SYNTH-33 · Guide
VO2max intervals
The interior of the rep is the prescription, and a flat 4×5 is a choice that needs a reason
The decision this framework makes
What SHAPE a VO2max session takes for this athlete, how often it can appear, and when the plain version is the right call anyway.
1. The principle everything else follows from
REF-041 (B, prospective single-arm intervention, n=22 well-trained cyclists, VO2max 67, Tier 3): across 21 sessions of 5 × 8 min at 40-min power with VO2 measured in every session, the % of VO2max sustained during the work intervals was the measure that best reflected the size of the adaptation — R² 0.54 for ΔVO2max, 0.44 for maximal aerobic power, 0.25 for Δpower at 4 mmol/L after nine weeks.
In this one programme adaptation tracked how high the rider's VO2 sat during the work, with the minutes identical for everyone.
Two consequences, and the second is routinely missed:
The interior of the rep is the prescription. "4×5 min at 110%" specifies almost nothing about the thing that predicts the outcome.
%HRmax is not a full surrogate for %VO2max (REF-041: %HRmax related to fewer outcomes and repeated worse session to session). Heart rate lags and plateaus; two sessions with identical HR traces can differ substantially in time ≥90% VO2max (inference from the poorer repeatability, not a measured pair). Prescribing and judging purely by HR under-reports exactly the variable that matters.
2. The formats, ranked by what this base actually shows
30/15 — 30 s @ 118% / 15 s @ 60% of 40-min power, in 6 × 8-min blocks (REF-046, B+, randomized crossover, n=19 well-trained cyclists, 11 women and 8 men, Tier 3): at identical mean power, 1123 s ≥90% VO2max versus 879 s for continuous work and 917 s for 60/60 — the best format tested here for time near VO2max — and the result did not differ between women and men and was not related to fibre-type proportion, capillarization or citrate synthase, so no phenotyping is needed before prescribing it. REF-257 (B+, crossover, 16 male team-sport athletes) adds that 15-s work bouts gave higher VO2 and more time ≥90% VO2max than 30-s bouts at the same volume, with lower lactate, lower leg and breathing RPE (overall RPE not significantly different) and less loss of knee-extension force — 15-s work at a lower fatigue cost than 30-s work; the study did not test 30/15, so carrying this over to 30/15, and the claim that 30/15 sits in a week more easily than 5-minute reps, are inference.
Decreasing ladder (HIDIT) — 3 → 2 → 1 → 0.75 → 0.5 min work with recovery 2 min → 20 s, same work and recovery power (the step sequence and ~117% / ~83% CP are from the entry's full text; the abstract gives 3 min → 30 s work, 2 min → 20 s recovery) and the same 3:2 ratio (REF-130, B in the base, crossover, n=12 amateur cyclists, Tier 2): 312 s above 90% VO2peak versus 182 s for classic short (30/20) and 179 s for classic long (3/2) intervals, with time to exhaustion, lactate and RPE not different.It is the curve of SYNTH-28 applied inside one session.
Fast-start, or over-under, long intervals (REF-700, B, crossover, n=10 well-trained XC skiers; REF-194, B, crossover, n=11 well-trained XC skiers): in speed- and duration-matched 5 × 5-min roller-ski intervals, a fast-start-then-decreasing rep and an over-under rep both produced more time ≥90% VO2peak and a higher mean %VO2peak than constant speed, were equivalent to each other, and cost the same HR, lactate, ventilation and RPE (REF-700); the fast-start format alone (1.5 min at 100% then 3.5 min at 85% of maximal aerobic speed vs a constant 90%) raised mean VO2 (87.6 vs 86.1% VO2peak) and peak VO2 (98.2 vs 95.4%) with a LOWER RPE (16.1 vs 16.5) and no difference in HR, lactate or leg-press power (REF-194).Not from a study; the base's own practice: the cycling translation "open hard, or alternate ~40 s @ ~120% FTP with ~1 min @ ~105% FTP" is an approximation of the skiers' MAS/LT2 anchors, not the studies' units.
