Why we built Tennis Tutor

A junior player gets an hour with a coach, sometimes two, and then goes away and practises. The practice is where the volume is. The coaching is where the correction is. Those two things happen on different days, and the space between them is where most of the improvement quietly fails to happen.

The scarce input is not court time. Most kids who are serious about tennis can find a wall, a ball machine, or a willing parent. The scarce input is someone watching closely enough to tell them what they actually did, as opposed to what they think they did. Those are different, and the gap between them is the whole problem. A player who believes they are meeting the ball out front, and is not, will practise that error several hundred times before anyone corrects it.

For kids inside a structured program, that correction arrives often enough. The moment they leave one, it stops almost entirely. That is the first thing we kept noticing: not that good coaching is rare, but that access to it is intermittent in a way the practice is not.

The second thing is the half of the sport that nobody in the tennis conversation seems to own. Strength and mobility work, when it happens at all, is rarely calibrated to where a particular kid is in their development. A twelve-year-old and a sixteen-year-old are not the same athlete scaled up and down, and a growing body has requirements that a general fitness plan is not built to respect. Most juniors are handed either nothing or an adult program with the weights reduced.

A camera and a phone should be able to fill some of that gap, and that is the claim the product rests on. Not to replace the coach. To cover the days the coach is not there.

So Tennis Tutor reads a recorded session frame by frame and reports on the things that are genuinely hard to see at full speed: joint angles, racket path, contact point. That last one matters more than it sounds. Where the ball is met relative to the body determines most of what happens afterwards, it changes by inches, and it is close to invisible in real time even to someone experienced who is standing in the right place. A camera does not get tired of looking at it.

The strength and mobility side is paired to the player's age and stage, drawn from pediatric sports-science guidelines rather than from what a professional does. This is the part we expect to be least exciting and most useful.

This is a category that routinely overstates what the machine does, so here is exactly where its job stops. The computer vision does the seeing. That is the entire scope of it. It identifies what is difficult to identify in real time and puts it in front of a person. The training plan that follows is rules-based and grounded in human expertise, because the question of what a fourteen-year-old should work on next is a coaching judgement and not a pattern-matching problem. The athlete is doing the work. Nothing here swings a racket.

There is a longer list of things it cannot do, and we would rather write it down than discover it in a review. It does not see intent. It cannot tell you that a player rushed a shot because they were nervous, or that a technically poor ball was the right choice under pressure. It has no opinion about tactics. It does not know what the coach said last Tuesday, or which flaw the player has been asked to leave alone for now while they fix something upstream of it. All of that is a coach's job and it stays a coach's job. This is a tool for between the lessons, not instead of them.

Tennis Tutor is not shipped yet. We are saying what it is for before it exists, which is uncomfortable to publish and useful to have on the record: if the thing we ship does not do what this note describes, the note is the evidence.