We study the selection of covariate adjustment sets for estimating the value\nof point exposure dynamic policies, also known as dynamic treatment regimes,\nassuming a non-parametric causal graphical model with hidden variables, in\nwhich at least one adjustment set is fully observable. We show that recently\ndeveloped criteria, for graphs without hidden variables, to compare the\nasymptotic variance of non-parametric estimators of static policy values that\ncontrol for certain adjustment sets, are also valid under dynamic policies and\ngraphs with hidden variables. We show that there exist adjustment sets that are\noptimal minimal (minimum), in the sense of yielding estimators with the\nsmallest variance among those that control for adjustment sets that are minimal\n(of minimum cardinality). Moreover, we show that if either no variables are\nhidden or if all the observable variables are ancestors of either treatment,\noutcome, or the variables that are used to decide treatment, a globally optimal\nadjustment set exists. We provide polynomial time algorithms to compute the\nglobally optimal (when it exists), optimal minimal, and optimal minimum\nadjustment sets. Our results are based on the construction of an undirected\ngraph in which vertex cuts between the treatment and outcome variables\ncorrespond to adjustment sets. In this undirected graph, a partial order\nbetween minimal vertex cuts can be defined that makes the set of minimal cuts a\nlattice. This partial order corresponds directly to the ordering of the\nasymptotic variances of the corresponding non-parametrically adjusted\nestimators.\n