488e4104520c6aab692863cc1dba45af-AuthorFeedback.pdf

The constraints in the theorem characterize the dual and primal36 quantities (dπ and Qπ), which can be used to estimate policy value, either alone or combined (lines 171-173, with37 a change-of-variableζ = dπ/dD). It is thus a natural starting point for OPE, which we will make explicit in the38 final version.