<p>Multi-Object Tracking (MOT) in dynamic environments with visually similar objects, such as sports and dance, remains a challenging problem that has yet to be fully solved. While MOT has seen significant advancements through techniques such as Intersection over Union (IoU) matching, re-identification (Re-ID), and Kalman filtering, these methods perform well in general settings but struggle in scenarios with similar appearances, frequent occlusions, and fail to re-associate objects that temporarily exit and re-enter the field of view, resulting in tracking discontinuities. In response to these limitations, we propose SportSORT, a novel approach designed for sports applications but adaptable to other domains facing similar challenges. Our method introduces three key innovations: Domain-specific Feature Matching, which leverages distinctive attributes such as jersey colors and numbers to reduce IoU-related misidentifications and improve tracking accuracy; Corrective Matching Stage, which resolves identity mismatches caused by long-term occlusions and Out-of-View Re-Association mechanism that enhances object re-identification upon re-entry into the scene. Experimental evaluations on the SportsMOT and SoccerNet-Tracking datasets indicate that SportSORT achieves HOTA scores of 81.3% and 88.0%, respectively, surpassing current state-of-the-art methods and performs robustly in complex sports tracking scenarios.</p>