Systems and methods are described for selecting tones for far-end crosstalk (FEXT) mitigation. In particular, systems and methods are described for performing far-end crosstalk (FEXT) mitigation within a digital subscriber line (xDSL) system. One embodiment is a method that comprises determining a first set of function values relating to a per-tone signal-to-noise ratio (SNR) for a user among a plurality of users for a number of tones, wherein the number of tones is less than or equal to a total number of tones available for transmission. The method further comprises estimating coefficients for performing FEXT mitigation for the number of tones, performing FEXT mitigation on the number of tones based on the estimated coefficients, and performing FEXT mitigation on a selected group of tones from among the number of tones, wherein the selected group of tones is less than or equal to the number tones.