We can now answer a question we could not answer a year ago: does it actually matter how often you film your strokes?

It does. And the size of the effect surprised us.

Across every tennis player on OnCourtAI with enough history to track, players who built up ten or more analysed sessions improved their technique score 74% of the time. Players with fewer than ten improved 55% of the time. Same platform, same feedback, same strokes. The difference is that one group kept coming back.

What We Measured, and How

We took every completed tennis analysis on the platform between October 2025 and September 2026 — 2,308 analyses from 792 players, covering 15,249 individual shots. For every player with at least three analyses we compared the average score of their earliest sessions with the average of their most recent ones, and grouped players by how much they had used the platform.

Here is what came out:

  • 3 to 4 sessions: 55% of players improved
  • 5 to 9 sessions: 55% improved
  • 10 to 19 sessions: 71% improved
  • 20 or more sessions: 80% improved

The average gain follows the same shape. Players in the 20+ group improved by roughly four times as many points as players in the 3-4 group.

We Tried to Break the Result

A finding like that deserves suspicion, because there is an obvious way it could be an illusion. If scores drift upwards for any reason at all — noise, seasonal form, the weather — then players with more sessions would show more drift, and it would look like improvement without anything real happening.

So we tested it. We took every player's set of scores, randomly shuffled the order, and re-ran the entire analysis. If the effect were an artefact of having more data points, shuffled data would show it too.

It did not. On shuffled data, the heavy-use group improved 50% of the time — exactly what you would expect from coin flips — with an average change of roughly zero. We ran this 300 times. Only 2 of those 300 random shuffles produced a result as strong as the real one.

In plain terms: the odds of getting this by chance are under one in a hundred. The ordering of the sessions matters. Players are genuinely getting better, not accumulating lucky readings.

Spacing Beats Cramming

Volume is not the whole story, and this is the part that changes how you should actually use the app.

We also grouped players by how many separate weeks they had filmed in, rather than how many videos they had shot in total:

  • All sessions in one week: 57% improved, average gain +0.7
  • Spread across two weeks: 52% improved
  • Spread across three or more weeks: 63% improved, average gain +1.6

Ten videos filmed in a single afternoon is not ten sessions of learning. It is one session, recorded ten times. You cannot improve your contact point between the third shot and the fourth shot of the same basket — the change has to be practised, and practice takes days.

Looking at elapsed time tells the same story. Players whose filming spanned one to three months improved 61% of the time with an average gain of +2.1 points. Players who did everything inside a single week improved 55% of the time, gaining +0.8.

The combination that works is regular sessions with real gaps between them. Not a burst of enthusiasm in March and nothing until August, and not fifteen videos on one Saturday.

Why the Gaps Are Doing the Work

This is not a quirk of our platform. It is how motor learning has always worked.

When you watch your forehand analysis and see that your contact point is behind your front hip, you have not fixed anything. You have identified something. The fix happens over the following days and weeks, on court, hitting balls with that one thought in mind, until the new pattern stops requiring conscious attention.

Filming again immediately measures the same unchanged stroke. Filming again in three weeks measures whether the work took.

There is a second, less obvious reason the gaps help. A single video is a small sample of a variable thing. Your forehand on a windy Tuesday when you are tired is not your forehand. Multiple sessions spread over time average out the noise, so that what you are looking at is your actual technique rather than one afternoon's version of it.

Which Strokes Respond Fastest

Not every stroke rewards consistency equally. Looking at each stroke separately across players with at least three sessions on that stroke:

  • Forehand: 61% of players improved, average gain +1.6 — the most responsive stroke
  • Backhand: 58% improved, average gain +0.6
  • Serve: 55% improved, average gain +0.2 — the most stubborn

The forehand is a shorter chain of movements, so improving your contact point tends to improve the whole stroke at once. The serve has more dependent parts, and fixing one can disturb another — which is exactly why the serve needs the longest runway and the most patience.

If you are starting out and want to see the system work, film your forehand. If you want the biggest long-term prize, film your serve and accept that it will take months rather than weeks.

What Good Practice Actually Looks Like

Based on what the data shows, here is the pattern we would recommend:

Film every two to four weeks, not every session. Often enough to catch drift, spaced enough for changes to bed in.

Take one fault away at a time. Look at your component scores, pick the lowest one you can realistically work on, and ignore the rest until the next video. Players who try to fix four things fix none.

Film the same stroke from the same angle. Comparisons are only meaningful if the setup is consistent. Our guide on filming for AI analysis covers this.

Keep going past the third session. This is the single clearest message in the data. The jump from 55% to 74% happens between the ninth session and the tenth. Most players who stop, stop before they get there.

Watch components, not the headline score. Your overall number can stay flat while two components move in opposite directions. The detail is where the instruction is.

The Honest Caveat

There is one thing this data cannot separate, and we would rather say it than let you find it later.

Players who film twenty sessions are, on average, more committed players. They are more likely to be having lessons, playing more matches and practising more deliberately. We cannot fully untangle "filming consistently causes improvement" from "people who improve are the sort of people who film consistently."

What the shuffle test does establish is that the improvement is real and ordered in time — these players genuinely got better, in sequence, rather than accumulating random high readings. Whether the camera caused it or merely recorded it, consistent measurement is a reliable marker of players who are actually getting better.

Our honest reading is that it is both. The feedback helps, and the habit of measuring keeps you honest about whether the work is landing.

Start the Clock

If you have filmed once and stopped, you are in the largest group on the platform and the group that gains least. The fix is not more effort in one afternoon. It is one video, once a month, for six months.

On our numbers, that puts you in the group where three players in four get measurably better.