Methodology
Built on published sports science. All of it cited.
Every number SteadySignal shows you comes from a named, peer-reviewed model — fitted to your own riding, never to an average athlete. This page is the full answer to the question every good coach asks: what is this training actually based on?
The foundation
Where the science comes from
These are the models the engine actually runs. They are decades old because they are still the field standard in 2026 — and each is paired with recent work that confirms or sharpens it.
Fitness, fatigue and form
CTL, ATL and TSB — your rolling fitness, your recent fatigue, and the balance between them (how fresh you are). The Banister impulse-response model (1975) and Coggan's Performance Management Chart, still the standard in a 2022 review of fitness-fatigue modelling.
Your engine, measured
We fit Critical Power (the hardest pace you can hold for a long effort) and W′ (your finite anaerobic battery) from your own rides. Your zones and FTP follow from that model — FTP is derived as CP × 0.95, not a guess off a percentage chart.
Sustainable progression
We watch how fast your fitness climbs — Coggan's guideline is a CTL ramp of ≤5 per week. We also compute ACWR (acute-to-chronic workload), but treat it as a descriptive ramp signal whose limits the research is clear about (Impellizzeri 2020) — never an injury verdict.
Overtraining early warning
Monotony and strain (Foster 1998): training every day the same, with no easy/hard variation, is an early marker of illness and maladaptation. We flag it against your own four-week baseline, in plain language.
How you feel is data
Your morning check-in — recovery, sleep quality, energy, mood — is scored against your own 28-day baseline, because subjective wellness tracks load and often leads the objective markers (Saw 2016, reinforced 2022). When your watch provides them via Health Connect, night-time HRV (7-day average vs your 60-day baseline), resting-heart-rate elevation, and sleep shortfall gently adjust the same readiness signal.
Arriving fresh
The taper evidence (Bosquet 2007, Wang 2023): cut volume, hold intensity, and peak into an event. We surface this as advice near your events — it is advisory, not automated.
Honesty first
What we don't claim
The fastest way to trust a training tool is to know what it refuses to invent. SteadySignal does not claim any of these, and never pretends to:
- No injury or illness prediction — HRV, resting HR, and sleep only nudge your readiness signal against your own baselines
- No single-night verdicts — objective signals count only as rolling averages vs your personal baseline (one bad night is noise)
- No diagnosis, ever — readiness is a training signal, not a medical assessment
- No device "readiness" or "body battery" score
- No blood markers, menstrual-cycle phase or any lab data
- No weight-loss target and no REDs (energy-deficiency) diagnosis
If a number isn't in your riding or your check-in, it isn't in your coaching — and this page is the plain-language version of the document our own code is audited against.
The modern layer
The science didn't stop in the nineties, and neither did we.
The foundational models are sound — but the field has genuinely added to them since 2015. These three are shipped features, computed from data you already produce, not slideware.
Intensity distribution
Polarized or pyramidal? We collapse your time-in-zone into the Seiler three-zone model (easy / threshold / hard) and classify how your last weeks actually split — from your own rides, not your intentions (Rosenblat 2019, Oliveira 2024).
Durability
The "fourth dimension" of endurance: how well your power holds after hours of work, not just how big it is when fresh. Measured fatigued, from the decline late in your long rides (Maunder 2021, Jones 2023).
Fuel for the work required
Per-session carbohydrate targets that scale to the demand of the session in front of you — fuel the long and hard days, ease the easy ones (Impey 2018, Morton 2026).
Transparency
What today's suggestion is based on
The daily answer — go hard, go easy, or go home — is a function of these inputs, all yours:
- Your rides — every session, from Strava, Wahoo or direct .FIT upload
- Your fitted Critical Power and W′ — the model of your own engine
- CTL / ATL / TSB — your fitness, recent fatigue and current freshness
- This morning's check-in, scored against your own 28-day baseline
- How close your next event is
- Your recent compliance — did the last sessions actually happen
- Your weekly available hours
Change any input and the suggestion changes — and the "why" is printed next to every recommendation.
Hours and volume
An elite week and a working week are not the same plan.
The plan builder distributes your weekly available hours across the block's sessions. A 4-hour week and a 14-hour week can share the same structure — the same shape of easy, threshold and hard days — at very different volumes. Hard-day length is capped in proportion to your hours, so a bigger week doesn't hand a time-crunched athlete a session they can't absorb. Your daily session is then sized again to today's readiness.
Honest boundary: the block is a structured, safety-checked template, individualized on your zones, targets, volume and adaptation — not a bespoke, hand-periodized programme from a human coach. We say so plainly rather than dress it up.
Fueling
Fuel for the work required — with the guardrails a coach would want.
Carbohydrate is fuel, and the right amount depends on the session. These are the cited starting ranges — a place to begin and practise in training, never a precise prescription.
Carbohydrate during a ride
By session duration
| Under 45 min | None needed |
| ~45–75 min (hard) | Mouth-rinse or very small amounts |
| 1–2 h | ~30 g carbohydrate/hr |
| 2–3 h | ~60 g carbohydrate/hr |
| Over 2.5–3 h | 60–90 g/hr (glucose + fructose mix needed above ~60 g/hr) |
Carbohydrate across the day
By training load (per kg body weight)
| Light | 3–5 g/kg/day |
| Moderate | 5–7 g/kg/day |
| High | 6–10 g/kg/day |
| Very high | 8–12 g/kg/day |
Protein: 1.6–1.8 g/kg/day, about 0.3 g/kg per meal.
