Are these reports from real users?

These are sample analyses generated on de-identified footage — either internal recordings or beta-player footage with explicit permission to use as marketing examples. Personally identifying details have been removed. They're representative of what a real user sees.

Can I see a full sample report?

Yes — sign up for the free tier and process your first session. The free tier produces the full report (no watermark, no feature gates), so the fastest way to see what AceSense outputs is to run it on your own video.

AceSense Examples — Sample Coaching Reports and Analyses

Six sample analyses from de-identified amateur matches. What the report looks like, what insights came out, and where the AI helped.

The fastest way to evaluate AceSense is to run it on your own footage on the free tier. But if you want a sense of what the report looks like before you record anything, here are six sample analyses from de-identified matches and serve sessions.

Each example below describes the session, summarises what the report surfaced, and calls out the specific insight that made it useful. None of these are dressed up — they're the kind of "one specific thing to fix" outcomes that the product is supposed to produce.

Example 1 — NTRP 3.5 singles, hard court, club tournament

Session: 2-set match, hard court, outdoor, late-morning sun. Recorded with a Pixel 8 fence-clip-mounted at 8 ft behind the baseline. 1080p / 60fps, 47 minutes total.

Player profile: right-handed, two-handed backhand, NTRP 3.5, lost the match 6-3 6-4.

What the report showed:

187 shots detected, classified across forehand (38%), backhand (29%), serve (24%), volley (5%), slice (4%).
Forehand stroke-quality score: 71/100 (component breakdown: trunk rotation 78, contact-point variance 82, racquet acceleration 58, follow-through 66).
Backhand stroke-quality score: 79/100.
Serve: average first-serve speed 142 km/h, second-serve 96 km/h.
Heatmap: clear bias toward forehand corner of the court on rallies, with the deuce-side service box receiving 64% of returnable serves.

The insight that mattered: the racquet-acceleration component on forehand was the lowest in the report at 58/100. Specifically, the player was decelerating into the contact point on deep forehands rather than accelerating through. The "top three things to work on" auto-summary led with: "Forehand: maintain or accelerate racquet head through contact on deep balls; current pattern shows deceleration on shots beyond ¾ court depth."

That's a precise, actionable cue for the next practice session — and it's the kind of thing a coach watching from the side might miss across a 47-minute match because the deceleration pattern only showed up clearly on the deeper forehands.

Example 2 — NTRP 4.0 singles, red clay, friendly match

Session: 1-set match, red clay outdoor, mid-afternoon. Recorded with iPhone 14 Pro on a 7 ft tripod behind the baseline (no fence available). 1080p / 60fps, 38 minutes.

Player profile: left-handed, one-handed backhand, NTRP 4.0, dropped the match 6-4.

What the report showed:

142 shots, including 8 slices (high one-handed-backhand share by category).
Slice F1 detection on this match: ran cleanly through the regression check; 7 of 8 slices correctly classified, 1 confused with a flat backhand.
Stroke quality on slice: 84/100 — the player's strongest stroke by component scoring.
Ball tracking degraded slightly on three rallies where the clay had been heavily kicked up; bounce-localisation confidence dropped on those specific bounces (flagged in the report).

The insight that mattered: the report surfaced that the player won 78% of points where the rally exceeded 6 shots, but lost 64% of points where the rally was 3 shots or fewer. The shot-mix breakdown showed an over-reliance on aggressive first-strikes after the serve that landed in the net. The auto-summary recommended building one extra shot before going for the winner — a tactical insight, not a mechanical one.

This kind of pattern is hard to see while you're playing. It's also the type of thing a coach picks up after watching three or four matches; AceSense surfaced it on one.

Example 3 — Solo serve practice, 32 first serves

Session: Solo serve practice on hard court. Tripod at 6 ft, 8 ft behind the deuce-side baseline, slightly off-centre. iPhone 15 at 120fps. 14 minutes.

Player profile: right-handed, NTRP 3.5, working on first-serve consistency.

What the report showed:

32 first serves analysed.
Average speed: 168 km/h. Max: 184 km/h. Min: 151 km/h. Standard deviation: 7.2 km/h.
Toss-height variance: ~24 cm peak-to-peak — flagged as high.
Contact-point variance: roughly 35 cm peak-to-peak — also flagged as high.
Trunk-rotation arc: consistent (low variance).
Per-serve stroke quality ranged from 62 to 86, mean 74.

The insight that mattered: the report found a strong correlation between toss height and stroke-quality score for this session — high tosses (above the player's mean) consistently scored 4-8 points lower than median tosses. The auto-summary led with: "Lower-than-median toss height correlates with higher-quality contact for this session. Try toning the toss down by ~10 cm and tracking whether quality improves."

