SSMT

What's real in sports tech right now

Signals — what's emerging, with the one question we're watching.   Grades — what we've actually assessed against the evidence.

Draft — a validation surface, nothing here is published. Signals are forward-looking and lighter than grades; a signal is never a verdict.
🔬 Lab-stage Computer Vision
One phone, propped on a chair, is now within 5.5° of a motion-capture lab.
A single phone camera is catching up to a motion-capture lab
The open question Untested outside a lab floor, on real patients, in normal clothes.
We graded this → SSMT-2026-0006
Graded Wearables & Recovery
Your wrist is honest on an easy jog and guessing on your last hard interval.
Partially supported drift 3/7
On the box “Accurate heart rate, right from your wrist.”
Accurate for steady cardio in healthy, light-skinned adults — not accurate, full stop.
Full evidence → SSMT-2026-0001
📈 Heating up Computer Vision
The match you watched on TV can now become 3D tracking data.
Ordinary match footage is turning into 3D tactical data
The open question Not yet checked independently against the stadium systems it would replace.
Not yet graded
Graded Wearables & Recovery
If you lay awake for an hour last night, your ring probably scored it as sleep.
Partially supported drift 5/7
On the box “See your deep, REM and light sleep every night.”
Directionally useful in aggregate, unreliable for any single night — and the accuracy that is well established is a different measurement from the one being sold.
Full evidence → SSMT-2026-0002
🌱 Just appeared AI Athletes
An AI is learning to move a human skeleton, one muscle at a time.
AI is learning to run a simulated human body, muscle by muscle
The open question A simulation milestone. No demonstrated transfer to a real athlete.
MyoSuite / MyoChallenge ↗
Graded Wearables & Recovery
It won't make you fitter than a fixed plan. It gets you there on fewer hard days.
Partially supported drift 4/7
On the box “Train when your body is ready.”
The supported finding is an efficiency one, not a superiority one.
Full evidence → SSMT-2026-0003
📈 Heating up Wearables & Recovery
A patch on your arm reads your sweat while you run.
Your sweat is becoming a live readout
The open question Whether acting on it beats drinking when you are thirsty is untested.
Microfluidic epidermal sweat sensing — Nature Biotechnology review literature ↗
Graded Wearables & Recovery
Your watch says rest today. We couldn't find one published test of that.
Unevaluated drift 7/7
On the box “Your readiness score tells you when to train hard and when to back off.”
Never tested, in any design. Currently unevaluable by independent parties.
Full evidence → SSMT-2026-0004
🔬 Lab-stage Wearables & Recovery
Your shirt is turning into the sensor.
Clothes that read your muscles
The open question Lab demonstrations only. Durability and washing have not been reported.
npj Flexible Electronics — textile strain-sensing literature ↗
Graded Computer Vision
It reads knee bend like a lab. Ask about hip rotation and it's guessing.
Partially supported drift 4/7
On the box “Lab-grade biomechanics from your phone.”
Genuinely good in the sagittal plane; not usable for rotations; and 'from your phone' is the least-evidenced configuration.
Full evidence → SSMT-2026-0006
👀 On our grading radar Wearables & Recovery
The ice bath makes you feel recovered. It may also be undoing the session.
Cold plunges are everywhere — and the evidence is genuinely split
The open question We are grading this next. The split appears to be about timing.
Not yet graded
Graded Biomechanics
The insole measures force well. Tomorrow the same athlete gets a different asymmetry number.
Partially supported drift 3/7
On the box “Know your injury risk before it happens. (measurement component)”
Force magnitude yes; the asymmetry number itself does not reproduce.
Full evidence → SSMT-2026-0007
🌱 Just appeared Biomechanics
The ball tells you your shot was flat before the coach does.
A basketball that grades your jump shot
The open question Whether that feedback actually improves shooting has not been shown.
Not yet graded
Graded Injury Science
836 runners were followed for six months. The lopsided ones got hurt less.
Contradicted drift 5/7
On the box “Know your injury risk before it happens.”
The best available evidence points the other way.
Full evidence → SSMT-2026-0008
👀 On our grading radar Wearables & Recovery
Built for diabetes. Now sold to tell you when to eat mid-race.
Glucose monitors jumped from diabetes to endurance sport
The open question We graded it: Unevaluated. During exercise it reads 14–34% off.
We graded this → SSMT-2026-0009
