The no-miracles argument is the main inference-to-the-best-explanation kind of argument for scientific realism, and the pessimistic induction is considered a main, if not the main, challenge for a NMA-based scientific realism. Doppelt advocates a new kind of inference-to-the-best-explanation supported scientific realism that he labels Best Theory Realism. If successful in replacing standard selective realism as the best version of scientific realism, BTR would be particularly good since it is not committed to the partial truth of past theories and thereby it (...) is immune to the antirealist strategy of finding cases of past, predictively successful theories with predictively essential components not retained by later theories. The goal of this paper is to raise doubts about Doppelt’s attempt and argue that, other benefits of his proposal notwithstanding, it fails. In section 1 I summarize the main tenets of standard, retentive selective realism relevant for the present discussion. In section 2 I show that Doppelt’s main arguments against retentive selective realism do not work. In section 3, I argue that the way BTR faces the challenge posed by the historical record that motivates PI is unsatisfactory and puts Doppelt into a fatal trilemma: either he is committed to two claims that are untenable together; or endorses an extremely implausible form of present-science chauvinism; or unjustifiably discriminates explanation against prediction in historical record. The conclusion is that BTR falls short of substituting standard retentive selective realism as the most plausible realist position, and that thereby the cases of past successful theories with predictively essential parts not retained by posterior theories are still a real problem for a plausible realist position. (shrink)