arXiv Science⌕ Search

arXiv · 2610.19646

The Big Match under translation-invariant charges

Abstract

We study two-player zero-sum stochastic games in which a stream of stage payoffs is evaluated by a translation-invariant charge on the set of stages. Every patient valuation, that is, every diffuse charge that obeys the time value of money principle, is translation-invariant, and by Mertens and Neyman each player has, for every $\varepsilon>0$, a strategy that is $\varepsilon$-optimal for all patient valuations simultaneously. We show that this robustness breaks down for translation-invariant charges. In the Big Match with $0$--$1$ payoffs, whose uniform value is $1/2$, every strategy of Player 1 is worthless: for every strategy of Player 1 and every $δ>0$, Player 2 has a pure reply, and there is a translation-invariant charge under which Player 1's payoff is less than $δ$. Consequently, the Big Match has no value with respect to all translation-invariant charges. This contrasts with finite Markov decision processes, where every stationary Blackwell optimal strategy is optimal for every translation-invariant charge. We pose questions on robust guarantees in general stochastic games and on optimality with respect to a fixed translation-invariant charge.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Abraham Neyman. 2026-10-14. The Big Match under translation-invariant charges. https://arxiv.org/abs/2610.19646

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

A Safe Approximation Based on Mixed-Integer Optimization for Non-Convex Distributional Robustness Governed by Univariate Indicator Functions

In this work, we present an algorithmically tractable safe approximation of distributionally robust optimization (DRO) problems that contain univariate indicator functions. The latter appear in different applications, but render the model nonlinear and nonconvex. The considered ambiguity sets can exploit moment information. Typically, reformulation approaches using duality theory need to make strong assumptions on the structure of the underlying constraints, such as convexity in the decisions or concavity in the uncertainty which cannot be assumed in our setting. We nevertheless present a semi-infinite minorant of the adversarial problem that admits a zero-gap dualization and is subsequently approximated by a discretized counterpart. Under mild assumptions, the latter provides a safe approximation that is a mixed-integer linear problem and can be solved by available standard software. Obtained solutions are guaranteed to be feasible for the original distributionally robust problem. We prove that the approximation of the adversarial problem converges to its true value for increasingly fine discretization, and that the values of the solutions of the safe approximations converge to the value of the true DRO problem under mild assumptions. In addition, the error bound yields an a posteriori enclosure of the true optimal value that can be evaluated in practice. On the practical side, the approach is made concrete for a challenging, fundamental task in material design, namely in particle separation. Computational results for a realistic setting show that the safe approximation meets the imposed quality requirement for every distribution in the ambiguity set, whereas the unprotected solution does not.

math.OC↗

Geometric Certificates for Deterministic Global Linear Convergence of Overparameterized Projected Gradient Descent

We study projected gradient descent (ProjGD) for the rank-constrained problem $\min\{f(X): \operatorname{rank}(X)\le r\}$ when the unconstrained minimizer $X_*$ has rank $r_*\le r$. Our main contribution is a quantitative geometric certificate relating the normal component of $\nabla f(X)$ to the smallest nonzero singular value of $X$. Under a rank-$(r+r_*)$ restricted curvature condition with constants $0<μ \sqrt{r_*/r}(L-μ)/(2Lμ)$. With an additional rank-$2r$ restricted smoothness assumption, we establish deterministic global linear convergence for $\sqrt{r_*/r}(L-μ)/(2Lμ)<η<1/L$. The result accommodates arbitrary rank overparameterization $r\ge r_*$ and is independent of the singular-value condition number of $X_*$. As a consequence, we obtain an alternative proof of the sharp threshold $δ<(1+\sqrt{r_*/r})^{-1}$ for the absence of spurious local minima in the asymmetric setting, where $δ=(L-μ)/(L+μ)$. This threshold and matching asymmetric examples were established in recent work.

math.OC↗

Congestion-Aware Charging Coordination for Electric Ride-Hailing Fleets under Stochastic Demand

In this study, we develop a dynamic charging scheduling method that integrates day-ahead charging planning and online reactive vehicle dispatch and charger assignment. The day-ahead planning model minimizes expected charging costs by considering anticipated waiting times and overnight energy replenishment costs. The online model then adapts this plan by maximizing the profit from vehicle dispatching and assigning vehicles to chargers through a reservation-based charging request scheduler to minimize the total charging-related operational time. The developed charging policy is tested on a Manhattan-like study area using synthetic data drawn from NYC yellow taxi data with a fleet size of 100 vehicles given the scenarios of 3000 and 4000 customers/day. The computational results show that our method outperforms different benchmark policies with up to +21.68% profit and +22.22% service rate for 4000 customers relative to the weakest benchmark. Sensitivity analyses are conducted for different system parameters, and managerial insights are discussed.

math.OC↗