{"resourceId":"uga-zenodo-di-football-gps-injury-ml-2026","versions":[{"version":"external-68ca006bd3759c6c8566f7596427dc05f6b04bf9bdb5346603e02b11ce0d05d1","resource":{"id":"uga-zenodo-di-football-gps-injury-ml-2026","title":"UGA Zenodo study: DI football GPS ML reaches ~0.878 ROC-AUC for non-contact injury risk amid severe imbalance","organization":"University of Georgia / collaborating NCAA Division I football program","sector":"College athletics / football sports science and injury-risk analytics","geography":"United States (NCAA Division I football; program unnamed in preprint)","publishedAt":"May 25, 2026 (Zenodo v1)","publicationDate":"2026-05-25","eventDate":null,"sourceName":"Zenodo","sourceLabel":"Academic preprint — NCAA DI football GPS injury-risk ML","sourceUrl":"https://zenodo.org/records/20385737","evidenceClass":"academic-research","outcomeClass":"mixed","topics":["infrastructure","knowledge-work","data-security","operating-model"],"finding":"Patel and Rasheed present an end-to-end ML pipeline on de-identified daily GPS monitoring from a prominent NCAA Division I football program to identify elevated non-contact injury risk. With roughly 60 injury cases versus about 15,000 healthy sessions, an LSTM reached ROC-AUC approximately 0.878; ensembles improved sensitivity/specificity balance. Interpretable drivers include sudden spikes in acceleration, deceleration, and session duration, plus elevated risk after rest. Force-plate readiness was largely unavailable. Authors position the work as decision-support for workload management, not perfect prediction.","sledRelevance":"Newly relevant NCAA DI football primary-data signal not in the archive: complements archived Monmouth tennis and USF concussion papers with varsity football GPS methods, and contrasts vendor injury-AI claims by foregrounding imbalance and incomplete readiness data.","evidence":"Zenodo PDF (10.5281/zenodo.20385737): abstract AUC; imbalance counts; feature patterns; Limitations section on imbalance, GPS-only scope, and CMJ gaps; player-independent evaluation language.","architectureImplications":"Interpretation: build athlete-day stores from GPS first; evaluate temporal models under player-independent folds; treat CMJ joins as optional until coverage is complete; ship scores only into staff tools.","governanceImplications":"Interpretation: keep alerts advisory; sports medicine retains unchallengeable medical authority; publish model cards stating imbalance and single-program limits.","securityPrivacyImplications":"Interpretation: de-identify exports; restrict who sees individual risk ranks; avoid secondary AI-training licenses on injury labels without consent review.","caveats":"Zenodo preprint; single unnamed program; extreme imbalance; GPS-external features dominate; not an operational outcome study.","streamIds":["college-athletics"],"roles":{"sales":"Open with sports-medicine and football performance staff drowning in GPS alerts—not a promise to eliminate non-contact injuries. This UGA/DI-football collaboration shows temporal models can rank high-risk athlete-days (LSTM ROC-AUC ~0.878) when positives are rare (~60 vs ~15,000), with acceleration/deceleration/duration spikes and post-rest reload as interpretable cues. Discover current GPS vendors, how often force plates are complete, and who owns load decisions. Bound value to decision-support triage, not guaranteed injury cuts. Do not sell peer-reviewed multi-school validation this preprint does not provide.","engineering":"Fit is athlete-day injury-risk modeling on GPS workload sequences with player-independent evaluation and class-weighted/ensemble methods. Integrate with existing GPS exports into a labeled athlete-day store; keep CMJ/force-plate joins optional until coverage stabilizes; prefer LSTM/ensemble scoring behind staff dashboards. Prerequisites: de-identification, injury labeling focused on last healthy session, and agreement that high sensitivity matters under imbalance. Constraints: single-program data; GPS-only features; Zenodo preprint. Security: treat GPS and injury labels as sensitive student-athlete health/performance data with role-based access.","delivery":"Delivery work is a bounded pilot: replicate labeling, train/evaluate with player-independent folds, and present spike/post-rest explanations to sports science before any athlete-facing score. Dependencies: GPS vendor exports, sports-medicine injury coding, and coaching buy-in that alerts are advisory. Ownership: sports performance owns load changes; sports medicine owns medical decisions. Skills: staff learn imbalance metrics (sensitivity/specificity/G-mean) over accuracy. Governance: no automated remove-from-play without clinician review; document CMJ gaps. Risks: alert fatigue, over-trusting AUC, and deploying without multi-season validation."