An athlete is a trajectory
Each athlete is compared with their own history first. Age group is context, never a verdict.
Data to Intelligence
Athlete development,
Design case
Athlete development,
one record at a time.
Better data before bigger promises.
This design case applies the SAVRN data principle to youth and team sport. It describes a system SAVRN has designed. It reports no performance or injury results, because none have been measured yet.
The coach still coaches
The system prepares the facts. Qualified coaches and medical staff make the decision.
Health data stays private
Recruiters, sponsors and marketing systems never receive pain reports, recovery data or medical history.
The gap
More sensors,
the same thin evidence.
Teams collect more readings every season. The published research behind training and injury decisions is still small, mostly male and rarely tested on a second group.
What athlete studies report about their athletes
Share of 842 athlete studies in sports medicine journals, 2017 to 2021.
Women as a share of study participants
Six sport and exercise science journals, 2014 to 2020: 5,261 publications. Team sport AI: 58 studies.
0 of 204 injury prediction models had been tested on outside athletes
One square per published model. A filled square would be a model tested on a separate group of athletes.
The record
Six things,
for every athlete.
The record connects six things for every athlete, so a later reader can tell an athlete who received a change in training from one who only had a similar reading.
What most datasets hold
What a model needs to learn which actions work
-
Context
Age, training history, sport, position, level of play and growth stage, recorded by trained staff with permission.
-
Plan
The session the coach approved: its purpose, exercises, intended duration and restrictions.
-
Activity
Minutes completed, session type and participation status for every practice and game, including the ones with no injury.
-
Response
The athlete's own report of soreness, fatigue and sleep, kept beside any device data.
-
Decision
What the coach or clinician decided, who approved it and the reason.
-
Outcome
Later test results, availability and health, with lost follow-up marked as unknown.
Illustrative workflow
A player reports new knee discomfort before practice.
- 01 / The report
The athlete says something is wrong.
A short check-in records new knee discomfort. Recent records show more match minutes than usual.
- 02 / The facts
The system shows what it knows.
The assigned staff see the report, the minutes and their dates. The training plan is marked as awaiting review. No diagnosis is generated.
- 03 / The decision
A qualified person decides.
A professional assesses the player. The coach receives only the approved participation instruction, and the modified session is logged.
- 04 / The outcome
The record follows up.
What happened next is recorded over the agreed period. Recruiting and sponsor systems receive none of the health detail.
Scope
What the system does,
and what it leaves to people.
The system supports coaches and medical staff. It does not replace their judgment, and it does not rank a young athlete's future.
The system does
- Keeps one dated record for each athlete, with the source of every entry.
- Routes a concern to the assigned staff and escalates if nobody acknowledges it.
- Shows missing information as missing.
- Attaches the date, the test protocol and the reviewer to every shared assessment.
- Keeps the core coaching service open to athletes who decline a wearable or research use.
The system does not
- Diagnose, prescribe rehabilitation or clear an athlete to return to play.
- Show an individual injury probability to an athlete or a coach.
- Decide scholarships, admission or roster places.
- Join health records to advertising, sponsor or recruiting systems.
- Treat a model built on one sport as proven in another.
Athlete
A short check-in and a clear explanation of the approved plan.
Coach
The roster's changes and missing information, in priority order, before practice.
Medical staff
A restricted queue of symptoms and assessments, separate from recruiting and marketing.
Parent or guardian
A progress view, permitted game video and control over optional sharing.
Recruiter
Dated assessments and approved footage, each with the name of the person who verified it.
Administrator
The means to revoke access and trace every later use of a record.
The evidence
What the research supports,
and what it does not.
Each study below supports one part of the design. None of them shows that software by itself improves development or prevents injury.
Evidence 01
Injury prevention exercise
Four pooled trials of one warm-up program found 39% fewer injuries in recreational and subelite soccer, a rate ratio of 0.61.
- Supports
- Delivering a proven exercise program every week and recording that it was done.
- Does not show
- An effect from software. The trials tested the exercises, in recreational and subelite players.
