Marker-based motion capture, a widely used method for three-dimensional motion analysis, entails important shortcomings, including soft tissue artifacts and constraints on experiment environments. By contrast, markerless systems require no reflective markers, show minimal inter-session variation, and remain unaffected by clothing, making them promising tools for athletic performance evaluation. This study was conducted to compare kinematic data obtained using the respective systems during change of direction (COD) maneuvers and to evaluate the applicability of markerless systems. Five trials of 90° COD maneuvers were performed by 23 male participants. Kinematic data were captured simultaneously using marker-based (Motion Analysis) and markerless systems (Theia Markerless Inc.). The markerless system used synchronized multi-camera deep learning to detect anatomical landmarks and to reconstruct full-body skeletal motion through triangulation and inverse kinematics. Trunk and lower-limb joint angles were calculated for both systems. Bland–Altman analysis, the intraclass correlation coefficient (ICC), root mean square deviation (RMSD), and normalized root mean square error (NRMSE) were used to compare the two systems. Both systems demonstrated good agreement for most joint angles. However, notable mean differences were found in ankle dorsiflexion (−10.92° [−18.38, −3.46]), knee flexion (−8.32° [−14.48, −2.13]), and hip external rotation (12.1° [−2.12, 26.33]). Most angles also showed good ICC values (>0.75), indicating measurement reliability between the systems. These findings suggest that markerless systems can capture kinematic patterns reliably during COD maneuvers. However, comparing the magnitudes of joint angles with those of marker-based systems demands caution. This method is valid for COD analysis if system-specific differences are considered.