4×8 min (REF-150, B+, RCT, n=35 trained recreational cyclists): the classic, and still defensible — 4×8 at 90% HRpeak (32 min of work) gave an 11.4% composite gain versus 5.5% for 4×4 at 94% HRpeak and 5.6% for 4×16 at 88%; adaptation depends on the interaction of intensity and total work duration, so accumulated minutes at a high enough intensity are the dose.
Norwegian national-team I-5 cycling models (Olympiatoppen guideline, level C — expert practice, not a study): 6 × 6 min / 3 min with a 30-s surge opening each rep (labelled I-4–5); 30/15 × 13 × 3 at 125-130% / 50% FTP with 3 min between sets; 5 × 5 min / 2.5 min; pyramid 8-7-6-5-4 min / 2 min — the last reps at the same or a higher external load than the first.Their rest rule: ~50-75% of the rep in I-5 (~50% in I-4, ~25% or 1-3 min in I-3), sometimes longer, e.g. before competitions (REF-1077, C). Practice descriptions, not tested against each other or against the formats above.
3. Frequency, and the ceiling
**REF-005 (B in the base; n=8 highly trained runners, VO2max ~72, uncontrolled, two consecutive 4-week phases) tested one vVO2max session a week and then three: one raised vVO2max (20.5 → 21.1 km/h, via better economy); three, in the following four weeks, changed no performance variable and raised end-of-test noradrenaline (11.1 → 26.0 nmol/L) — a strain marker, not overtraining syndrome.
Two a week — the usual cap — was never tested**, and because the phases were sequential, "three a week" is also "the second month". Treat the cap as coaching convention sitting between a tested success and a tested "nothing extra", and rotate formats (ladder / short-short / classic) rather than repeat one. At 5-8 h/week that is the entire quality budget (SYNTH-26).
Cadence: Not from a study; the base's own practice: cue a moderate 85-95 rpm during VO2max work.
What REF-255 (A- in the base; systematic review, 13 articles) actually reports is that cycling at a high cadence AND high intensity impaired neuromuscular function and performance more than low-cadence cycling, that it remains unclear whether cadence affects the onset of fatigue, and that the heterogeneity of the studies prevents a cadence prescription — so the cue is "not unusually high", not a validated band.
4. The honest counterweight
REF-421 (B+ in the base; within-subject crossover, n=22 well-trained cyclists, VO2max 69.5, 1 woman) is the entry that stops this becoming dogma: a one-week block of MODERATE-intensity intervals (5–7 × 10–14 min, RPE ~14.5) improved 15-min power as much as a HIT block of 5 × 8.75 min of short intervals (4.9% vs 2.8%, no difference), improved power at 4 mmol/L MORE (4.5% vs 2.1%, p = 0.03), and VO2max change did not differ between blocks. Individual responses varied widely (PO15min +4.9 (8.7)%).
The "% VO2max predicted the gain only within the HIT block" analysis once quoted here is not in the paper's abstract and has been removed from the entry — do not cite it.
So: shape the rep when the session's target is VO2max, but do not tell an athlete that unshaped moderate work does not develop them, and do not conclude a block failed because one athlete responded less.
For the genuinely time-crunched, REF-696 (A-, meta-analysis of 16 RCTs, 318 healthy young participants, age ~23) is worth knowing: low-volume all-out sprint intervals raised aerobic capacity by ~3.6 mL/kg/min (~8%), a moderate-to-large effect against no-exercise controls and not different from endurance-training controls — efficient, but not a full replacement where the event demands volume (the base's inference, not a finding of the meta-analysis), and measured in healthy young people rather than trained athletes.
Choosing the format — who gets which
No single format wins for everyone. Keep two kinds of evidence apart: acute studies count time near VO2max in one session (a proxy — see Limitations); training studies measure change over weeks. In this base only 30/15 (elite and well-trained cyclists), 4×8 min (trained recreational cyclists) and sprint intervals (mostly non-athletes) have training evidence; hard-start, over-under, ladder and 15/15 have acute evidence only.