Graded Wearables & Recovery
Built for diabetes. During exercise it can read a third off.
Unevaluated drift 4/7
On the box “Fuel smarter. See how your body responds in real time.”
Bordering Partially supported for the narrower claim that CGM tracks glucose trends during exercise.
Full evidence → SSMT-2026-0009
📈 Heating up Biomechanics
A powered belt makes running measurably cheaper. It's also banned in competition.
Exoskeletons have started making runners measurably faster
The open question Metabolic savings in a lab are not race results.
Not yet graded
Graded Biomechanics
Same lift, two devices. One is accurate to ±0.03 m/s. The other isn't close.
Supported drift 3/7
On the box “Know exactly how fast you lifted.”
Supported for linear transducers and camera devices. Contradicted for IMU/accelerometer devices.
Full evidence → SSMT-2026-0010
📈 Heating up Computer Vision
Your gym can now film your squat and tell you which knee is cheating.
Markerless capture has left the lab and turned up in commercial gyms
The open question Side-on angles hold up. Rotation still does not. Ask which one they are selling.
We graded this → SSMT-2026-0006
Graded Biomechanics
It won't add to your squat. It might add to your jump, on less volume.
Partially supported drift 3/7
On the box “Autoregulate every set.”
Equal on strength; small real gains on jump, sprint and change of direction, often at lower volume.
Full evidence → SSMT-2026-0011
🔬 Lab-stage Computer Vision
One rig that claims to do gait, running, tennis and volleyball.
One rig, any sport: markerless is going general-purpose
The open question General-purpose often means validated for nothing in particular.
We graded this → SSMT-2026-0006
Graded Wearables & Recovery
The boots make you feel better. Across 8 trials, nobody performed better the next day.
Partially supported drift 4/7
On the box “Recover faster. Train harder tomorrow.”
Accurate about how you feel; unsupported about what you can then do.
Full evidence → SSMT-2026-0012
🔬 Lab-stage Computer Vision
Markerless Motion Capture in Routine Clinical Upper Limb Assessments: Validity and Insights Beyond Ordinal…
The Action Research Arm Test (ARAT) is a widely-used upper limb outcome measure in neurorehabilitation, but its ordinal scoring is subjective and suffers from limited sensitivity and specificity. We evaluated whether artificial-intelligence (AI)-based marke…
The watch NOT GRADED — auto-discovered from arXiv and not yet read by an editor. A pointer to the source, not a verdict.
Tim Unger et al. — arXiv ↗
🔬 Lab-stage Computer Vision
Pose-to-Biomechanics: Bridging 3D Human Pose Estimation and Biomechanical Attribute Prediction
Recent progress in 3D human pose estimation has made markerless recovery of skeletal motion increasingly accurate and scalable. However, most pose estimators remain optimized for geometric keypoint accuracy, while many real-world applications in rehabilitat…
The watch NOT GRADED — auto-discovered from arXiv and not yet read by an editor. A pointer to the source, not a verdict.
Ayda Eghbalian et al. — arXiv ↗
🔬 Lab-stage Computer Vision
Biomechanics-aware Multi-view Markerless Motion Capture of Dexterous Hand Movements
Markerless motion capture (MMC) techniques have been widely beneficial in biomechanical analysis of human movement; however, application to complex motions of the hand lags other musculoskeletal systems. The primary goal of this study was to evaluate the pe…
The watch NOT GRADED — auto-discovered from arXiv and not yet read by an editor. A pointer to the source, not a verdict.
Pouyan Firouzabadi et al. — arXiv ↗
🌱 Just appeared Computer Vision
qualcomm/MediaPipe-Pose-Estimation
A public model on Hugging Face — 522 downloads · 111 likes.
The watch NOT GRADED — auto-discovered from Hugging Face and not yet read by an editor. A pointer to the source, not a verdict.
qualcomm — Hugging Face ↗
🌱 Just appeared Computer Vision
MCG-NJU/SportsAction
Dataset Card for MultiSports Dataset Summary Spatio-temporal action localization is an important and challenging problem in video understanding. Previous action detection benchmarks are limited in aspects of small numbers of instances in a tri…
The watch NOT GRADED — auto-discovered from Hugging Face and not yet read by an editor. A pointer to the source, not a verdict.
MCG-NJU — Hugging Face ↗
🌱 Just appeared Computer Vision
GEM/sportsett_basketball
SportSett:Basketball dataset for Data-to-Text Generation contains NBA games stats aligned with their human written summaries.
The watch NOT GRADED — auto-discovered from Hugging Face and not yet read by an editor. A pointer to the source, not a verdict.
GEM — Hugging Face ↗