Hydration: drink to thirst; add sodium for long (over 2 h), hot or salty-sweat efforts. More fluid is not safer — over-drinking risks hyponatremia.
Where we stop. Every fueling number is general education, not individual medical or dietetic advice. We set no weight-loss, calorie-deficit or race-weight targets, and we do not diagnose, screen or traffic-light REDs (Relative Energy Deficiency in Sport) — that is physician-led (IOC 2023). Consult a sports dietitian or physician, especially with any under-fueling symptoms.
Adaptation
Miss a session? The plan bends. It doesn't break.
A missed session is detected after about 36 hours. Once a week the plan is reviewed and, if needed, adjusted — at most one change per week, from three transparent rules: ease (back off after a hard stretch), trim (shorten what's ahead), or upgrade (add back when you're fresh and compliant). For coached athletes, every change is held for your coach's approval, and every change is undoable.
Honest non-claim: it does not cram missed workouts back in. Lost training is lost — the plan adapts forward rather than pretending the week still fits.
The AI
The AI narrates. The models decide.
The training decisions on this page are made by deterministic models — the same inputs always produce the same answer. The AI's only job is to write the plain-language summary that explains those numbers back to you, grounded strictly in your own data (on the paid tier). When it is unavailable, a deterministic fallback writes the summary instead.
The AI never designs your training and never adapts it. It explains; it does not decide.
References
Every model, named and public
The complete citation list behind every claim above. Follow any of them — this is the whole point of the page.
- 1.Banister EW (1975). A systems model of training and overtraining; Coggan & Allen, Training and Racing with a Power Meter (CTL/ATL/TSB). Source ↗
- 2.Stephens Hemingway / Imbach et al. (2022). The Use of Fitness-Fatigue Models for Sport Performance Modelling. Sports Medicine – Open. DOI 10.1186/s40798-022-00426-x. Source ↗
- 3.Coggan A. CTL ramp-rate guidance (≤5 CTL/week sustainable). TrainingPeaks.
- 4.Impellizzeri FM et al. (2020). Acute:Chronic Workload Ratio — methodological critique. Source ↗
- 5.Impellizzeri FM et al. (2021). Acute-to-random workload ratio is as associated with injury as ACWR: time to dismiss ACWR and its components. IJSPP.
- 6.Foster C (1998). Monitoring training in athletes with reference to overtraining syndrome (monotony/strain). Med Sci Sports Exerc. Source ↗
- 7.Saw AE, Main LC, Gastin PB (2016). Subjective wellness measures track training load. BJSM. Source ↗
- 8.Subjective vs objective monitoring (2022). Integrative Proposals of Sports Monitoring: Subjective Outperforms Objective Monitoring. Sports Medicine – Open. DOI 10.1186/s40798-022-00432-z. Source ↗
- 9.Bosquet L et al. (2007). Effects of tapering on performance: a meta-analysis. Med Sci Sports Exerc. Source ↗
- 10.Wang et al. (2023). Effects of tapering on performance in endurance athletes: a systematic review and meta-analysis. PLoS ONE 18(5):e0282838. Source ↗
- 11.Stöggl T & Sperlich B (2015). Training intensity distribution among elite endurance athletes. Frontiers in Physiology.
- 12.Rosenblat MA, Perrotta AS, Thomas SG (2019). Polarized vs threshold training — meta-analysis. J Strength Cond Res. Source ↗
- 13.Oliveira et al. (2024). Training intensity distribution. Sports Medicine. DOI 10.1007/s40279-024-02034-z. Source ↗
- 14.Cycling-specific intensity-distribution review (2023). Systematic review of training intensity distribution in cyclists. Source ↗
- 15.Maunder E, Seiler S, Mildenhall MJ, Kilding AE, Plews DJ (2021). The Importance of 'Durability' in the Physiological Profiling of Endurance Athletes. Sports Medicine. DOI 10.1007/s40279-021-01459-0. Source ↗
- 16.Jones AM (2023/2024). Physiological resilience / durability as the fourth dimension of endurance performance.
- 17.Jeukendrup A (2014). A step towards personalized sports nutrition: carbohydrate intake during exercise. DOI 10.1007/s40279-014-0148-z. Source ↗
- 18.Thomas DT, Erdman KA, Burke LM (2016). ACSM/AND/DC Joint Position Stand: Nutrition and Athletic Performance. Source ↗
- 19.Morton et al. (2026). Revisiting the 90 g/hr carbohydrate ceiling. J Nutr, PII S0022-3166(26)00091-X.
- 20.Impey SG et al. (2018). Fuel for the Work Required: a practical approach to amalgamating train-low paradigms. DOI 10.1007/s40279-018-0867-7. Source ↗
- 21.Hew-Butler T et al. (2015). 3rd International Exercise-Associated Hyponatremia Consensus. Source ↗
- 22.NATA (2017). National Athletic Trainers' Association fluid-replacement position statement. Source ↗
- 23.Mountjoy M et al. (2023). 2023 IOC consensus statement on Relative Energy Deficiency in Sport (REDs). BJSM 57(17):1073. Source ↗
SteadySignal is a training analysis tool, not a medical device. Nothing on this page is medical advice. For the full internal specification, see how it works for athletes.
Train on evidence, not vibes.
Free to start — read your own signal against models you can look up.