Optimal Stopping Values and Conditional Expectations

Summary

The document raises questions about interpreting an optimal stopping value function, the conditioning notation, and why a strategy that waits one time step is valued by an expectation over future states. It also asks whether the inequality between the current value and expected future value follows from a supermartingale property. These are conceptual questions about stochastic processes and dynamic programming rather than a worked derivation or trading strategy.

No explanatory answer or supporting evidence is included, so the document does not resolve the questions. Its usefulness lies in identifying distinctions a reader must clarify: the value is evaluated given current information and state, while a future value is random before the next state is observed and must be averaged. Any conclusion about supermartingales depends on how the value process and admissible stopping rules are defined.

Key ideas

• The optimal stopping value is expressed as the best expected payoff over eligible stopping times.
• Conditional expectation notation needs interpretation in terms of the information and state available at the evaluation time.
• A future value is generally random before the next state is observed, so its current value is represented by an expectation.
• The document asks whether a value inequality follows from a supermartingale property but provides no answer.

Tags

# A fundamental question on optimal stopping time need clarification


# A fundamental question on optimal stopping time need clarification












I am currently studying optimal stopping time.Under this topic there is a basic concept which confuses me. I would appreciate some clarification. So we define $\tau$ a stopping time, and $\phi (\tau,x_t)$ being the pay off function that realises the optimal value function $$V(t,x)=\, sup_{t\leqslant \tau \leq T}\, E_{t,x}[\phi (\tau,x_t) ]$$ Firstly I don't understand both the mathematical and finance implication of this? $$E_{t,x}[\cdot ]\rightarrow x_{t}=x$$ Secondly,I don't understand that if I have a strategy that waits until $t+\Delta t$ then uses$\tau_{t+\Delta t}$ , why is the value of this strategy $$E_{t,x}[V(x_{t+\Delta t},\tau_{t+\Delta t})]$$instead of $$V(x_{t+\Delta t},\tau_{t+\Delta t})?$$ This brings the question of at what time point $t$ are we evaluating all these stopping time strategy? Lastly, I would like to know if this $$V(x_{t},\tau_{t})\ge E_{t,x}[V(x_{t+\Delta t},\tau_{t+\Delta t})]$$is purely based on the assumption of supermartingale property?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.