SYNTH-32 · Guide
Durability
Every threshold you prescribe against was measured rested, and by hour three the athlete is a different athlete
The decision this framework makes
Whether this athlete's ceiling is aerobic capacity or their ability to still express it after three hours — and, if it is the latter, what the training and the prescription change to.
1. The finding that should unsettle every zone chart
REF-423 (B, endurance-trained cyclists, n=10 per the full text, 2000) is the founding observation: about an hour of moderate cycling (~60% VO2peak) cut 5-min performance power (−12 W) and 30-s sprint power, gross efficiency fell, and the loss of efficiency tracked the loss of 5-min performance (r = 0.91). Fresh numbers stopped describing the rider after an hour of easy work.
REF-427 (B, n=20 well-trained amateur cyclists) and REF-144 (B in the base; n=12 elite/professional cyclists, VO2peak 73.6) take it further: after a 1000 kJ preload at 80% of fresh 20-min power, amateurs lost 10.1% of 20-min power and 10.8% of 5-min power (SDs ± 6.5 and ± 7.8 in the abstract, swapped between the tests in the paper's Table 1; corrected 2026-10-12), and the size of the drop was not correlated with VO2max or relative FTP; after 4 h of intermittent riding at 100 g CHO/h, elites lost 10% of 6-min TT power and 6% of peak power, and none of the fresh laboratory measures (LT1, gross efficiency, maximal fat oxidation, VO2peak) predicted the size of the loss.
In these two cohorts durability behaved as an independent parameter of endurance performance; one professional cohort found the opposite (REF-424, §5). Two athletes with the same FTP are not the same athlete at hour four.
2. What actually moves, and by how much
The moderate-to-heavy transition falls during prolonged exercise, non-linearly (REF-179, B, n=14 trained cyclists/triathletes: after 2 h at 90% VT1 power, VT1 power fell ~10% (217 → 196 W) while the heart rate at that boundary rose 9-10 bpm; REF-143, B in the base, n=12 trained cyclists aged ~40, VO2peak 52: VT1 power fell along a second-order polynomial in 11 of 12 riders, time to a 5% VT1 drop averaged 139 ± 78 min, and that time correlated with time to task failure, 234 ± 66 min, and with fat-oxidation rates from the rested incremental test).Power that was comfortably endurance-domain at the start is heavy-domain by the second or third hour. The athlete's zones move under them while the workout file says they did not — and in the opposite direction for HR and power.
CP fell 9% after 2 h of heavy-intensity cycling on placebo (260 → 236 W), having been unchanged at 40 and 80 min; W′ went earlier and further, −18% at 80 min and −23% at 2 h (REF-145, B+, n=16). Nothing beyond 2 h was measured in that trial; the 3-h moderate-riding result is in the carbohydrate bullet below (REF-1122); a 5-hour figure would still be extrapolation.The practical corollary: late in a long event, avoid surges above CP — W′ is shortened, carbohydrate does not bring it back ([REF-145]), and its reconstitution slows with repeated bouts (REF-131: the slow time constant lengthened from ~388 s to ~716 s after a second bout — SYNTH-28's curve, degraded).
Carbohydrate blunts the decline. 60 g/h negated the 2-h CP fall (254 W vs 236 W on placebo) without a detectable change in muscle glycogen depletion (REF-145; pooled crossovers find a small sparing, ~24 mmol/kg dry weight over ~110 min, REF-1123, B+); after 3 h at 95% of the gas exchange threshold CP fell 14.7% on water, 7.4% at 60 g/h and 4.2% at 120 g/h, and W′ fell with no difference between doses (REF-1122, B, n=16, crossover); after 150 min of moderate riding, VT1 power fell −3 ± 2% with carbohydrate vs −6 ± 4% without (REF-181, B, n=12, crossover), though the 5-min TT afterwards fell in both conditions (−4% vs −10%, not significantly different).Durability is therefore partly a fuelling variable, not purely a trained one — and REF-144 shows even elites at 100 g/h still lose 10% of 6-min power over 4 h, so fuelling raises the floor rather than removing the fade.
