Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

If $X$ is a standard 1d brownian motion and $M_t$ $= \mbox{max}\{X_S: 0 \le s \le t\} $, what can we say about $M_t/t$ as $t \rightarrow \infty$?

Mainly, what can we say about the behavior of this martingale?

My attempt: P(Msubt > a) = 2P(B(t) >a), but what integral does this fit in with?

share|cite|improve this question
Hint: The reflection principle gives that $M_t$ has the same distribution as $|B_t|$. – Chris Janjigian May 2 '12 at 23:46
Sorry I used $B_t$ as my notation for a Brownian motion there. – Chris Janjigian May 2 '12 at 23:52
It also might help you at some point when estimating the tail of the normal distribution that there is a really stupid bound: $e^{-x^2}\leq xe^{-x^2}$ for $x$ large. – Alex R. May 3 '12 at 0:27
So you are saying to use Hoeffdings inequality? will it help? – mary May 3 '12 at 1:59
Perhaps I was being too cryptic. Chris suggested that the distribution of $M_t$ is the same as $|B_t|$. You can easily calculate the probability that $|B_t|\geq at$ for $a$ a positive constant, as a Gaussian integral but you'll need to do some estimates for how this "tail" probability behaves for large $t$. I believe my prior hint will help. – Alex R. May 3 '12 at 2:53

Since $B_t$ is symmetric, the result recalled in the post reads $\mathrm P(M_t\gt at)=\mathrm P(|B_t|\gt at)$ $(\ast)$.

Since $B_t$ equals $\sqrt{t}B_1$ in distribution, $(\ast)$ is $\mathrm P(|B_1|\gt a\sqrt{t})$. Since $|B_1|$ is almost surely finite and $a\sqrt{t}\to+\infty$ for every $a\gt0$, the last probability goes to zero, hence $\mathrm P(M_t/t\gt a)\to0$ for every $a\gt0$. That is, $M_t/t\to0$ in probability.

One can strengthen this convergence, noting that the series $\sum\limits_n\mathrm P(|B_1|\gt a\sqrt{n})$ converges since $\mathrm P(|B_1|\gt a\sqrt{n})$ is $\mathrm e^{-\frac12a^2n+o(n)}$. By the first Borel-Cantelli lemma, $M_n/n\to0$ almost surely when the integer $n$ goes to infinity. For every $t\geqslant1$ there exists an integer $n\geqslant1$ such that $n\leqslant t\lt n+1$, and $(M_t)_t$ is nondecreasing hence $M_t/t\leqslant M_{n+1}/n\leqslant 2M_{n+1}/(n+1)$. This proves that $M_t/t\to0$ almost surely when the real number $t$ goes to infinity.

The same identity $(\ast)$ shows the convergence in $L^p$ for every finite $p\geqslant1$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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