},"roleDetails":{"sales":{"customerProblem":"Football performance staff collect dense GPS workload data but struggle to turn rare non-contact injury events into trustworthy daily risk triage.","stakeholders":["Head football strength/performance","Sports medicine / ATC","Football coaching staff","Athletics IT/analytics","Compliance/privacy"],"discoveryQuestions":["Which GPS vendor and export grain (athlete-day vs minute) are in use?","How complete are force-plate/CMJ readiness feeds?","Who may change practice load when a model flags risk?","Have any prior models been tested player-independent under class imbalance?"],"valueHypothesis":"If GPS-based temporal models flag elevated-risk days with interpretable spike/post-rest drivers, staff can prioritize human review—without claiming injury elimination.","potentialEngagement":"Four-to-six-week methods pilot: label non-contact events, evaluate LSTM/ensemble baselines with player-independent folds, and brief sports medicine on alert thresholds.","unsupportedClaims":["Guaranteed reduction in non-contact injuries","Multi-school external validation already completed","That force-plate readiness is fully modeled in this study"]},"engineering":{"fit":"Decision-support ML on collegiate football GPS athlete-day sequences for elevated non-contact injury risk.","architectureIntegration":"Batch ingest daily GPS features; build athlete-day labels from sports-medicine coding; train temporal/tabular models offline; serve risk scores and feature explanations to staff tools only.","prerequisites":["De-identified GPS history with injury labels","Player-independent evaluation protocol","Clear human owner for load changes"],"constraints":["Extreme class imbalance (~60 vs ~15,000)","Incomplete CMJ/force-plate coverage","Single DI program; preprint status"],"security":"Restrict injury and GPS identifiers; no athlete-facing raw scores without clinical gate; retain access logs.","proofOfValue":["Held-out ROC-AUC with sensitivity/specificity/G-mean","Documented false-alarm burden at chosen threshold","Sports-medicine review of flagged high-risk days"]},"delivery":{"work":"Pilot labeling pipeline, model card, coach/ATC briefing, and threshold policy for advisory alerts.","dependencies":["GPS data license/export","Sports-medicine coding time","Performance staff scheduling"],"ownership":"Sports performance owns operational use; sports medicine owns medical authority; analytics owns model maintenance.","skillsAdoption":"Staff train on imbalance metrics and interpreting spike/post-rest drivers rather than accuracy headlines.","governanceCheckpoints":["Written advisory-only policy","Clinician sign-off before any remove-from-play automation","Seasonal re-validation"],"acceptanceCriteria":["Player-independent metrics documented","Alert review workflow live for one unit","CMJ gap explicitly listed as out-of-scope"],"risks":["Alert fatigue from low specificity","Overfitting to one program","Privacy incidents from broad score sharing"]}},"retrievedAt":"2026-09-23T03:11:05Z","enrichedAt":"2026-09-23T03:15:27Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Interpretation: less-resourced DI programs need clear imbalance education so AUC headlines do not outrun staffing for alert review.","procurementImplications":"Interpretation: require player-independent metrics, sensitivity at usable specificity, and readiness-data coverage disclosures in injury-AI RFPs.","operatingModelImplications":"Interpretation: separate model ops from medical decision rights; performance staff adjust load only within clinician-approved policies.","sourceVerification":{"openedUrl":"https://zenodo.org/records/20385737","referenceExcerpt":"the LSTM reaching a ROC-AUC of approximately 0.878","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}