What each format has behind it
| Format | Acute: time near VO2max | Training outcome | Who was studied |
|---|---|---|---|
| 30/15 | Best tested (REF-046, B+, §2) | 9 sessions in 3 weeks of 3 × 13 × 30/15: 20-min power +4.7% vs −1.4% and peak aerobic power +3.7% vs −0.3% for effort-matched 4 × 5 min; VO2max change not different (REF-045, B, randomisation not stated). Five 30/15 sessions in 6 days, then 6 easy days: power at VO2max +3.7% vs +0.7% for regular training, VO2max not different (REF-048, B, non-randomised) | REF-046: 19 well-trained cyclists, 11 women (Tier 3); REF-045: 18 elite cyclists, VO2max 73 (Tier 4-5); REF-048: well-trained cyclists |
| 15/15 | Beat 30/30 at a lower cost (REF-257, B+, §2) | None | 16 male team-sport players, cycling; transfer is inference |
| 40/20 | No entry | No entry | Not from a study; the base's own practice: a 30/15 variant for a rider who finds 30/15 too easy |
| Even-paced long reps, 4-8 min | Beaten by the same work reshaped (§2) | 7 weeks at 2/wk: 4×8 min at 90% HRpeak +11.4% composite vs +5.5% for 4×4, +5.6% for 4×16, +4.2% for low-intensity only (REF-150, B+) | 35 trained recreational cyclists, VO2peak 52, 6 women (Tier 2) |
| Hard-start / over-under long reps | Beat speed-matched flat reps at the same HR, lactate and RPE (REF-700, B); the fast-start gain is small, "a refinement, not a new stimulus" in its entry (REF-194, B) | None | 10 and 11 well-trained XC skiers (Tier 3); transfer is inference |
| Decreasing ladder | Beat 30/20 and 3/2 (REF-130, B, §2) | None | 12 amateur cyclists (Tier 2) |
| SIT (30 s all-out, or 10 s all-out on ~4 min rest) | Not a time-near-VO2max format | Aerobic capacity ~+8%, no different from endurance training (REF-696, A-, 16 RCTs, 318 healthy young people). 10-s sprints on 4-min rest raised VO2max as much as 30-s sprints (+9.2% vs +9.3%); on 2-min rest they did not raise it (REF-050, B+, 48 participants, 2 weeks, training status not given). In XCO riders 6 weeks of HIIT (7-10 × 4-6 min) ≈ SIT (8-12 × 30 s): race time −5% vs −3% (trial summarised in REF-742, A, systematic review) | Mostly non-athletes; XCO riders in REF-742 |
Who gets which (session counts, rep schemes and hour bands in this table are the base's own practice unless a source is named for them)
| Athlete / situation | Default | Why, and how sure |
|---|---|---|
| Tier 1 — new to structured intervals | 1, then 2 short sessions a week: 30/15 or 1 min hard / 1 min easy, 10-20 min of intervals including recoveries | REF-749 (A, meta-analysis of 69 RCTs in previously inactive adults, Tier 0-1): 80% of the VO2max gain at ~11 min a session including recoveries (~5 min of hard work; ~15 min a week), flat beyond ~23 min; applying it to athletes is inference. REF-243 (B+, 82 inactive adults): self-paced 30-s sprints and 1-min intervals gained equally; higher power in the 1-min format gained more, a higher HR did not |
| Tier 2 — trained recreational, ~5-8 h | In a VO2max block, 2 a week, alternating 4×8 min at the hardest pace held evenly for all four with 30/15 | 4×8 is the only format with a training trial at this level (REF-150, B+, Tier 2 cyclists); 30/15's advantage is acute and measured one level up (REF-046, Tier 3). The alternation is house practice |
| Trained, 6-12 h (Tier 2-3) | 30/15 as the anchor — 3 × 13 × 30/15 with 3 min between series, the REF-045 protocol (the ~120% FTP target is house practice) — plus one shaped long-rep session (fast-start or over-under 5×5, or the ladder); 2 a week in a block | REF-045 tested elites; its entry calls the extension below elite an inference. The shaped long reps rest on acute data only (REF-700, REF-194, REF-130) |
