The movement evidence room · Independent explanations, cautious conclusions
Claims under examination

Questions About Risk Claims

Clear distinctions make better conversations. Open a question to inspect the claim and its limits.

Can AI predict exactly who will be injured?

No. A model may estimate probabilities for a defined outcome and time window, but it cannot establish an individual future with certainty. Its usefulness depends on validation and the setting.

Read how risk models learn →
Is movement asymmetry always a problem?

No. Differences can reflect task demands, anatomy, sporting adaptation, symptoms or measurement error. Meaning depends on the feature and context, not a universal label.

Put differences in context →
Does a high risk score diagnose an injury?

No. A statistical estimate is not a diagnosis. Symptoms and clinical concerns require assessment by an appropriate qualified professional.

Does a low score mean it is safe to return to sport?

No. A low score cannot rule out injury or establish readiness. Return-to-sport decisions involve a wider assessment of health, function, demands and the athlete’s situation.

Why is a high accuracy figure not enough?

Accuracy can be misleading when outcomes are uncommon or evaluation data are not independent. Ask about sensitivity, specificity, predictive values, calibration and uncertainty.

Inspect performance claims →
What is independent validation?

It is evaluation using data separate from model development, chosen to test the intended use. Another team, season or population can reveal limitations hidden by internal testing.

Will changing a flagged movement prevent injury?

Not necessarily. Association does not show that changing the feature changes injury occurrence. The intervention and the model-guided decision need appropriate evaluation.

Keep outcomes separate →
Are phone cameras less useful than laboratory systems?

Not automatically. Suitability depends on the task, feature, required precision and validation. Convenience does not establish validity, and laboratory complexity does not guarantee practical benefit.

Compare measurement methods →
What should happen to athlete videos and scores?

Organisations should define the purpose, lawful basis, access, retention and sharing arrangements. Athletes should understand how results may affect decisions. Health-related inferences need particular care.

Can I send this website a personal movement video?

No. The evidence desk handles general editorial questions, not personal assessment. Please do not send movement videos, medical records or identifiable health information.

Send a general editorial query →