Your Pickleball · April 23

01 · Discovered

You played four kinds of shots.

We didn't tell the model what a dink or a drive looks like. We ran unsupervised clustering on 237 of your wrist waveforms and four acceleration fingerprints emerged. The labels below are our best guesses — tell us which ones match how you actually played them.

Four clustered shot waveforms auto-discovered from 237 of your shots

Each panel shows all 30+ shots in one cluster (faint lines) and the cluster centroid (bold). X-axis = 800ms window around peak contact. Y-axis = normalized acceleration.

Cluster 0

Hard drives (our guess)

75 shots

Sharp symmetric peak at 8.2g mean — the highest of any cluster. Rise ~85ms, fall ~115ms. These are your power moments — body engaged, clean force.

Signature: ∧ · peak 8.2g · dur 200ms

Cluster 1

Quick dinks & volleys (our guess)

61 shots

Tight narrow peak — fastest in/out of any cluster. Rise 44ms, fall 58ms. The kitchen touch shots — compact, no wind-up, no follow-through.

Signature: ▲ · peak 5.5g · dur 101ms

Cluster 2

Loaded shots (serves? backhand setups?)

48 shots

Slow 155ms rise, sharp 71ms fall — the wind-up signature. Your arm loads before release. Most likely serves or high backhand setups.

Signature: ╱│ · peak 5.5g · dur 226ms

Cluster 3

Trailing-finish shots (backhands? transitions?)

53 shots

Asymmetric — rise 93ms, fall 122ms. The lingering-finish pattern. Possibly backhand drives, or shots played from the transition zone.

Signature: │╲ · peak 5.7g · dur 214ms

Why this matters to you

No other consumer pickleball wearable does this. Once you label a handful of shots ("this was a serve, this was a dink"), the app learns your personal shot alphabet and tracks how each evolves session by session. Next week you'd get coaching like "your serves got 12% faster — what did you change?" instead of just "you hit 98 shots."

02 · Form signatures

What your wrist hints at about your form.

The wrist sensor can't see your paddle face or grade technique the way a coach watching from the sideline can. But the SHAPE of each swing signature tells us something about how you're generating force. Here's what we can honestly read, and what still needs a video.

Your loaded shots (Cluster 2, ~48 shots)

You're actually winding up — that's good.

Your Cluster 2 signature has a 155ms rise followed by a 71ms fall. That's a 2.2:1 load-to-release ratio — meaningful wind-up followed by committed finish.

Most rec players have near-symmetric signatures on serves (just swing at it). Yours shows you're building momentum through the arm before release. If these are your serves, that's real serve form, not just a tap.

Needs video to confirm: the SOURCE of the load (hip rotation vs shoulder-only) — but the load itself is there.

Your hard drives (Cluster 0, ~75 shots)

Your drive finish is committed.

Rise 85ms, fall 115ms. Your fall time is ~35% longer than your rise. That's a follow-through signature.

If you were arming the ball — swinging from the shoulder only and stopping on contact — we'd see a fast rise and an equally fast stop. Your longer fall means your paddle keeps moving past contact. That's body-driven force, not arm-only.

Needs video to confirm: whether the follow-through is directional (forward = good) or wristy (across = less good).

Your dinks & volleys (Cluster 1, ~61 shots)

Your hands are quiet on touch shots.

44ms rise, 58ms fall. No wind-up, no follow-through. Pure reactive paddle use. That's the signature of hands that are already in front of you, not swinging through the ball.

This is GOOD. Rec players often "swing" their dinks, which shows as a wider signature with a bigger decay. Yours are narrow = committed kitchen stance, paddle out in front.

Needs video to confirm: paddle face angle (closed/open) on your dinks — the sensor can't see this.

Your trailing-finish shots (Cluster 3, ~53 shots)

Something is different about this group.

Rise 93ms, fall 122ms. The asymmetry is stronger than your hard drives. Long decay after contact — paddle keeps moving well past the ball.

Plausible reads: (1) these are backhand drives with extended finish, (2) you're reaching for balls out of your strike zone, (3) these are shots played on the move rather than set. In pickleball, the tell of a reaching shot is usually the trailing decay — paddle continues because your body wasn't positioned to stop it cleanly.

If you can tell us "these were my backhands" OR "I was stretched out a lot" — we can train the detector to confirm.

Your overall form signal

Across 237 shots the top-5 nearest-neighbor similarity averaged 0.97 — which means the motor programs you're running are extremely consistent. You're not flailing differently every swing. Your body knows four patterns and executes them repeatably. That's the baseline condition for any real coaching to work.

What the wrist can never tell you

• Paddle face angle at contact (open/closed)

• Contact point on the paddle

• Spin magnitude in RPM

• Where on the court you stood

• Your split-step timing

• Your grip (Continental, Eastern, etc.)