| Elite (Tier 4-5) | 30/15 short intervals first; a loading microcycle (5 HIT sessions in 6-7 days, then easy days or 1 a week) a few times a season | REF-045 (B) is the only elite training trial. REF-048 (B, well-trained, non-randomised) and REF-136 (B, n=15 trained cyclists, not elite: 5 HIT sessions in one week then 1 a week for 3 weeks, VO2max +8.8% vs +3.7% for 2 a week spread evenly) support the block; "a few times a season" is house practice. REF-468 (B+, narrative): chasing the biggest acute response has no demonstrated link to adaptation — dose it so it can be repeated |
| Runners (cyclists: the rows above — nearly all format evidence here is cycling or skiing) | 1 session a week at vVO2max speed (the other quality day at threshold, SYNTH-47): short reps such as 30 s fast / 30 s jog, or 2-5-min reps | No entry compares formats in running. REF-005 (B, 8 highly trained runners, uncontrolled): 1 a week raised vVO2max; 3 a week in the next four weeks added nothing. REF-090 (B, correlational, 85 elite male runners): short intervals correlated with performance (r ≥ 0.53), long intervals barely (r ≥ 0.22). Rep lengths are house practice |
| Time-crunched (<5 h) | 2 short sessions a week: one 30/15 set (~20 min including recoveries), and either SIT (4-6 × 30 s all-out, or 10 s on ~4 min rest) or a second 30/15 | REF-749 (A, inactive adults): the curve flattens past ~23 min; that a short session is therefore not a compromise for a trained time-crunched athlete is inference — the entry says not to transfer the number to trained athletes. REF-696 (A-, healthy young people): SIT matched endurance training on aerobic capacity — efficient, not a replacement where the event needs volume (inference). Budget from SYNTH-26, house practice |
| Masters | Same formats and intensity; short-short as the default; 48-72 h between hard days | No masters trial of VO2max formats in the base. SYNTH-35 (C+): keep intensity, widen gaps on recovery signals; 48-72 h is its house default. REF-595 (B, former elite male runners): ~6% VO2max loss per decade with continued hard training vs ~15% after stopping (amount of training graded, not intensity). REF-118 (B+, 22 veteran cyclists, 47 y): 12 weeks of 5 × 6 min at 40 rpm and 73-82% HRmax changed nothing — moderate low-cadence reps are not a substitute. Short-short preference is inference from REF-257's lower fatigue cost (16 male team-sport players) |
| Women | Same default | REF-046: no sex difference in time ≥90% VO2max for 30/15 (11 women, 8 men, well-trained) |
| XCO / cyclocross | VO2max intervals and SIT both, not one | REF-742 (A): in XCO riders HIIT ≈ SIT; burst formats in SYNTH-46 |
Frequency, block length, progression
Sessions a week: 2 in a VO2max block — tested as 2 a week for 7 weeks in trained recreational cyclists (REF-150) and 3 a week for 3 weeks in elites (REF-045); 3 a week for four weeks added nothing over 1 in eight runners (REF-005). 1 a week to maintain between blocks was tested only as the maintenance weeks of REF-136 (n=15).
Block length: the tested windows are 3 weeks / 9 sessions (REF-045), 7 weeks (REF-150) and 9 weeks / 21 sessions (REF-041); a narrative review puts 6-8 sessions over 2-4 weeks at a 2-4% gain in intense-exercise performance for well-trained athletes (REF-035, B+). Not from a study; the base's own practice: 3-6 weeks (6-12 sessions), a recovery week, a retest, then a change of emphasis.
Progression: Not from a study; the base's own practice: add work before intensity — more series or reps (2 → 3 series of 30/15; 4×5 → 5×5 → 4×8), and raise the target only when the last rep holds it. Reasoned from REF-150 (32 min at 90% HRpeak beat 16 min at ~95% in Tier 2 cyclists) and, for beginners, capped by REF-749's plateau at ~23 min a session. A session that visibly compromises the next days is over-dosed (house practice carried by REF-468).
