We revisit the "fareytail expansions" of elliptic genera which have been used in discussions of the AdS 3 /CFT 2 correspondence and the OSV conjecture. We show how to write such expansions without the use of the problematic "fareytail transform." In particular, we show how to write a general vector-valued modular form of nonpositive weight as a convergent sum over cosets of SL(2, Z). This sum suggests a new regularization of the gravity path integral in AdS 3 , resolves the puzzles associated with the "fareytail transform," and leads to several new insights. We discuss constraints on the polar coefficients of negative weight modular forms arising from modular invariance, showing how these are related to Fourier coefficients of positive weight cusp forms. In addition, we discuss the appearance of holomorphic anomalies in the context of the fareytail.