Computer-aided sports training is very advantageous for aspiring sports persons. In the games like cricket, maintaining an optimal pose is of prime importance as the performance of the batsmen is completely dependent on the pose. The pose can be achieved by long hours of practice. Towards this end, this paper aims to recognize the pose of the batsman, thereby recognizing the shot that is offered by the batsman. Straight drive, pull and cut shots are to identified from the video frames, and this is done by processing the joint locations that are extracted from a multi-person pose estimator, namely the AlphaPose network. The ‘merit’ of the joint information that is associated with a novel entropy measure formulated for this work. Extensive experimental studies indicate that the proposed entropy-based joint selection nearly outperforms the other measures that are considered for the study. A training tool based on the entropy-assisted feature selection is proposed in the latter part of the paper. This tool assesses the video file of the shot taken at a preset distance and provides feedback about the pose correction that is required to execute the shot properly.