For those we need either a paddle sensor or a phone video paired to the IMU. Both on the roadmap.

03 · Your rallies, per YOU

Important context: you see only your shots.

In a doubles rally, 4 players are on court but only your wrist has a sensor. The PPA pro tournament stat of 11.5 shots per rally counts shots from all 4 players. Your 2.9-shots-per-rally average is just the shots YOU struck — your partner + your opponents struck the rest. If we assume roughly balanced doubles participation, a ~3 shots-per-rally from you = a total rally of ~8-12 shots across all 4 players. That's actually a healthy rally length.

Session 1 — how often you struck in each rally (12s gap threshold, beta)

1 (your serve/return)

2–3 strikes

4–5 strikes

6–8 strikes

9+ strikes ★

You struck 2.9 shots per multi-shot rally on average. In doubles that implies real rallies totaling ~8-12 shots — pretty normal rec-competitive shape. The few 4-5 strike rallies (13 of them) are where you stayed engaged for more touches in a single point. Our detector is still in beta; if you remember a 20-shot dink battle that didn't show up, the threshold is probably splitting it.

Your biggest rally

of your strikes in 8 seconds

At minute 24 you personally hit 9 shots in 8 seconds — about one per second. That's a dialed-in kitchen volley exchange. Total rally including your partner and opponents was probably 15+ shots. HR was in the 130s; you were locked in.

Session 1 · 23:54 → 24:03

04 · Watched happen

A 6.5-minute rest quietly changed your game.

At minute 33 you took a 6.5-min break. Your HR dropped to 108. When you came back, the stats say "same pace" — 57 shots vs 57 shots, nearly identical peak G. But something shifted underneath.

Before rest (min 20–33)

Shots57

Peak G avg6.28

HR avg119 bpm

Dinks20

Drives27

Hard drives10

Smashes1

Your hardest shot14.0g

→

6.5-min rest
HR 131 → 108

After rest (min 39–52)

Shots57

Peak G avg6.25 (−0.5%)

HR avg127 (+8 bpm)

Dinks16 (−4)

Drives34 (+7)

Hard drives7 (−3)

Smashes0 (−1)

Your hardest shot9.8g (−30%)

What the headline misses: you replaced dinks AND hard drives with middle-ground drives. Your hardest shot dropped 30%. Your heart worked 8 bpm harder for the same average force. Either you consciously shifted to a more sustained grinding style — or the rest gave you cardio recovery while your neuromuscular system quietly slipped into fatigue. Either read is coachable. Do you remember feeling any different after the break?

Why this matters

Real coaches watch the game. Wearables watch YOU. Same-player, same-session, quantified pre/post comparisons are the thing only a wearable can do at 50Hz resolution. This is the daily coaching moment — your 14.0g swings disappeared after the rest; your cardio went up 8 bpm — that tells you something about your own body your opponent never saw.

Your top priority next session · synthesized from 3 AI coaches in parallel

Rebalance your shot diet — more dinking, fewer drives.

You hit 167 drives vs 104 dinks across both sessions — drives outnumbered dinks almost 2:1. A common coaching principle: establish a sustained dink exchange before you speed the ball up, rather than attacking on the first attackable ball. You're not wrong to drive — you're just not building enough dink pressure first.

Your drill

10 min of cooperative cross-court dinking before games (aim for 30+ shot rallies, no drives allowed). Then play points where you or your partner must hit 3+ dinks before anyone can speed up.

How to measure it

Your longest rally should grow past 9 strikes of yours. Your Cluster 1 volume (quick dinks) should increase from 61 toward 90+ in a similar session. Your dink % should rise from 31 toward 40.

Your shot mix today

Dinks31%

Drives49%

Hard drives19%

Smashes1%

Only 2 shots above our 12g smash threshold — could mean high-percentage play, or our threshold is too high for your smashes. Do you remember your hardest overheads today?

Methodology

Data

Apple Watch Series 10, 100Hz 9-axis IMU filtered to 50Hz, stored in WR1 binary format. HR via HKWorkoutSession. GPS via watch. Events defined as ≥4g peak wrist acceleration with 400ms debounce.

Shot vocabulary

Unsupervised K-means (k=4) on 800ms-normalized accel magnitude waveforms. No shot labels used. Cluster interpretations are post-hoc based on signature shape + peak G.

Rally detection (beta)

Your shots within 12-second gaps grouped as same rally. Single-shot "rallies" likely serves or isolated returns. Only your shots are tracked — in doubles, ~¼ of total rally shots.

Coaching discipline

Claims labeled SENSED (direct measurement) vs INTERPRETED (possible meaning, alternatives listed). No technique claims from wrist alone — the sensor can't distinguish athletic from uncoordinated motion without video.

Paige — here's what your wrist told usabout how you played today.