The effectiveness of Tier 2 interventions may depend on the variability of student skills within an intervention group. We investigated the effect of initial variability within intervention groups using data from a large-scale study of ROOTS, a Tier 2 kindergarten mathematics intervention. Our research questions were as follows: Does mathematics pretest variability within ROOTS groups predict student gains from the intervention, and is this relationship moderated by (a) intervention group size (2- versus 5-student groups), (b) quality of explicit instruction, or (c) group-level initial skill? We found that ROOTS groups with greater pretest variability on a closely-aligned mathematics measure experienced smaller gains from the intervention. This association was more impactful in larger intervention groups (i.e., 5-student groups compared to 2-student groups), and in groups with higher quality of explicit instruction. We unpack these findings and conclude with recommendations for practice and future research. [This paper was published in "Exceptional Children" v90 n3 p331-346 2024.]