correlated optimins are the Pareto-optimal points of the image of v. The paper asserts existence of a correlated optimin in every finite game and, for any correlated equilibrium, the existence of a correlated optimin that weakly improves every player's guaranteed payoff. In two-player zero-sum games the concept coincides with correlated equilibrium (and the maximin value); outside zero-sum games it can strictly Pareto-dominate all correlated equilibria, as illustrated by a 2×2 game with a unique correlated equilibrium.

Significance. If the existence and improvement results hold, the paper supplies a new correlated solution concept that is Pareto optimal with respect to a well-defined guaranteed-payoff vector and that weakly dominates correlated equilibrium in that vector. The explicit 2×2 illustration of strict improvement and the reduction to maximin value in zero-sum games are concrete strengths. The extension from Ismail (2025) is presented as a genuine generalization rather than a circular reuse.

major comments (1)

[Existence theorem and proof of existence] The existence theorem (presumably the main result in §3 or §4) asserts that a Pareto-maximal element of the image of μ ↦ v(μ) exists for every finite game. The stress-test concern is load-bearing here: because the predicate “strictly profitable” partitions the simplex into open regions separated by hyperplanes where expected gain equals zero, the opponent deviation set can enlarge discontinuously when crossing such a hyperplane. This produces downward jumps in v_i(μ), violating upper semi-continuity. A non-usc map on a compact domain need not have closed image, so the partially ordered set of attainable guaranteed-payoff vectors may lack maximal elements. The proof must either establish upper semi-continuity after all, exhibit an alternative compactness argument, or construct the maximizer directly (e.g., by enumerating the finitely many regions induced by the finite action sets).

minor comments (2)

[Definition of correlated optimin (early section)] In the definition of v(μ), clarify whether a single opponent or all opponents may deviate simultaneously; the abstract’s phrasing “opponents may either obey … or make unilateral … deviations” is ambiguous on this point.
[Illustrative example] The 2×2 example should display the explicit payoff matrix in the main text (or at least in an appendix) so that readers can verify the unique correlated equilibrium and the claimed strict improvement.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed and insightful report. The concern regarding potential discontinuities in the guaranteed-payoff map v is well-taken and we address it directly below by revising the existence argument to exploit the finite action spaces rather than relying on upper semi-continuity.

read point-by-point responses

Referee: The existence theorem asserts that a Pareto-maximal element of the image of μ ↦ v(μ) exists for every finite game. The stress-test concern is load-bearing here: because the predicate “strictly profitable” partitions the simplex into open regions separated by hyperplanes where expected gain equals zero, the opponent deviation set can enlarge discontinuously when crossing such a hyperplane. This produces downward jumps in v_i(μ), violating upper semi-continuity. A non-usc map on a compact domain need not have closed image, so the partially ordered set of attainable guaranteed-payoff vectors may lack maximal elements. The proof must either establish upper semi-continuity after all, exhibit an alternative compactness argument, or construct the maximizer directly (e.g., by enumerating the finitely many regions induced by the finite action sets).

Authors: We agree that v may fail to be upper semi-continuous because the set of admissible unilateral recommendation-contingent deviations can jump discontinuously across the hyperplanes where expected gain equals zero, producing downward jumps in the min. We therefore revise the existence proof to avoid any appeal to upper semi-continuity. Because every player has a finite action set, each opponent possesses only finitely many possible recommendation-contingent deviation functions. This induces a finite collection of possible subsets S of active deviations. For each such subset S we consider the corresponding closed polyhedral region consisting of all μ at which every deviation in S yields non-negative expected gain for its owner. On this compact set the restricted map v_S (minimum payoff over exactly the fixed finite collection S) is continuous, being the pointwise minimum of finitely many continuous functions of μ. Its image is therefore compact and possesses Pareto-maximal elements. We then select, among the finitely many candidate maximal vectors obtained from all such regimes, a vector that is Pareto-undominated in the overall attainable set. Any correlated optimin must arise in one of these regimes, so the selected vector is a correlated optimin for the original game. The revised proof will appear in the next version of the manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity; derivation self-contained via new evaluation map on finite action spaces

full rationale

The paper defines correlated optimin directly from the vector-valued map v(μ) of guaranteed payoffs under the paper's own unilateral deviation predicate, then asserts existence of Pareto-maximal points and weak improvement over correlated equilibria. This construction does not reduce any claimed result to a fitted parameter, a self-referential definition, or a load-bearing self-citation chain; the Ismail (2025) reference supplies only the original optimin notion for mixed profiles and is extended rather than reused to force the correlated case. The finite-game setting supplies compactness of the simplex and finiteness of deviation sets, allowing the existence argument to rest on the paper's definitions without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The paper relies on standard finite-game assumptions and introduces the correlated optimin as a new selection criterion; no free parameters or invented physical entities appear.

axioms (1)

domain assumption Every finite normal-form game has finite action sets and real-valued payoff functions.
Invoked implicitly for the existence claim in every finite game.

invented entities (1)

Correlated optimin no independent evidence
purpose: Pareto-optimal correlated distribution under the worst-case guaranteed-payoff vector against recommendation-contingent deviations.
Newly defined in the paper; no independent evidence outside the definition is provided.

pith-pipeline@v0.9.0 · 5682 in / 1407 out tokens · 46714 ms · 2026-05-20T07:16:03.756380+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1]

Aumann, R. J. (1974). Subjectivity and correlation in randomized strategies. Journal of Mathematical Economics\/ 1\/ (1), 67--96

work page 1974
[2]

Ismail, M. S. (2025). Super-nash performance. International Economic Review\/ 66\/ (4), 1487--1503

work page 2025
[3]

and J.-P

Moulin, H. and J.-P. Vial (1978). Strategically zero-sum games: the class of games whose completely mixed equilibria cannot be improved upon. International Journal of Game Theory\/ 7\/ (3--4), 201--221

work page 1978
[4]

Nash, J. F. (1951). Non-cooperative games. Annals of Mathematics\/ 54\/ (2), 286--295

work page 1951

[1] [1]

Aumann, R. J. (1974). Subjectivity and correlation in randomized strategies. Journal of Mathematical Economics\/ 1\/ (1), 67--96

work page 1974

[2] [2]

Ismail, M. S. (2025). Super-nash performance. International Economic Review\/ 66\/ (4), 1487--1503

work page 2025

[3] [3]

and J.-P

Moulin, H. and J.-P. Vial (1978). Strategically zero-sum games: the class of games whose completely mixed equilibria cannot be improved upon. International Journal of Game Theory\/ 7\/ (3--4), 201--221

work page 1978

[4] [4]

Nash, J. F. (1951). Non-cooperative games. Annals of Mathematics\/ 54\/ (2), 286--295

work page 1951