This paper focuses on the issue of mentor-mentee match for mentoring beginning teachers in elementary schools. While grade-level match is often an explicit goal of mentor assignment in many existing teacher mentoring programs, empirical evidence is limited regarding its effects on teachers' instructional practices. Data come from a large-scale study of comprehensive teacher induction programs. Taking advantage of the study design in which only one mentor was assigned to multiple teachers within the same school, I analyze an adaptive centering with random effects model. The results suggest that exposure to mentors with expertise in the same grade levels may benefit beginning teachers' classroom management and lesson implementation, but not lesson content.