Convergence of classes of subpramarts and games which become better with time (Q805062)

Let (\(\Omega\),F,P) be a probability space, D a directed set filtering to the right with the order \(\leq\), and \((F_ t,D)\) an increasing family of sub-\(\sigma\)-algebras of F. A stochastic process \((X_ t,D)\) is a subpramart, if for each \(\epsilon >0\) there exists a simple stopping time \(\sigma_ 0\) s.t. \[ (1)\quad P(X_{\sigma}-E(X_{\tau}| F_{\sigma})>\epsilon)<\epsilon \] for all simple stopping times \(\tau \geq \sigma \geq \sigma_ 0\). A process \((X_ t,D)\) is a game which becomes better with time (GBT), if (1) holds for \(s_ 0,s,t\in D\), \(s\leq t\). A process X is a pramart, if both X and -X are subpramarts, and X is a game which becomes fairer with time (GFT), if both X and -X are GBTs. These definitions generalize the notions of a submartingale and a martingale. The author proves various (stochastic and essential) convergence results and optimal stopping theorems for subpramarts and GBTs.