AI Watches Your Cricket Swing and Coaches You Like a Pro

A team of three researchers has built an AI system that watches cricket batting on video and instantly identifies which shot the player just played — a flick, pull, defense, or drive — with 97.68% accuracy in their tests. More importantly, it explains how the body moved, not just what happened. The work is described in a short six-page academic paper.

The trick is that the system mostly ignores the picture and focuses on the skeleton. Software called YOLO (a tool that spots people in video) first finds the batter in each frame, then a second tool maps 30 key points on the body — shoulders, elbows, knees — into three-dimensional space. That means awkward camera angles, poor lighting, and busy backgrounds no longer matter. Older systems relied on raw video pixels, which broke easily when conditions changed.

Why you should care: good cricket coaching is expensive and not available to everyone. This kind of tool could one day run on a phone propped up at the nets, comparing your joint angles to an expert's and flagging habits that lead to injury — especially useful for young players whose bodies are still developing. The same approach is already showing up in golf, tennis, and physiotherapy, so expect it to reach other sports and home fitness apps.

The catch: the system only recognizes four shots, and it hasn't been tested on large amounts of messy real-world footage. It also needs a reasonably clear view of the player, so a phone buried in a kit bag won't work. It is research, not a product — nobody can buy it today. And filming players, especially children, raises real questions about who owns and stores that video.