Applications such as autonomous driving and video surveillance rely on Multiple Object Tracking (MOT) technology to accurately identify objects in video data. Real-time MOT systems are often challenged for continuously improving computational efficiency while maintaining the acceptable level of accuracy. For every frame, advanced algorithms for detection and tracking is used to identify and track objects. However, it is still unclear if they should be used across all frames, or whether it would be better to use the algorithm only for selected frames. This is an empirical question, best answered by experimental research. Here, we explore how frame skipping during object detection impacts tracking accuracy and speed in real-time MOT systems. We examined the trade-off between skipping and tracking robustness for given MOT tasks. The consequences of frame skipping were evaluated using publicly available MOT datasets (KITTI, MOT16, MOT17 and MOT20) in different skipping image frequencies. Frame skipping allowed us to achieve a negligible drop in the MOTA and HOTA score while giving us a big 80% boost in speed over regular baseline configuration.