No black box. Here's the science.
Every number in Motus comes from an established sports-science model, computed from your own data. Here's exactly what we use — and where you see it.
Training load & fitness
Banister · Coggan TSSYour fitness (CTL), fatigue (ATL) and form (TSB) come from the Banister fitness-fatigue model, driven by Training Stress Score (TSS) computed from power, heart rate or pace. The same math serious coaches rely on — calculated transparently from your activities.
In Motus: the Performance Manager Chart, and every plan decision.
Power & capacity
GoldenCheetah CP2/CP3 · W'balYour critical power and anaerobic work capacity (W') are fitted from your own efforts with the 2- and 3-parameter CP models (GoldenCheetah-derived). W'bal then tracks how much of that capacity you've spent through a session, second by second.
In Motus: the Power Duration Curve and the W'bal widget.
Aerobic fitness (VO2max)
VO2max estimationVO2max is estimated from the relationship between your pace or power and heart rate on aerobic efforts, then tracked over time — so you see a fitness trend, not a one-off number.
In Motus: the VO2max trend widget.
Injury-risk load (ACWR)
Acute:Chronic Workload RatioThe Acute:Chronic Workload Ratio compares your recent (acute) load against your longer-term (chronic) base. Sharp spikes flag elevated injury risk — a well-studied early-warning signal.
In Motus: the ACWR dashboard.
Race prediction
Riegel + load adjustmentExpected race time uses the Riegel endurance model across each leg, adjusted for your fitness (CTL) and freshness (TSB) on race day. Your goal probability then breaks down into seven transparent factors — so you can see exactly how it was computed.
In Motus: goal probability and expected-time.
Biomechanics from video
MediaPipe BlazePose · Butterworth · Phase portraitPose is estimated with MediaPipe BlazePose, smoothed with Butterworth filtering, and analyzed with phase-portrait and coordination (CRP) methods to measure joint angles, cadence, symmetry and ground-contact time.
In Motus: Virtual Bike-Fit, Running Gait Analysis and Stroke Analysis.
Research-grade pose estimation from 2D video — best with clear, side-on footage. It's a training aid, not a medical or clinical assessment.
AI never guesses your training.
AI does what it's genuinely good at — reading your technique on video and explaining your data in plain language. But your training load, zones, plan and predictions are computed by the models above, not generated by a chatbot. Every recommendation traces back to one of them — and you can overrule any of it.