English

Thresholds for sensitive optimality and Blackwell optimality in stochastic games

Computer Science and Game Theory 2025-06-24 v1

Abstract

We investigate refinements of the mean-payoff criterion in two-player zero-sum perfect-information stochastic games. A strategy is Blackwell optimal if it is optimal in the discounted game for all discount factors sufficiently close to 11. The notion of dd-sensitive optimality interpolates between mean-payoff optimality (corresponding to the case d=1d=-1) and Blackwell optimality (d=+d=+\infty). The Blackwell threshold αBw[0,1[\alpha_{\sf Bw} \in [0,1[ is the discount factor above which all optimal strategies in the discounted game are guaranteed to be Blackwell optimal. The dd-sensitive threshold αd[0,1[\alpha_{\sf d} \in [0,1[ is defined analogously. Bounding αBw\alpha_{\sf Bw} and αd\alpha_{\sf d} are fundamental problems in algorithmic game theory, since these thresholds control the complexity for computing Blackwell and dd-sensitive optimal strategies, by reduction to discounted games which can be solved in O((1α)1)O\left((1-\alpha)^{-1}\right) iterations. We provide the first bounds on the dd-sensitive threshold αd\alpha_{\sf d} beyond the case d=1d=-1, and we establish improved bounds for the Blackwell threshold αBw\alpha_{\sf Bw}. This is achieved by leveraging separation bounds on algebraic numbers, relying on Lagrange bounds and more advanced techniques based on Mahler measures and multiplicity theorems.

Keywords

Cite

@article{arxiv.2506.18545,
  title  = {Thresholds for sensitive optimality and Blackwell optimality in stochastic games},
  author = {Stéphane Gaubert and Julien Grand-Clément and Ricardo D. Katz},
  journal= {arXiv preprint arXiv:2506.18545},
  year   = {2025}
}
R2 v1 2026-07-01T03:29:16.816Z