From Scores to Sessions: Building Training Plans From Your Own Data

The hardest part of improving at tennis is not finding out what is wrong. It is working out what to do on Tuesday evening.

You film your forehand, you get your scores back, and you learn that your spin generation is 38 and your kinetic chain is 48. Useful. Now what? Which of those do you fix first? What drill actually targets a kinetic chain? How many weeks should it take?

That gap — between a diagnosis and a plan — is what My Plan was built to close.

What My Plan Does

My Plan reads the analyses already sitting in your account, finds the components where you are genuinely weakest, and builds a structured multi-week training programme around them. It is generated from your own numbers, not from a generic template with your name at the top.

You tell it three things: how many weeks you want, how many sessions a week you can realistically do, and how long each session can be. It handles the rest — session structure, warm-up blocks, main blocks, specific drills, and what each drill is actually targeting.

Across the plans built so far, the typical plan runs five weeks, three sessions a week, about 64 minutes a session. Those are the numbers real players chose when asked what they could commit to, and they are a reasonable starting point if you are unsure.

How It Decides What You Should Work On

This is the part that matters, so it is worth being specific.

Every analysis you run scores your stroke across its individual components — preparation, contact point, kinetic chain, balance, wrist lag, follow-through, spin generation, footwork, swing mechanics, and more depending on the stroke. My Plan pools those component scores across all your recent videos, averages them, and ranks them from weakest to strongest.

Averaging across multiple videos is deliberate. A single bad reading on a windy afternoon should not set your training priority for the next five weeks. The plans in our system are typically built from five to thirteen analyses per component, which is enough to tell a genuine weakness from a bad day.

The weakest components become the plan's focus. Everything else is left alone, on purpose — a plan that tries to fix nine things fixes none.

What the Data Says Players Are Actually Weak At

Pooling the component scores across every plan generated so far produces a fairly brutal picture of the amateur game. Here are the averages, worst first:

• Spin Generation: 35
• Kinetic Chain: 48
• Hip Rotation: 50
• Knee Bend: 51
• Weight Transfer: 54
• Wrist Lag: 54
• Balance: 55
• Contact Point: 61
• Follow-through: 61
• Footwork: 67
• Preparation: 70
• Ball Toss: 80

Two things jump out.

First, spin generation is the weakest component in the amateur game by a distance, and it is not close. It was the single highest-priority weakness in 13 of the 17 plans we could analyse. If you have never been told your topspin is the thing holding your forehand back, the numbers suggest it probably is.

Second, the components clustering at the bottom — spin generation, kinetic chain, hip rotation, knee bend, weight transfer — are all describing the same underlying fault from different angles. They are all about using the ground and the body to generate racket speed, rather than swinging with the arm.

That is the defining technical difference between a club player and a good one, and it is exactly what a well-built plan should attack.

Inside a Session

A generated session is not a list of vague suggestions. Each one has a type, a structure and timed blocks.

Sessions come in three types, and the mix is deliberate: roughly 69% technique, 16% match play, 15% fitness across everything generated so far. Technique dominates because that is what the analysis measures, but the plan does not pretend tennis is only technique — match play sessions exist to put the new pattern under pressure, and fitness sessions exist because a knee bend score of 51 is sometimes a strength problem rather than a technique problem.

Within a session you get a warm-up block and a main block, each with named drills, durations and — crucially — the component each drill is targeting. A real example from a forehand plan:

• Live Ball Split Step Rally — 20 minutes
• Wall Drill for Forehand Contact Point — 15 minutes
• Weight Transfer and Balance Drill — 15 minutes
• Targeted Forehand Practice — 20 minutes, targeting kinetic chain

The system has generated 628 sessions containing 2,925 individual drills to date. Every one of those drills is tied back to a component score that came from a player's own video.

Ticking Things Off

Plans are only useful if you actually follow them, so the plan is interactive. You mark drills complete as you do them, and sessions complete as you finish them. The plan tracks where you are in the programme rather than becoming a document you generated once and never opened again.

Plans have a status too — draft, active, completed or archived — so you can generate one, look at it, and decide whether to commit before it starts cluttering your account.

How This Connects to the Rest of the Platform

My Plan sits at the end of a loop rather than standing on its own:

Film a handful of shots. Analyse them to get component scores. Generate a plan from your weakest components. Work the plan for four to six weeks. Film again and find out whether it moved.

That last step is the one people skip, and it is the one that makes the whole thing work. Our own data on consistency is unambiguous here: players who build up ten or more analysed sessions improve 74% of the time, against 55% for players who film only a handful. A plan gives the gap between videos a purpose.

It also means the second plan you generate is better than the first, because it is built on more data about you.

Padel Too

Plans are not tennis-only. Padel players get the same treatment, built around padel's own component set and stroke types, with the plan length and session structure adjusted to match. Padel plans in our data run slightly longer per session — around 70 minutes — and fewer sessions per week, which reflects how most padel players actually get court time.

Where It Falls Short

Two honest limitations.

The plan is built from what the camera can see. It knows your kinetic chain is poor; it does not know you have a shoulder that has been grumbling since April, or that your club only has court time on Sundays. Treat the output as a strong starting proposal, not an instruction.

And a plan cannot replace a coach for the hardest faults. Spin generation is the clearest case — it is the weakest component we measure and the one that responds least well to self-directed work, because it needs a swing path rebuilt rather than a detail corrected. A plan can point you at it, structure the repetitions and tell you when it moves. Someone standing on the other side of the net feeding balls will get you there faster.

Getting Started

You need at least a few analyses in your account first — the plan is only as good as the data it reads, and one video is not enough to separate a weakness from a bad afternoon. Film two or three sessions across a couple of weeks, then open My Plan in the OnCourtAI app, set your weeks and availability, and see what it proposes.

Then, six weeks later, film the same stroke again. That is the only part of this that actually proves anything.