In existing Multi-Object Tracking(MOT) algorithms, most trackers primarily rely on motion information and appearance information to accomplish tasks. Although these cues possess strong discriminative capabilities, their effects can be weakened in crowded scenes due to non-linear target motion and occlusions. In this paper, we discover that the judicious utilization of pseudo-depth and height change direction can markedly enhance the tracker’s association capabilities. To this end, we propose two additional cues to enhance the tracker’s performance: pseudo-depth weighted hierarchically according to confidence(PD-WHC), and Intersection over Union modulated by both height and height change direction( \(\mathrm {H^{2}DMIOU}\) ). With the assistance of these two cues, our method AMC-SORT has achieved outstanding performance on the MOT17, MOT20, and DanceTrack datasets. On the most challenging DanceTrack dataset, AMC-SORT has surpassed the existing state-of-the-art algorithms, demonstrating its robust association capabilities.