The science behind the statistics
What each derived figure is, how it is computed, and how far to trust it.
Raw figures (time, distance, heart rate) are shown as recorded. Every figure derived from them has an entry here, and a figure without one is not shown. Ratios without support, such as the acute:chronic workload ratio, are left out on purpose.
- Validated: backed by the literature.
- Validated, with limits: backed by the literature, within the limits given.
- Estimate: useful but without solid validation, always labelled in the app.
Session RPE
Validated. Perceived effort (0 to 10) × minutes. It is the load measure best supported across sports, including where heart rate reads poorly.
RPE (0-10) × minLimits: It relies on how you rate your sessions: only rated sessions count.
Sources: Foster et al. 2001, Haddad et al. 2017
TRIMP
Validated, with limits. Training load of a session: its duration weighted by heart rate, which counts more and more as effort rises.
min × ΔHR × 0.64 e^(1.92 ΔHR); ♀ 0.86 e^(1.67 ΔHR)Limits: It only sees heart rate, so it reflects short efforts and intervals poorly, and depends on your maximum and resting heart rate being right. Comparing it between sports is unreliable.
Sources: Wallace et al. 2014
Fitness (CTL)
Estimate. Your load of the last weeks, each day fading with a 42-day time constant. It describes what you have absorbed.
CTL(d) = CTL(d-1) + (1 - e^(-1/42)) × (load - CTL(d-1))Limits: A description, not a measure of your fitness: it does not predict a performance. It needs about four months of history to settle.
Sources: Clarke & Skiba 2013, Imbach et al. 2022
Fatigue (ATL)
Estimate. Your load of the last days, each day fading with a 7-day time constant. It describes your recent training.
ATL(d) = ATL(d-1) + (1 - e^(-1/7)) × (load - ATL(d-1))Limits: A description of recent load, not a measure of how tired you feel.
Sources: Clarke & Skiba 2013, Imbach et al. 2022
TSB (Training Stress Balance)
Estimate. Fitness minus fatigue: negative when recent load is above what you are used to, positive when it is below.
TSB = CTL - ATLLimits: The -20 and +25 bands are conventions, not validated limits. Form describes a balance: it diagnoses neither overreaching nor detraining.
Sources: Clarke & Skiba 2013, Imbach et al. 2022
Intensity
Validated. Time spent below, between and above your two thresholds, from your heart rate zones: zones 1-2, 3 and 4-5 on a five-zone setup.
5 zones: Z1-2 | Z3 | Z4-5Limits: It describes your training and depends on your zones being right. No distribution is proven better than the others.
Sources: Seiler & Kjerland 2006, Casado et al. 2022
Efficiency
Estimate. What you produce per heartbeat on steady outings of 40 minutes or more: rising means more output for the same heart rate.
output / HRLimits: Heat, hydration and terrain move it too: read the trend, not one outing.
Sources: Rothschild et al. 2025
Aerobic decoupling
Estimate. How much efficiency drops from the first half of a steady outing to the second.
(EF1 - EF2) / EF1Limits: Heat and hydration raise it, and it has no standard definition. The often-quoted 5% limit has no solid basis, so none is shown.
Sources: Rothschild et al. 2025, Coyle & González-Alonso 2001
Runs far beyond your usual longest
Validated, with limits. Runs over 10% longer than your longest of the 30 days before. In a study of 5,205 runners, such single jumps were linked to more injuries than usual.
distance > 1.1 × longest (30 d)Limits: An association across many runners, not your personal risk, and it comes from one observational study.
Sources: Frandsen et al. 2025
Critical speed
Estimate. The speed your best efforts of 2 to 20 minutes point to, fitted as distance = reserve + speed × time. It sits near the limit of what you can sustain.
distance = D′ + CS × t (2-20 min)Limits: These are bests from training, not maximal tests, so it can read low: on real races the method predicted the marathon within about 8%. Not tested on trails, so not shown there.
Sources: Smyth & Muniz-Pumares 2020, Jones et al. 2010
Heart rate by output
Estimate. Your median heart rate at each speed adjusted for slope, or at each power. Lower at the same output usually means fitter.
median HR by band of grade adjusted speed or powerLimits: Field tracking linked to progress with moderate correlations (r 0.43 to 0.61). Heat, fatigue and hydration move it too.
Sources: Vesterinen et al. 2014
How efforts hold
Estimate. How much output dropped, or heart rate rose, from the first to the last third of outings of an hour or more.
(last third - first third) / first third, 60 min or moreLimits: No standard definition exists: this is ours. Heat and dehydration raise it, and no threshold has a solid basis.
Sources: Smyth et al. 2022, Coyle & González-Alonso 2001
Do your records agree?
Estimate. Each record against the pace another one implies, with the usual distance-time exponents. A dot above the band: that distance has room to improve.
T2 = T1 × (D2 / D1)^k, k 1.04-1.08 (to 1.15 for the marathon)Limits: Exponents vary across runners, and bests from training are not always maximal. A check, not a prediction.
Sources: Vickers & Vertosick 2016
Likely race times
Estimate. A range drawn from one best, with the usual distance-time exponents. The marathon range is wider, as most runners fade more than the formula.
T2 = T1 × (D2 / D1)^k, k 1.04-1.08; marathon k 1.07-1.15Limits: It assumes training for that distance and flat roads. It does not predict trail or ultra races, and a training best can stay below your level.
Sources: Vickers & Vertosick 2016
HRV (7-day ln rMSSD)
Validated, with limits. The 7-day rolling mean of the logarithm of your overnight variability. Read it against itself: a lasting change matters more than one night.
mean(ln rMSSD, 7 days)Limits: Watch readings need good nights of data, sources are not interchangeable, and it is not a readiness score.
Sources: Plews et al. 2013, Buchheit 2014
Critical swim speed
Validated, with limits. The pace given by the line between your 200 m and 400 m times. It is linked to swimming performance.
CSS = (400 - 200) / (T400 - T200)Limits: A test pace, not a physiological threshold: it differs from the maximal lactate steady state. It only holds for times swum flat out.
Sources: Wakayoshi et al. 1992, Dekerle et al. 2005
Critical power
Validated, with limits. The power your best efforts of 2 to 20 minutes point to, fitted as work = reserve + power × time. It sits near the limit of what you can sustain.
work = W′ + CP × t (2-20 min)Limits: These are bests from training, not maximal tests: compared with a lab protocol, individual gaps can reach tens of watts. Only hard efforts make it reliable.
Sources: Jones et al. 2010, Karsten et al. 2014
Monotony (7 days)
Estimate. How alike your days of load were this week: the mean over the spread. Even days give a high figure, hard and easy days a low one.
mean(daily load, 7 d) / SD(daily load, 7 d)Limits: From a small study of 25 athletes, and the often-quoted limit of 2 is unverified: read it as a description.
Sources: Foster 1998
GAP
Validated, with limits. The flat pace that costs the same energy as your pace on this slope, from energy costs measured on runners.
v × Cr(i) / Cr(0); Cr = 155.4i⁵ - 30.4i⁴ - 43.3i³ + 46.3i² + 19.5i + 3.6Limits: Measured on a treadmill: it knows nothing of technical terrain, and is less reliable on trails.
Sources: Minetti et al. 2002