1. Athlete Library
Run Robot stores athlete profiles in this browser and can export the same record as Athlete Brain JSON.
Shared Athlete Brain Β· Command Centre
HEALTH β FITNESS β PERFORMANCE status will appear here when a shared athlete is loaded.
V2 reads and updates the existing JBST-AthleteBrain-1.1 record. It preserves swim, bike, goals, nudges, evidence and history while updating the run branch.
2. Athlete Basics
3. Power & Physiology
CP is used to create a race-power range, not a single magic wattage. Imported files will later let Run Robot learn how this athlete actually converts power into pace.
4. Stand-alone Run PBs
Distance
PB time
Date
5 km
10 km
21.1 km
42.2 km
When two or more PBs are present, Run Robot estimates the athlete's own fatigue / durability exponent rather than relying only on a generic conversion formula.
5. Triathlon Run PBs
Tri run
PB time
Date
Sprint Β· 5 km
Olympic Β· 10 km
70.3 Β· 21.1 km
Ironman Β· 42.2 km
As matching fresh and triathlon PBs are added, Run Robot can estimate the athlete's observed bike-to-run penalty.
6. Garmin / Stryd / Run File Import EVIDENCE
Drop or choose run files
Garmin / Stryd FIT, TCX, GPX and common Stryd CSV exports are accepted. V2 analyses the file, classifies the run, and extracts efficiency / drift where record data permit. Direct Garmin account sync still requires Garmin OAuth/backend access.
Garmin / Stryd FIT, TCX, GPX and common Stryd CSV exports are accepted. V2 analyses the file, classifies the run, and extracts efficiency / drift where record data permit. Direct Garmin account sync still requires Garmin OAuth/backend access.
No run files imported.
7. Imported Evidence
Sessions
0
imported
Latest distance
β
km
Latest pace
β
min/km
Latest power
β
watts
Latest HR
β
bpm
V2 Run Intelligence
Race / TT evidence
0
auto-identified
Z1 / aerobic runs
0
usable trend sessions
Z1 efficiency
β
m/s per 100 W
Efficiency trend
β
recent vs earlier
Median HR drift
β
aerobic decoupling
Long-run durability
β
late-session efficiency
Latest % CP
β
power / Critical Power
Data quality
STARTER
records with power + HR
Import several runs to establish athlete-specific efficiency and durability trends.
Z1 Efficiency History
Automatic Session Read
No imported sessions yet.
Run History Β· editable classification
| Source / date | Class | Distance | Time | Pace | Power | % CP | HR | Efficiency | HR drift |
|---|---|---|---|---|---|---|---|---|---|
| No imported sessions yet. | |||||||||
8. Race / Course Model
Add at least one PB, then run the model.
9. Course Profile
Optional GPX race course
Upload a course to add distance and elevation context.
Upload a course to add distance and elevation context.
No course loaded. Flat / neutral course assumed.
10. Fuel & Energy Model CARB β FAT
Model anchor: RER/RQ β 1.00 represents carbohydrate-dominant oxidation; β0.70 represents fat-dominant oxidation. V2 assumes carbohydrate dominance near CP, then progressively lowers estimated RER as intensity falls and duration extends. This is an estimate until athlete-specific gas-exchange data exist.
Fuel Model Output
Total energy
β
distance-based kcal
Power cross-check
β
kcal Β· not primary
Mean est. RER
β
duration-adjusted
Carbohydrate used
β
g total
Fat used
β
g total
Carb demand
β
average g/h
Suggested intake
β
g/h range
Planned absorbed CHO
β
g total
Endogenous CHO draw
β
g from stores
Estimated CHO remaining
β
editable starting pool
Fuel risk
β
scenario flag
Energy cost drift
β
ultra fatigue allowance
Run the performance model to calculate fuel use.
| Time | % CP | Est. RER | CHO energy | Fat energy | CHO demand | Fat use |
|---|---|---|---|---|---|---|
| No fuel model yet. | ||||||
RUN ROBOT Β· PERFORMANCE MODEL
Build the run brain.
Store the athlete, add PBs and import real running files. Run Robot will combine performance history, durability and scenario adjustments into a race model.
Coach read
No model yet. The most useful first evidence is a recent PB plus a longer-distance PB or a Stryd/Garmin run file.
Model Results
Predicted time
β
scenario total
Target pace
β
min/km
Fatigue exponent
β
athlete-specific if possible
Power target
β
range from CP
Power / kg
β
target range
PB anchor
β
prediction source
Tri penalty
β
fresh β tri run
Course gain
β
metres climbed
Carbohydrate
β
race intake range
Evidence level
β
model confidence signal
Performance Curve
1% Better Opportunities
Run a model to identify the next useful data or performance opportunity.
What V1 is learning
Speed
Fresh PB curve across 5K, 10K, half and marathon.
Fresh PB curve across 5K, 10K, half and marathon.
Durability
How quickly performance drops as race duration grows.
How quickly performance drops as race duration grows.
Tri conversion
How much slower the athlete runs after the bike versus fresh.
How much slower the athlete runs after the bike versus fresh.
Power
CP-based race ranges plus imported real-world power evidence.
CP-based race ranges plus imported real-world power evidence.
Efficiency
Future versions can compare pace per watt, HR drift, shoes and terrain.
Future versions can compare pace per watt, HR drift, shoes and terrain.
Health β Fitness β Performance
The Run Brain is designed to grow into durability, injury, footwear and S&C decisions.
The Run Brain is designed to grow into durability, injury, footwear and S&C decisions.
V2 Science / Model Guardrails
Running calories: primary estimate uses body mass Γ distance Γ an editable default of 1.00 kcal/kg/km. Published trained-runner values cluster close to ~0.97β1.09 kcal/kg/km.
Running power: power is used for intensity, efficiency and a secondary calorie cross-check. V2 deliberately does not treat Stryd watts as identical to bicycle crank watts.
Substrate split: estimated RER is converted to carbohydrate/fat energy proportions using the conventional non-protein range of ~0.70 (fat) to ~1.00 (carbohydrate). The CPβRER curve is a coaching heuristic, not a substitute for athlete-specific indirect calorimetry.
Long duration: 6 h and 24 h treadmill studies show RER falls during prolonged running; the 24 h study found the main metabolic changes occurred in the first ~8 h. Energy-cost findings are mixed, so V2 uses only a modest, editable-style guardrail of up to ~5% cost drift rather than assuming a large deterioration.
Ultra prediction: predictions beyond marathon distance are explicitly extrapolations. They become much stronger once long-race / ultra evidence is imported; fuelling, terrain, heat, GI tolerance and muscular durability increasingly dominate.