Chen et al. present the fundamentals of a method to determine trajectories of athletes for the analysis of tactics. This could be an incredibly valuable tool for the sports and broadcast industries. The technique is completely independent of human intervention, which has both positive and negative implications. The approach is a layered technique that progresses from low-level image processing to high-level trajectory segmentation and classification. The technique has been tested on basketball players, but the authors make it clear that other researchers have attempted to use the same method on soccer, and that it should be applicable to most team sports.

An overview of the primary ideas is presented in a very short but properly concise section 2. However, the real details start in section 3, where the authors discuss the fundamental process of determining camera calibration based on the court lines. This step is imperative for successful calibration and extraction, because the video is not consistent from game to game or even from scene to scene. Section 4 addresses the task of detecting, extracting, and tracking the players as individuals and as members of a team. The final step, described in section 5, involves the abstraction of the screen tactic from the video by detection and classification of the motion patterns.

The analysis provided is sufficiently detailed at each level of the abstraction process. There are real performance results for court line extraction and calibration, player/team detection and tracking, and final tactic classification.

The paper as a whole is well formed and articulate with regard to the research performed, the results presented, and the analysis completed. However, there are still multiple spelling, formatting, and grammatical errors that should have been corrected prior to publication. The figures are informative, but a few are not clearly valuable to the paper. The fundamental tactic that was detected was a screen, which, while not unique to basketball, is not a widely used technique in most sports. I still question the broader impact of this research on automatic sport tactic analysis.

Overall, I believe this this paper has merit in advancing the ability to automatically detect, identify, and classify specific examples of common tactics in basketball. As a result of reading this paper, I am convinced that the major sports organizations should further investigate this technique as a tool to improve performance without having to increase cost.