The escape rate of favorite sites of simple random walk and Brownian motion. (Q1879848)

For a simple symmetric random walk on the integers, denote by \(V(n)\) any site which maximises the number of visits by the walk up to time \(n\). Erdős and Révész initiated the study of the asymptotic properties of the `favourite sites' \(V(n)\). It is natural to conjecture, for example, that with positive probability, \(V(n)=0\) infinitely often. A surprising result of Bass and Griffin disproves this conjecture by showing that \(\lim | V(n)| =\infty\) almost surely. In the present paper the rate of escape is given by showing that \(\liminf_{n\to\infty} | V(n)| (\log n)^{\gamma}/n^{1/2}\) is zero or infinity according as \(\gamma\leq 1\), or otherwise. The proof uses a strong invariance principle of Révész to infer this from an analogous result for Brownian motion, which of course is of independent interest. To prove the Brownian motion case the authors replace fixed times by random times and use the Ray-Knight theorem to obtain delicate estimates for Brownian local times from an analysis of Bessel processes. The authors suggest that their method offers a valuable tool to attack further open problems on favourite sites.