
What the AI found
“Acute load spiked only when morning resting heart rate dropped below 48 bpm alongside high volume, signalling parasympathetic saturation rather than adaptive fitness.”
Before
28 CSV exports and guessing at fatigue
After
one 4-minute prep note per athlete
The same system, three states — real screens, not a screenshot
| Athlete 04 | Garmin load: 740, Sleep: 7.8h, Soreness: 2/5 |
| Athlete 07 | Polar load: 920, Sleep: 6.1h, Soreness: 4/5 |
| Athlete 11 | Oura readiness: 88, Load: 810, Notes: 'felt flat' |
| Athlete 12 | No export submitted |
Prompt
Analyse these 6 weeks of training stress, HRV rMSSD, and resting heart rate for Athlete 04. Cross-reference subjective fatigue scores against sudden drops in resting HR. Flag non-obvious overreaching markers.
Analyse these 6 weeks of training stress, HRV rMSSD, and resting heart rate for Athlete 04. Cross-reference subjective fatigue scores against sudden drops in resting HR. Flag non-obvious overreaching markers.
AI
Athlete 04 shows a counter-intuitive pattern: their 3 most severe recovery plateaus were preceded by morning RHR falling below 48 bpm (baseline 52) with high TSS. This suggests parasympathetic saturation mimicking fitness, followed within 72 hours by subjective fatigue spikes to 4/5.4 mins / athlete
Review prep
89% agreed in check-in
Flag precision
3 of 14 this week
Flagged athletes
Finding the Hidden Pattern in Athlete Overreaching
How a sports physiotherapist built an integrated review stack that isolates overreaching before athletes break down.
A private sports physiotherapist managing endurance athletes, EU
Tools used
The real tools used here — swap any for your own equivalent. Each links to how we’d set it up.
Starting state
Before anything was set up
Every Sunday night smelled of cold coffee and screen fatigue. The physiotherapist toggled between fourteen open browser tabs: Garmin Connect exports, Polar training logs, and three separate spreadsheet templates. Athletes reported feeling fine right up until their Achilles flared or systemic fatigue forced a mid-season shutdown. The numbers sat in disjointed columns, offering precision without coherence, leaving the clinician to guess whether an athlete was adapting to training volume or teetering on exhaustion.
| Athlete 04 | Garmin load: 740, Sleep: 7.8h, Soreness: 2/5 |
| Athlete 07 | Polar load: 920, Sleep: 6.1h, Soreness: 4/5 |
| Athlete 11 | Oura readiness: 88, Load: 810, Notes: 'felt flat' |
| Athlete 12 | No export submitted |
Working state
All-Access, doing its job
On Sunday evening, the physiotherapist pasted six weeks of anonymised tabular metrics into Claude using the practitioner system prompt. Instead of generic recovery platitudes, the model parsed the correlation between cardiac autonomic indicators and training stress balance. Within seconds, it highlighted a repeating anomaly across three runners that human visual scanning had missed: an unexpected drop in resting heart rate was masking deep autonomic fatigue rather than signalling cardiovascular improvement.
Prompt
Analyse these 6 weeks of training stress, HRV rMSSD, and resting heart rate for Athlete 04. Cross-reference subjective fatigue scores against sudden drops in resting HR. Flag non-obvious overreaching markers.
Analyse these 6 weeks of training stress, HRV rMSSD, and resting heart rate for Athlete 04. Cross-reference subjective fatigue scores against sudden drops in resting HR. Flag non-obvious overreaching markers.
AI
Athlete 04 shows a counter-intuitive pattern: their 3 most severe recovery plateaus were preceded by morning RHR falling below 48 bpm (baseline 52) with high TSS. This suggests parasympathetic saturation mimicking fitness, followed within 72 hours by subjective fatigue spikes to 4/5.Use case implemented
The finished system, running on its own
Fourteen endurance athletes now upload their training export every Sunday evening into a shared Google Drive folder. A single structured template parses the training stress balance, sleep duration, and subjective fatigue ratings into three distilled observation points. The physiotherapist reviews the entire squad in under an hour on Monday morning, speaking directly to flagged recovery bottlenecks before the athletes load their training blocks for the week.
4 mins / athlete
Review prep
89% agreed in check-in
Flag precision
3 of 14 this week
Flagged athletes
What an outside observer would notice
65 mins vs 4.5 hrs
Weekly review time
93% up from 58%
Athlete check-in adherence
The stack — build it yourself
Standardises disparate CSVs from Garmin, Polar, and Apple Watch into uniform weekly rows.
Excels at parsing dense tabular inputs and surfacing subtle non-linear physiological interactions without hallucination.
Low-friction, calm interface for four-line athlete talking points during Monday morning consults.
These are the tools used in this story. Any can be swapped for an equivalent you already trust.
Go deeper
Do this yourself
Assemble your practitioner stack
This story runs on All-Access. The tools and prompts above are the real build — swap any tool for your own equivalent and follow the same steps.