A specific, testable hypothesis from a 14-minute practice video.

Example 4 — Indoor doubles, mixed-level club night

Session: Doubles match, indoor LED-lit hard court, evening. Galaxy S23 fence-clip at 9 ft. 1080p / 60fps, 52 minutes.

Player profile: mixed pair, NTRP ~3.5 average. Two of four players right-handed; one player left-handed at the net.

What the report showed:

Aggregate shot detection: 268 shots detected at the match level.
Per-player attribution: reliable for both baseliners (right-handed and left-handed) at 92% accuracy. Net players' attribution dropped to ~78% on rapid net exchanges — flagged in the report's beta-feature warning banner.
Heatmap: showed the doubles team's collective coverage gap in the deuce-side mid-court — both players tended to drift toward the alleys on returns.
Stroke quality ran cleanly for the two baseliners; net-player quality scoring was attached to the partial attribution and flagged as preliminary.

The insight that mattered: the heatmap revealed a coverage gap that neither player on the team had noticed. The simple fix — one player covers the centre on returns, the other guards the alley — came directly from looking at the bounce-density map.

This example is also a deliberate showcase of where AceSense currently has doubles support in beta — we're transparent about what works and what doesn't.

Example 5 — NTRP 4.5 returner, hard court, training session

Session: Practice match against a stronger sparring partner. Hard court outdoor, partly cloudy. Pixel 9 Pro at 60fps, fence-clip 8 ft. 35 minutes.

Player profile: right-handed, two-handed backhand, NTRP 4.5, focused on return-of-serve.

What the report showed:

156 shots, of which 47 were returns of serve.
Return stroke quality: forehand return 83/100, backhand return 87/100.
Return depth heatmap: ~60% of returns landing inside the service line (i.e. short).
The auto-summary flagged: "Return depth significantly biased short. Strong stroke quality but low court penetration on returns; consider adjusting contact point further back on the toss-arrival axis."

The insight that mattered: stroke quality was good, but the player was costing themselves court position by playing returns short. This is a different category of insight from "fix your forehand" — it's a tactical gap that high-stroke-quality scores might otherwise hide.

Example 6 — NTRP 3.0 returner, public-park court, no fence

Session: Match on a public park court with no perimeter fence. Tripod at 6 ft, 8 ft behind the baseline, sandbag-weighted base. iPhone 13 at 30fps (older phone, no 60fps option). 41 minutes.

Player profile: right-handed, two-handed backhand, NTRP 3.0, returning to tennis after a long break.

What the report showed:

134 shots — slightly fewer than typical for the duration because of longer between-point pauses.
30fps input flagged at upload; the system warned that ball-tracking confidence on serves would be reduced.
Bounce localisation worked normally; serve-speed estimates flagged as ±15 km/h confidence interval rather than the usual ±6.
Stroke quality scored cleanly across the match (the pose pipeline doesn't depend on frame rate as heavily as ball tracking does).

The insight that mattered: the player's serve fault rate was 38% on first serves — high. The report's auto-summary cross-referenced this with toss-position variance and identified a left-side toss drift correlated with faults. Practical fix: tone down toss-arm reach across the body.

This example is also a demonstration that AceSense produces a useful report on a no-fence public court with a sub-recommended frame rate. We're explicit about the confidence-interval degradation rather than papering over it.

What's not in these examples

We deliberately chose representative-not-best examples. The reports above are real (de-identified). They are not "the best report we've ever seen." Three things you would also see if you used AceSense yourself:

Failure modes are flagged in-report. When the system has low confidence on a stretch of footage, it says so on the output. We don't pretend perfection; the accuracy methodology page documents every known failure mode.
Reports get more useful over time. Session 4 of a player's history is more useful than session 1, because the timeline view shows what's improving and what's regressing.
The "top three things to work on" auto-summary is opinionated. It picks three. Sometimes you'll disagree. Coaches will too — that's part of the value of the coach-share workflow, where the coach reads the auto-summary and tells you whether they agree or not.

Generate your own example

The fastest way to actually understand what AceSense produces is to run it on your own footage. The free tier gives you 3 analyses per month with the full report. Film one match by following the filming guide, upload, and you'll have a report in your inbox within 5-7 minutes that's identical in structure to the six above.

Read next: How AceSense works · Accuracy methodology · Filming guide · Pricing · FAQ

Frequently asked questions

Are these reports from real users?: These are sample analyses generated on de-identified footage — either internal recordings or beta-player footage with explicit permission to use as marketing examples. Personally identifying details have been removed. They're representative of what a real user sees.
Can I see a full sample report?: Yes — sign up for the free tier and process your first session. The free tier produces the full report (no watermark, no feature gates), so the fastest way to see what AceSense outputs is to run it on your own video.