I'm attending a stocahstic processes course. I have some trouble with the intuition behind a stopping time. I will consider the discrete case to make it simpler.

a stopping time is given by Wikipedia as:

A stopping time with respect to a sequence of random variables $X_1,X_2,X_3,…$ is a random variable $\tau$ with the property that for each $n$, the occurrence or non-occurrence of the event $\{\tau \le n\}$ depends only on the values of $X_1,X_2,X_3,…,X_n$.

Being more general, $\tau$ is a stopping time if $\{\tau \le n\}$ is $\mathcal{F}_n$-measurable for any $n$, where $\mathcal{F}_n$ is the natural filtration of the process up to time $n$.

Wikipedia makes the example of the Gambler's ruin, starting from 100€ and winning or losing 1€ at each $n$. $X_n$= amount of money at time n.

Let's consider $\tau=\inf \{n \ge 0: X_n=0\}$.

Is $\tau$ a stopping time for any $n$?

I would think that for any $n<100$ $\{\tau \le n\}$ is equal to $\emptyset$ so it belongs to $ \mathcal{F}_n$ for any $n<100$, and for any $n>100$ $\{\tau \le n\}$ can be written as the intersection of {losing 1€ at time $m$} for any $m$ from $1$ to $100$, so it still belongs to $\mathcal{F}_n$. How do I concile this with the definition of WIkipedia say for $n<100$ (It depends on the future outcomes)? Is this the proper way to procede?

If so why $\tau=\inf\{ n\ge 0: X_n=2*X_0$ i.e. the $n$ at which I doubled my initial amount of money$\}$ is not a stopping time? Proceding as before I could construct an union of events belonging to $\mathcal{F}_n$ for any $n$. But ipse dixit it's not a stopping time.

I guess there's something big I'm missing here, I beg for mercy. Thank you

  • $\begingroup$ If $ X_0$ is fixed (not random) then I don't see the problem... $\endgroup$ – Ian Jan 31 '15 at 15:24
  • $\begingroup$ It is fixed, it's 100€. Still, I can't figure it out. Intuitively I would say that there's a chance I will never double the money because I run out of it before time n. Yet, given the definition the event of doubling the initial money when I've run out of money is impossible, so it's {} which belongs to any Fn.. $\endgroup$ – Tommaso Guerrini Jan 31 '15 at 15:33
  • $\begingroup$ That isn't a problem. If you ran out already then the doubling event doesn't happen for any n, which is fine. $\endgroup$ – Ian Jan 31 '15 at 15:36
  • $\begingroup$ You say it's a stopping time then? $\endgroup$ – Tommaso Guerrini Jan 31 '15 at 15:37
  • $\begingroup$ Looks like it to me. What exactly have you seen to the contrary? $\endgroup$ – Ian Jan 31 '15 at 15:47

Looks like the article on Wikipedia is requiring $\mathbb{P}(\tau < \infty) =1$. It is stated on the definition that sometimes it is required.

So, if you add this requirement, then of course your variable is not a stopping time, as you may also lose forever. If you don't (and I actually didn't add this requirement during my stochastic processes course) then it is a stopping time.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.