Evidence 02
Strength training for young athletes
A meta-analysis of 43 studies found moderate gains in strength and jump height, and smaller gains in sprint, agility and sport-specific tests.
- Supports
- Qualified, age-appropriate strength training, with separate tests for strength, jump, sprint and skill.
- Does not show
- A training plan written by a model, or a link from a better test score to better match play.
Evidence 03
Asking the athlete
A review of 56 studies found athletes' own reports tracked training load with better sensitivity and consistency than objective measures.
- Supports
- Keeping what the athlete reports beside the device data.
- Does not show
- One readiness score built from a few questions. A review of 21 studies found single questions related to load anywhere from not at all to very strongly.
Evidence 04
Growth and maturation
A review of 30 studies found growth-related injuries peaked during the adolescent growth spurt, and called for more research in female athletes and in more sports.
- Supports
- Recording growth stage as context, measured by trained staff, in private and by choice.
- Does not show
- One growth threshold for every athlete, or protection from measuring alone.
Evidence 05
Youth development
The International Olympic Committee describes adolescence as non-linear, with physical, psychological and social development moving at different speeds.
- Supports
- Comparing each athlete with their own history before comparing them with a group.
- Does not show
- A forecast of which athlete will succeed.
Evidence 06
Injury prediction
A review of 38 machine learning studies found that small datasets and inconsistent injury definitions hold the models back. A review of 204 models found none tested on outside athletes.
- Supports
- Prediction research that runs in shadow and is tested on outside data.
- Does not show
- An injury probability shown to an athlete, a parent or a coach.
Evidence 07
Workload rules
A randomized trial assigned 18 of 34 elite youth soccer teams to plan training by a common workload ratio for a 10-month season. Health problems did not fall compared with teams that trained as normal, with a relative risk of 1.01.
- Supports
- Recording exposure for every session and leaving the judgment to the coach.
- Does not show
- One safe zone for training load that applies to every athlete.
Evidence 08
Video feedback
A review of 11 school studies found that video feedback with verbal feedback seemed to teach movement better than verbal feedback alone. It said class size, time, equipment and data protection must be weighed first.
- Supports
- A coach-selected clip tied to one learning goal, one practice task and a later review.
- Does not show
- Long-term gains in match play from automated highlights.
The minimum dataset
Ten groups of records,
each with a control.
Ten groups of records make up the minimum dataset. Each one has a control that protects the athlete or the quality of the data.
Identity and permissions
Athlete, guardian relationship, team, permitted purposes and their dates.
The operating ID is kept apart from the research ID.
Development context
Age, training history, sport, position and approved growth observations.
No public body ranking. Growth is assessed by trained staff only.
Planned work
Session purpose, approved exercises, intended duration and restrictions.
The plan is saved before anyone edits it.
Completed exposure
Minutes, participation status, session type and perceived effort.
Missing data never counts as an injury-free session.
Athlete response
Approved questions on symptoms and well-being, suited to the athlete's age.
The wording, the scale and the version are kept.
Performance
Chosen sprint, jump, endurance and skill tests.
The same equipment and protocol each time, with measurement error recorded.
Clinical outcomes
Review, diagnosis where qualified, onset, recurrence, time lost and clearance.
Medical access only. Clinicians own the labels.
Decisions
The alert reviewed, the decision, the approver, the reason and any override.
The record shows what people did, not only what a model said.
Media
Game, source footage, timestamps, roster mapping, review and rights.
A clip inherits the permissions of the recording it came from.
Follow-up
Participation, the outcome window, transfers and dropouts.
Lost to follow-up is kept separate from healthy.
Measures
How a pilot
is measured.
A pilot tests whether the system works. It does not show an effect on development or injury, and the two are measured apart.
Proposed pilot targets
-
Required session and exposure records complete in the final four pilot weeks
Target: 90%
-
Staff time preparing weekly reports, against a time study run before the pilot
Target: 25% less, against the time before the pilot
-
Safety alerts routed to the assigned staff and acknowledged within the agreed service level
Target: 100%
-
Opted-in families that receive a progress summary each reporting cycle
Target: 90%
Injury incidence
New injuries divided by athlete exposure hours, times 1,000. Training and competition are reported apart.
