Skip to main content

DFA a1 — aerobic threshold (top of Zone 2) from your heart rate

Feature Request

Automatically estimate your aerobic threshold (the top of Zone 2) from your ride data using DFA a1, a metric derived from your heart-rate variability. It's the boundary that decides how much easy endurance work you should be doing, and the hardest one to pin down without a lab test.

With a heart-rate chest strap and the AlphaHRV app, IntervalCoach would read this value after your rides and use it to sharpen your Zone 2 and tailor your workouts more precisely.

(Aerobic threshold / LT1 only for now, not a replacement for your FTP.)

Posted by Martijn Russchen ·
2 comments

Log in to comment and vote

Comments2

  • Martijn Russchen

    Team•

    Jul 28

    This is exactly the implementation bar we'd hold it to, thanks. Your artifact-rate point is the deciding one: without per-ride signal quality gating (device allowlist, artifact percentage threshold, fatigue-state exclusion) the number would be confidently wrong for most outdoor files. If we build this, it starts indoor-only on validated straps, with the artifact rate shown next to the result.

  • Sacha Ban

    •

    Jul 27

    DFA a1 correlates with LT1 in controlled studies (DFA a1 = 0.75 corresponds to VT1, R² up to 0.83 against measured oxygen uptake). The concept is sound. Implementation quality will decide whether it's useful or noise.

    Device choice matters. Studies validating DFA a1 use research-grade ECG or the Polar H10, which samples RR intervals at high precision. A Polar H7 comparison against ECG showed an average bias of 4 bpm at the detected threshold. Any chest strap with lower sampling precision or looser electrode contact will shift the result in ways that don't show up in the app, they just produce a wrong number with high confidence.

    Artifact correction is the main failure point. Missed beats as low as 1-3% introduce measurable bias in raw DFA a1 values. At 6% artifact rate, proportional bias reaches up to 19%, depending on the correction method applied. Outdoor rides with cadence changes, climbing efforts, or road vibration produce exactly this kind of noisy RR data. A chest strap on a smooth indoor trainer session will give a cleaner signal than the same strap on a rolling outdoor ride.

    Fatigue breaks the model. DFA a1 only tracks ventilatory threshold reliably in a non-fatigued state. Research on runners found the relationship between DFA a1 and VT1 shifts under fatigue, meaning a session ridden on tired legs can register as a different zone than it actually is. Feeding DFA a1 from a Tuesday ride two days after a hard weekend will not reflect the same threshold as a fresh Monday ride.

    Reliability and validity are not the same thing. A 2024 study found DFA a1-derived thresholds reproducible session to session, but flagged that whether they actually represent the physiological LT1/LT2 boundary is still open. Bias also varies by sex and fitness level, meaning a single fixed 0.75 cutoff does not generalise cleanly across users.

    Net: useful as a directional signal over many rides, not reliable as a single-session number, and only meaningful with a validated chest strap, clean artefact handling, and a non-fatigued state at the time of the ride. None of this is disclosed in the feature description as written. Given the caveats above, it might be worth setting expectations early, perhaps as a beta flag or a short note on requirements, so users get value from it rather than inconsistent readings that undermine trust in the metric.