The mechanism, from the other side. Making 2 h of moderate riding (90% of VT1, water only) burn ~19% more of the body's own carbohydrate — niacin plus heated thighs, same external work, similar plasma glucose — nearly doubled the VT1 loss: ~7% after the ride alone, ~13% with the faster burn.By a calculated split (n = 7), the extra loss came through cycling efficiency (REF-1121, B, n=11 crossover; substrate data on 7). The 3-min all-out end power (the CP estimate) did not differ between trials, not even after the ride alone. W′ was ~25% lower after the ride (EX vs rested alone P = 0.063), with no extra loss from the faster burn.So in this protocol faster carbohydrate use moved the first threshold but did not detectably move CP or W′. Glycogen itself was not measured.
The fade is partly central. After 3 h of cycling without glucose, blood glucose fell to 3.0 mM and sustained maximal knee-extension force dropped to 197 N versus 222 N with glucose and 248 N at baseline, with reduced CNS activation only in the placebo trial (REF-482, B, crossover; n=8 endurance-trained men) — one reason a well-fuelled athlete "keeps their head" late.
In trail running the drift is biomechanical as well as physiological (REF-548, B, single-session pre-post, n=23 trained trail runners, 11 women, 180 min + three 12-min uphill TTs): uphill TT distance fell 6.6%, stride length shortened and ground-contact time rose, HR and RPE climbed while oxygen cost and leg stiffness were preserved — small, coordinated drifts the authors read as HIGH durability in trained runners.Whether that drift is trainable was not tested (inference).
Sex: one laboratory comparison found women's thresholds slid less (REF-1132, B, 16 women and 16 men recruited above sex-specific VO2peak floors): after 90 min at 110% of the first threshold on water only, power at GET fell 16 ± 15% in men and 2 ± 13% in women (sex × time P = 0.013) and at RCP 13 ± 10% vs 6 ± 9% (P = 0.049), while peak power, VO2peak and neuromuscular function fell alike.The men did 25% more work per kg and rode slightly nearer their second threshold (21% vs 17% of the GET-RCP gap, P = 0.07), and individuals overlapped widely, so part of the sex gap may be a dose gap (inference). No carbohydrate was given in the ride and no performance was measured.
3. When durability is the limiter
Reach for this framework when (Not from a study; the base's own practice for the thresholds):
their best 20-minute power is respectable but their races fall apart late, which is the classic presentation;
decoupling is high on long sessions — Blocks computes HR drift; REF-179 measured the direction (HR at the re-measured boundary +9-10 bpm (142 → 151 bpm) after 2 h) and the ">10% over 3 h" flag is a house convention, not a validated cut-off;
Conversely, when the event is under an hour and the athlete's power-duration curve is simply low, durability is not the answer — build the engine.
4. What changes when it IS the limiter
Not from a study; the base's own practice — no entry here compares durability training methods:
Train the quality in the fatigued state, not fresh. The point of the long ride stops being volume and becomes the work done AFTER two hours: intervals late in the ride, uphill inserts in a fatigued state (the measurement protocol of REF-548, not a training trial).
One long ride, not two short ones (REF-159, B, n=14 well-trained runners, acute): one 6×10-min moderate session showed HR, lactate and RPE rising from first to second half, while the same work split 6.5 h apart reset — the drift is the stimulus we are after, and splitting removes it; adaptation was not measured.
Match the duration you need. REF-427 measured 5-min and 20-min durability; that short-effort durability needs short hard efforts placed INSIDE the long ride is inference from specificity, not a finding of that paper. A rider who has to attack a climb at hour four does not get that from steady riding alone (inference).
Fuel to protect the ceiling — carbohydrate is the one within-ride lever with trial evidence here (heavy strength training has two small controlled trials, REF-146 in cyclists and REF-1144 in runners, §5) (60 g/h in REF-145, CHO vs none in REF-181, 100 g/h as the single tested intake in REF-144, 0 vs 60 vs 120 g/h in REF-1122, where each step kept more CP and none kept all of it; REF-1121 supports the mechanism from the other side — burning more of the body's own carbohydrate in the ride lowered VT1 further); 80-100 g/h as a working target is house practice from the fuelling ladder (SYNTH-02), and the gut needs training to tolerate it.
Read races and long sessions for the fade, not the average. Normalized power for a 5-hour ride hides the shape that matters.