Injury burden
Days lost divided by athlete exposure hours, times 1,000, with symptoms and modified participation beside it.
Availability
Fully available athlete days divided by observed athlete days. Unknown is reported apart from available.
Development
Change within each athlete on tests chosen in advance, with the same protocol each time.
Data quality
Completeness by source and by group, duplicate rate, unresolved identity matches and delay.
Staff value
Reporting time, review time for each athlete, and alerts acted on against alerts dismissed.
Injury definitions, exposure recording and the split between incidence and burden follow the International Olympic Committee recording standard.
Planning sequence
Five phases,
five gates.
Each phase ends at a gate. The next one does not start until the gate is passed.
Weeks 1 to 12 are drawn to scale. The last two phases have no fixed end.
- 01 / Weeks 1 to 2
Discovery and rights
Confirm the scope, map the workflow and select the pilot teams and measures.
Sign-off from the sporting, clinical and privacy leads.
- 02 / Weeks 3 to 6
Data foundation and baseline
Set up identity, consent, roster, exposure and check-ins. Run the baseline time study and test protocols.
Imports reconcile, access tests pass and baseline records are complete.
- 03 / Weeks 7 to 12
Supervised pilot
Two to four teams, about 60 to 120 athletes. Coach workspace, development record and family reports.
Operating measures met, with no open safety or security finding.
- 04 / A full season or longer
Extended evaluation
Development and health are tracked forward. Any model runs in shadow.
Enough follow-up and events, and a locked independent evaluation.
- 05 / After the gates
Controlled expansion
More teams, more video and the models that passed.
Separate scientific, commercial and capacity approval.
Privacy and safeguarding
One program,
separate permissions.
Enrolling in a program is not permission for everything. Service, research, model training, recruiting, media and marketing each need their own.
Children under 13
In the United States, covered online services need verifiable parental consent, and separate consent before a child's data goes to a third party. The rule limits how long data is kept and counts biometric identifiers as personal information.
Federal Trade Commission, January 16, 2025College athletes
Federal guidance places the student health records a campus clinic keeps, at most colleges, under the student privacy law known as FERPA. They are excluded from the health privacy law known as HIPAA.
US Department of Health and Human Services and US Department of Education, 2019Commercial separation
Sponsors receive permitted audience information. They never receive an individual athlete's injury, readiness, growth or recovery data, and health records are never sold.
Safe to speak up
A young athlete should never have to choose between describing a problem and protecting a recruiting profile. Nothing in the system rewards playing through pain.
Where SAVRN fits
The record, the compute
and the review.
Athlete records are small. Video is not. One open smartphone motion capture system needed 31 hours of computing to process sessions for 100 people. Uhlrich et al, PLoS Computational Biology, 2023
Permissions, approvals and audit
Governed records
SAVRN provides the ingestion, the records and the documented path from a recommendation to a human decision.
Capacity sized from measured work
Video and model compute
Capacity is sized from camera hours and processing hours measured in the pilot, not from the number of athletes.
Nine steps
The method
No model reaches a coach until it has run in shadow and passed a test on outside data.
A practical first conversation
Start with the decisions your coaches make.
Bring one sport, the teams you would start with and the decisions you want to support. We will map the record, the permissions and the measures for a pilot.
Is this a deployed system with results?
No. It is a design case. It describes the record, the rules and the measures SAVRN would use. It reports no performance or injury results.
Does the system predict injuries?
No. Published reviews found that injury prediction models are limited by small datasets and have not been tested on outside athletes. Prediction stays in research until a model passes an outside test.
Who sees an athlete's health information?
Authorized medical staff only. Coaches receive the approved participation instruction. Recruiters, sponsors and marketing systems receive none of it.
What happens if an athlete declines a wearable?
The core coaching service stays available. Optional monitoring and research use are separate choices, and declining them carries no penalty.
