2026/06/30 by Magnus Perninge
Mathematics · #math.OC
arxiv created 2026/07/30 · arxiv updated 2026/07/31
We consider a finite-horizon, zero-sum stochastic differential game in which one player controls a path-dependent stochastic system, while the opponent is given the opportunity to terminate the game prematurely. We introduce a control randomization formulation, which allows us to establish that the upper and lower value functions coincide. Our approach also yields a representation of the common game value in terms of a nonlinear Snell envelope, where the underlying stopped process is given by the unique maximal solution of a backward stochastic differential equation (BSDE) with constrained jumps.