5. Predicting, testing and dosing it
Whether a fresh profile predicts it is split. In 10 young professionals the share of CP kept after 5 × 8 min at 105-110% CP correlated with relative VO2max (r = 0.84) and gross efficiency (r = 0.87) (REF-424, B); in 12 professionals the 2.9% loss of 20-min power after ~4 h was unrelated to ventilatory threshold, peak power or VO2max (REF-196, B); in 13 well-trained riders muscle citrate synthase, complex I and carnosine did not predict VT1 durability (REF-425, B).
Hard prior work costs more than a lot of easy work: in 14 professionals, 5 × 8 min above CP lowered later 1-s and 15-s power more than ~2000 kJ of low-intensity riding, despite less work; 12-min power and CP changed in neither (REF-193, B).
Late in a ride breathing may mark the boundary better than watts (the authors' proposal, not tested as a guidance tool): ventilation at VT1 did not change significantly after 2 h while its power fell ~10% (REF-180, B, n=14).
Fatigued power from ordinary files is reliable: in 18 professionals, best 10-s to 20-min powers after 0-40 kJ/kg were repeatable (SEM < 5%, ICC > 0.90) and declined clearly from ~20 kJ/kg (REF-479, B; amateurs untested). Compare within the athlete only — no gold-standard test exists (REF-141, B+, perspective).
The same reading can be made on foot, against gradient (one case): a speed-gradient-duration model fitted to 77 training sessions' GPS data gave a world-class mountain runner a critical speed for every gradient; in a 100-mile race he averaged 84.8% of it, held about 87.4% for some 5 h, then lost 1.45 percentage points an hour to 74.0% (REF-1194, C+, case study; the authors' method, critical speed extrapolated from training efforts of 10 s-12 min; not validated in other runners).
What may build it, on data its authors call scant (REF-141): years of consistent volume, long sessions with race-pace-or-harder bouts, heavy strength and plyometrics; 12 weeks of heavy lifting raised 5-min power after a 185-min ride from 371 to 400 W in well-trained cyclists (REF-146, B+).In 28 well-trained male runners, 10 weeks of twice-weekly heavy lifting plus jumps cut the rise in oxygen cost over a 90-min run at ~80% VO2max from 4.7% to 2.1% and lengthened a run to exhaustion at 95% vVO2max straight afterwards by 35% (247 → 324 s), with no change in the running-only group and none in fresh economy (REF-1144, B+; completers only, the lifters trained 0.8 h a week more, an open-ended test rather than a race).
How much long riding, by event — the base's own practice, not from a study: crit, track or TT under 40 min, one long ride a month; road race of 60-90 min, a long ride is optional — spread the hours if the week works better that way (SYNTH-84); gran fondo or gravel of 3-5 h, one 4-5 h ride a week in base with efforts late in it; marathon MTB or ultra, two long rides a week in peak preparation, some of 6 h or more; stage race, long rides on consecutive days.All fuelled (SYNTH-02).
Practical Application
When an athlete says "I'm fine until 3 hours and then I die", stop discussing FTP. That sentence is a durability diagnosis, and the block should change accordingly.
Use decoupling as the tracked metric (house convention; the measured alternatives are the fatigued power profile, REF-479, and the REF-427 field protocol) — within this athlete, over time, on comparable rides. Improvement means the same drift at a longer duration or a higher power, not a better fresh number. A field protocol exists: 5-min + 20-min fresh, then 80% of fresh 20-min power to 1000 kJ and repeat; ~10% loss is the amateur average in REF-427.
Put the hard work at the END of the long ride, and say explicitly that being tired going in is the point (house practice).
For a time-crunched athlete this framework mostly does not apply. A 5-hour ride is not available at 6 h/week; the honest version is one longer ride with efforts late in it, and a clear statement that full durability work needs volume they do not have.
What NOT to do
Do not split the long ride to make it fit (REF-159; inference from 60 min of moderate running intervals, acute responses only — the entry does not cover long low-intensity work).
Do not chase durability for a criterium or a 40k TT — wrong limiter, wrong block.
Do not assume a woman will fade less late in a ride because of her sex (REF-1132: one unfuelled laboratory comparison, a ±13% spread within the women, work per kg not matched); read her own decoupling or repeat test.
