The maximum maximum of a martingale constrained by an intermediate law (Q5936994)

Let \(\mu_1,\mu_2\) be zero mean probability measures. Consider the family \({\mathcal M}(\mu_1,\mu_2)\) of martingales \(M=(M_t)_{0\leq t\leq 2}\), with \(M_0\equiv 0\), and law \(\mu_t\) at time \(t=1,2\). Let \(\nu_2(M)\) denote the law of the terminal maximum \(\overline M_2=\sup_{0\leq t\leq 2}M_t\) and \({\mathcal P}(\mu_1, \mu_2)= \{\nu_2(M) \mid M\in{\mathcal M}(\mu_1, \mu_2)\}\). Using excursion theory, the authors construct an element \(M^*\in {\mathcal M}(\mu_1,\mu_2)\) whose maximal law \(\nu_2(M^*)\) dominates any \(\nu\in{\mathcal P}(\mu_1,\mu_2)\). This construction is an extension of the Azema-Yor-Skorokhod embedding. It extends results of \textit{D. G. Hobson} [in: Séminaire de probabilités XXXII. Lect. Notes Math. 1686, 250-263 (1998; Zbl 0935.60028)], where no intermediate law is imposed. Finally, an application is given to the robust hedging of a lookback option.