SCT-RPE: An 8-minute research-backed cycling fitness test you can try in Intervals.icu

Thanks for the thoughtful feedback!

I agree that experienced cyclists generally have a better sense of their perceived effort. That said, our sample was relatively heterogeneous despite all participants being classified as trained cyclists, with VO2peak ranging from 55 to 78 mL/kg/min and Wpeak from 211 to 417 W. Despite these differences, participants were able to self-regulate their effort using the standardized RPE anchors, with encouraging test-retest reliability. Of course, whether this approach works equally well in less-experienced cyclists remains to be investigated. Full participant characteristics are available in the paper, which is currently free to access.

Regarding your suggestion, that’s actually quite close to what we’re proposing! The SCT-RPE measures power and HR during predominantly submaximal exercise, but instead of prescribing a fixed power or HR target, cyclists regulate their effort using standardized RPE anchors. This removes the need for prior threshold calibration or separate RPE recordings after each stage.

As for your main question, the goal is primarily longitudinal fitness monitoring rather than FTP/CP prediction. We’re looking for a practical, repeatable, low-burden way to monitor endurance-related fitness between maximal performance tests, not replace them.

Our initial results showed promising cross-sectional validity and test-retest reliability. Whether the test can reliably track within-athlete changes over time is precisely what we’re investigating in our ongoing longitudinal study.

On a related note, I think the new automatic submaximal fatigue testing feature @david implemented in Intervals.icu is really cool! Combining HR responses and RPE with HRV, sleep and other wellness metrics, along with explicit decision rules for interpretation, is a great example of how complementary approaches can be used for athlete monitoring.

Your ML approach sounds interesting too! I’d be curious to hear more about the inputs you’re using to estimate CP.