Limitations
The durability work here is mostly trained-to-elite cyclists in the laboratory (REF-144 elite; REF-179, REF-143, REF-181, REF-427, REF-1122, REF-1132 trained amateurs, Tier 2-3; REF-145 n=16 with no training status in the abstract).The DIRECTION transfers; the magnitudes (~9-10% at 2 h or after 1000 kJ; ~15% after 3 h of moderate riding on water in REF-1122) were measured under single preload protocols and should not be quoted at a recreational rider as though measured on them, nor extrapolated to 5 h.
[REF-423] gives no sample size or VO2max in its abstract (n=10 is full text; tagged Tier 2-3, cycling) and pre-dates the durability vocabulary; REF-427 is a cross-sectional cohort, not a trial.
No entry here compares durability TRAINING methods head to head. That late-ride efforts build durability better than the same efforts done fresh is mechanistically reasonable and widely practised, but this base does not contain the trial. Present §4 as reasoning.
Decoupling thresholds are conventions. The >10% over 3 h figure is a house flag, not a validated cut-off, and decoupling is confounded by heat, fuelling and drift in the sensor itself.
Beyond the trail, running durability here rests on one randomised strength trial over a 90-min treadmill run (REF-1144, 28 men), one cross-sectional study linking long runs to better economy durability (REF-160) and one acute session-split study (REF-159); none ran longer than 90 min, and the biomechanical component of road running likely differs from trail.
Not in the base: durability in women cyclists beyond one unfuelled 90-min laboratory comparison (REF-1132; no field, race or fuelled data, work per kg not matched), the dose-response of long rides, masters-specific durability, which interventions stack.
How this entry was checked and changed
Why this level
regraded 2026-09-27 from B+: the experimental work under it is B/B+ (small crossovers and cohorts); no A-range entry is cited
Content correction 2026-09-27
Headline and §2 "CP can fall 15-20% at 5+ hours without fuelling (REF-145)" -> CP unchanged through 80 min, −9% at 2 h on placebo (260 → 236 W), negated by 60 g/h; W′ −18% at 80 min, −23% at 2 h; nothing measured beyond 2 h.
"REF-427: long rides develop ~20-min durability; 1-min durability needs short efforts inside the ride" -> the paper tested 5-min and 20-min power (−10.8% / −10.1%) and no training; duration-specificity relabelled inference.
"REF-144: elites at 100 g/h lose ~36% glycogen over 4 h" -> abstract: 6-min TT power −10%, PPO −6%, not predictable from fresh LT1/GE/MFO/VO2peak; the glycogen figure is not in the abstract.
REF-143 "Tier 4-5" -> 12 trained cyclists, age ~40, VO2peak 52 (Tier 2-3); abstract numbers added (Δ5%VT1 139 min, task failure 234 min, fat-oxidation correlation); ">10% HR over 3 h" labelled house convention.
"REF-159 (A)" -> B (crossover n=14, runners, acute); "80-100 g/h (REF-144, REF-482)" -> house target (REF-144 tested 100 g/h only; REF-482 gives no dose); REF-548 "form degrades and is trainable" -> small coordinated drifts with efficiency preserved, high durability, trainability not tested; REF-598 removed from the "masters compound the fade" claim; REF-482 numbers added (n=8, 248/222/197 N).
Level parentheticals synced to the cited entries' current Level lines: REF-423 B+ -> B; REF-427 B+ -> B; REF-144 A -> B; REF-179 A- -> B; REF-143 A -> B; REF-145 A -> B+; REF-181 A- -> B.
Level B+ -> B: the parenthetical claimed A-level experimental work; every cited study is B or B+.
Resync 2026-09-28: "REF-423 carries no tier or sport tags in this base" and "(B, n=10 trained cyclists)" -> the entry now carries Tier 2-3 and Sport cycling; its abstract gives no n or VO2max (n=10 is from the full text). Limitation and §1 parenthetical restated.
Revision 2026-10-01: merged SYNTH-12 and SYNTH-15 in as §5. From SYNTH-12: the predictor evidence with the professional split (REF-424 predicts, REF-196 does not), REF-425, REF-193, REF-180, REF-141 with REF-146, the "how much for whom" matrix as labelled practice (SYNTH-50, 58 and 63 quote it), and the research gaps. From SYNTH-15: REF-479 (field durability is repeatable, SEM < 5%, ICC > 0.90, decline clear from ~20 kJ/kg).
