How many probes are needed to compute the maximum of a random walk?

Odlyzko (1995) proves that, in the average, cn+o(n) probes are necessary to compute the maximum of a simple symmetric random walk with n steps, in which c appears under the form of a triple integral. In this paper we prove that (log n/|log p−log q|)+o(log n) probes are necessary to compute the maximum of a simple asymmetric random walk with n steps. We also give c under closed form.