In this study, we used Bayesian latent change score modeling (LCSM) to compare models of triannual (fall, winter, spring) change on elementary math computation and concepts/applications curriculum-based measures. We used data from elementary students in Grades 2–5 with approximately 700 to 850 students in each grade (47%–54% female; 78%–79% White, 10%–11% Black, 2%–4% Hispanic/Latino, 2%–4% Asian, 2–4% Native American or Pacific Islander; 13%–14% English learner; 10%–14% had special education individualized education plans). Our results converged with common nonlinear growth patterns from the assessment norms and prior independent findings. However, Bayesian LCMSs captured practically-relevant sources of change not observed in prior studies. Practical and methodological implications for screening and data-based decision-making in multi-tiered systems of support, limitations, and future directions are discussed.