This study examined the heterogeneity in temporal shifts of unrealistic optimism (UO) by analyzing students' grade expectation throughout a semester. UO was defined as the gap between students' estimated and current course grades, with a larger gap indicating higher UO. Final course grades were viewed as the outcome of UO. A total of 206 participants completed baseline measures of personal factors during the first week of the semester and repeated measurements at five subsequent time points. Using person-centered trajectory analysis (e.g., latent growth mixture models), we identified three distinct UO trajectories: UO-Persistent (6.8%; consistently high UO), UO-Decreasing (26.2%; diminishing UO), and Realistic (67.0%; consistently low UO). High perfectionistic standards and dysphoria predicted UO-Persistent group membership, while perfectionistic discrepancy, self-efficacy, and belief in optimism's power did not. The Realistic group achieved higher final grades than the UO-Decreasing group. Unexpectedly, no difference was found in final grades between the strongest UO group (i.e., UO-Persistent) and other two groups. These findings suggest that UO's temporal shift is not a unitary construct, and distinct UO patterns may be associated with different academic outcomes. This study underscores the significance of comprehending temporal shifts and employing person-centered analysis in UO related to academic achievement. The discussion addresses both research and practical implications.