How are young athletes protected?
A guardian gives permission and the athlete gives assent in terms suited to their age. In the United States, services covered by the children's privacy rule need verifiable parental consent for children under 13.
How long before the system can show an effect?
A pilot of about 12 weeks can show that records arrive, permissions hold and staff can act. An effect on development or injury needs at least a full season and a comparison group.
Sources for this page
- Rates of Reporting and Analyzing Race and Ethnicity in Athlete-Specific Sports Medicine Research. Sonnier et al, Orthopaedic Journal of Sports Medicine, 2024. Systematic review of 842 athlete studies, 2017 to 2021.
- Invisible Sportswomen: The Sex Data Gap in Sport and Exercise Science Research. Cowley et al, Women in Sport and Physical Activity Journal, 2021. Audit of 5,261 publications and 12,511,386 participants, 2014 to 2020.
- Current Approaches to the Use of Artificial Intelligence for Injury Risk Assessment and Performance Prediction in Team Sports. Claudino et al, Sports Medicine Open, 2019. Systematic review of 58 studies and 6,456 participants.
- Just How Confident Can We Be in Predicting Sports Injuries? Bullock et al, Sports Medicine, 2022. Systematic review of 30 studies and 204 injury prediction models.
- Effect of specific exercise-based football injury prevention programmes on the overall injury rate in football. Thorborg et al, British Journal of Sports Medicine, 2017. Meta-analysis of 6 cluster-randomized trials in recreational and subelite soccer.
- Effects and dose-response relationships of resistance training on physical performance in youth athletes. Lesinski et al, British Journal of Sports Medicine, 2016. Meta-analysis of 43 intervention studies in athletes aged 6 to 18.
- Monitoring the athlete training response: subjective self-reported measures trump commonly used objective measures. Saw et al, British Journal of Sports Medicine, 2016. Systematic review of 56 studies.
- Single-Item Self-Report Measures of Team-Sport Athlete Wellbeing and Their Relationship With Training Load. Duignan et al, Journal of Athletic Training, 2020. Systematic review of 21 studies in adult team sport athletes.
- Associations between growth, maturation and injury in youth athletes engaged in elite pathways. Parry et al, British Journal of Sports Medicine, 2024. Scoping review of 30 studies.
- Growth and Maturation Assessment in Youth Sport: Balancing Benefits and Risks. Lundberg et al, Scandinavian Journal of Medicine and Science in Sports, 2026. Narrative review that proposes five principles for growth assessment.
- IOC consensus statement on elite youth athletes competing at the Olympic Games. Bergeron et al, British Journal of Sports Medicine, 2024. International Olympic Committee consensus statement.
- Machine learning approaches to injury risk prediction in sport: a scoping review with evidence synthesis. Leckey et al, British Journal of Sports Medicine, 2025. Scoping review of 38 studies.
- Does load management using the acute:chronic workload ratio prevent health problems? Dalen-Lorentsen et al, British Journal of Sports Medicine, 2021. Cluster randomized trial of 34 elite youth soccer teams. 482 players were on the 25 teams that stayed in the trial.
- Video-based visual feedback to enhance motor learning in physical education: a systematic review. Modinger et al, German Journal of Exercise and Sport Research, 2022. Systematic review of 11 studies in primary and secondary schools.
- International Olympic Committee consensus statement: methods for recording and reporting of epidemiological data on injury and illness in sport 2020. Bahr et al, British Journal of Sports Medicine, 2020. The recording standard for injury and illness data in sport.
- FTC Finalizes Changes to Children's Privacy Rule Limiting Companies' Ability to Monetize Kids' Data. Federal Trade Commission, January 16, 2025. Announcement of the amended children's online privacy rule.
- Joint Guidance on the Application of FERPA and HIPAA to Student Health Records. US Department of Health and Human Services and US Department of Education, 2019. Federal guidance, December 2019 update.
- OpenCap: Human movement dynamics from smartphone videos. Uhlrich et al, PLoS Computational Biology, 2023. Validation in 10 adults and a field study of 100 people.