Revision 2026-10-01: contradictions with SYNTH-12 resolved in favour of the newer frameworks, none of the old lines carrying stronger evidence. Omega-3 as a durability tool dropped (REF-190 shows no performance change and did not measure durability; SYNTH-67: not ergogenic).Deliberately under-fuelled long rides and fasted Z2 as a durability stimulus dropped (REF-157 is an acute substrate shift only; SYNTH-50, SYNTH-69 and this card fuel the long ride). "90-120 g/h" dropped — grams per hour stay with SYNTH-02.SYNTH-12's power-drift bands (<3% / >8%) and retention grades (>95% excellent …) were unsourced and are not moved; this card keeps its own labelled >10% decoupling flag and REF-427's ~10% cohort average. Not moved: REF-426 (no located source; this card never cited it).Not moved from SYNTH-15: the four-construct matrix, the event → construct tree and percentage weights (house priors on a title-verified viewpoint, REF-477); REF-480 stays as its own entry.
Verification 2026-10-01: 18 cited REF entries opened in KB.md (REF-426/477/480/190/157 appear only in the revision log and were skipped). Every number in merged §5 matches its entry: REF-424 r = 0.835/0.869, n=10; REF-196 −2.9%, n=12; REF-425 n=13; REF-193 n=14, ~2000 kJ, 5×8 min; REF-180 n=14; REF-479 n=18, SEM < 5%, ICC > 0.90, ~20 kJ/kg; REF-141 'data scant'; REF-146 371→400 W.
Verification 2026-10-01: the merge added the REF-424 counter-evidence to §5 but left the lead, §1 and the first cross-reference line asserting the absolute 'fresh measures do not predict'. This is the one substantive defect.
Verification 2026-10-01: all level/design/n parentheticals match the head blocks (REF-423 B, 427 B, 144 B, 179 B, 143 B, 145 B+, 181 B, 482 B, 548 B, 159 B, 424/196/425/193/180/479 B, 141 B+, 146 B+). Level line B is honest; no LEVEL or LEVEL-OVERALL findings.
Verification 2026-10-01: §3 thresholds, all of §4 and the §5 long-ride matrix are labelled house practice in their headers; the 80-100 g/h working target matches the amateur 3-5 h cell of SYNTH-02. No cross-framework contradiction found in a spot-check of SYNTH-02 and SYNTH-50.
Revision 2026-10-08: the 60-90-min road-race row no longer prescribes a weekly 3-h ride; it points to SYNTH-84 (long session optional for events up to ~2 h). Other rows, decision and level unchanged.
Revision 2026-10-13: added REF-1121 (Maunder 2026, J Appl Physiol, read from the full text) to §2, §4 and Cross-References.
Revision 2026-10-13: folded REF-1122 (Norte 2026, B, n=16, full text): CP after 3 h of moderate riding fell 14.7% / 7.4% / 4.2% at 0 / 60 / 120 g/h, W′ fell with no difference between doses. "Without changing muscle glycogen depletion (REF-145)" -> "without a detectable change", with REF-1123 (meta-analysis, small sparing). Headline, §2, §4, Cross-References and Limitations updated; decision and level unchanged.
Revision 2026-10-13: added REF-1132 (Pastorio 2026, Scand J Med Sci Sports, B, n=32, full text) to §2 (new sex bullet), What NOT to do, Cross-References and Limitations; "Not in the base: durability in women cyclists" narrowed to "beyond one unfuelled 90-min laboratory comparison". Decision and level unchanged.
Revision 2026-10-16: added REF-1144 (Zanini 2025, MSSE, B+, RCT n=28 well-trained men, full text) to §4 (heavy strength now has two small controlled trials for durability, REF-146 cyclists and REF-1144 runners), §5 'What may build it', Cross-References and Limitations (the 'nothing covers running durability outside the trail' line now names what does: REF-1144, REF-160, REF-159, none longer than 90 min). Decision and level unchanged.
Revision 2026-10-26: REF-1194 (Mougin 2026, C+) added to section 5 as a field reading of durability in running against grade-specific critical speed; no rule changed; level unchanged.