How to tell it is working — by power, not heart rate (SYNTH-16, B)
In the session: time at or above target power and whether the last rep still holds it (SYNTH-16, power dose and subjective tiers); not kJ above CP/FTP (REF-130). Heart rate only as a within-athlete trend on the same format on a similar day (SYNTH-16) — it did not separate fast-start from flat reps (REF-194, skiers) and did not predict the VO2peak gain in REF-243 (inactive adults; exploratory tertile analysis), where power did in the 1-min format.
After the block: the markers that moved in the trials were power at VO2max or peak aerobic power and 20-min power (REF-045, REF-048) and a VO2peak-plus-power composite (REF-150); VO2max itself did not separate the formats in REF-045 or REF-048. So retest ramp/MAP and 20-min power, not only an estimated VO2max (adding a 5-min power test is house practice). Not from a study; the base's own practice: retest after the easy days (REF-048 tested after 6).
If it did not work: one flat block is not a verdict — in REF-498 (B+, 42 untrained middle-aged adults, the same 8 weeks done twice) the individual VO2max response barely reproduced (ICC 0.19). Check recovery, nutrition and whether the intensity actually landed (REF-431, B+, narrative) before changing the format.
Evidence gaps
Below the elite level no training trial compares 30/15 with long reps; REF-045 is 18 elite cyclists over 3 weeks. Candidate: Rønnestad et al. 2015, Scand J Med Sci Sports (10 weeks, short vs effort-matched long intervals, well-trained cyclists).
No running trial compares short-short with long reps. Candidate: Helgerud et al. 2007, Med Sci Sports Exerc (4×4 min vs 15/15 vs continuous training).
No masters- or women-specific training trial of VO2max formats.
Practical Application
Prescribe the interior, not just the envelope. Say what happens inside the rep — hard start, over/under alternation, or short-short — and why.
Reuse the catalog rather than composing a flat block from scratch. Blocks already implements these shapes ("Fast-start 4×5", "VO2 ladder", "30/15", "3×9").
Default to 30/15 for a first VO2max block (REF-046, REF-257): highest time in zone in the cycling comparison, no fibre-type or sex caveat; the lower fatigue cost was shown for 15-s versus 30-s bouts in team-sport players, not for 30/15 (inference). That default is for a trained cyclist; beginners, elite riders, runners, the time-crunched and masters are in "Choosing the format — who gets which" above.
Outside a block, cap at 2 sessions a week and rotate the format (house practice). REF-005 shows 3/wk added nothing over 1/wk in eight runners; 2/wk is coaching convention, not a finding.
Cue a moderate cadence (house practice; REF-255 warns against high cadence at high intensity, nothing more precise).
A flat rep is a legitimate CHOICE — state the reason. First VO2max session for someone new to the format; a rider who cannot hold structure outdoors; a session deliberately kept simple. That is honest coaching; defaulting to flat without noticing is not.
No power meter: shape by RPE and by feel — "start the rep clearly too hard, then settle" is executable without watts — and say plainly that HR alone is a rough, less repeatable read of the stimulus (REF-041).
What NOT to do
Do not use work above CP as the effectiveness metric (REF-130's entry warns against it explicitly; the dissociation is from its full text).
Do not claim moderate-intensity work is wasted — it is not (REF-421).
Do not prescribe 30/15 outdoors to someone whose roads make 30-second precision impossible; the format that gets executed beats the format that reads best (house practice).
Limitations
The format comparisons measure time ≥90% VO2max or mean %VO2max — an acute proxy, not adaptation. Only REF-041 links the fraction sustained to actual nine-week outcomes, and it is one single-arm prospective study. The chain "shaped rep → more time in zone → more adaptation" is well-motivated but only its first and last links are separately evidenced.
[REF-005] is eight runners without a control group, and REF-255 cannot support a cadence band; both are quoted here for less than they are usually quoted for.
Individual response varies substantially — none of these formats is guaranteed for a given athlete, and a block that underperforms is not automatically a programming error (the responder variability is read from REF-421's large SDs, e.g. PO15min +4.9 (8.7)%).
Nothing here compares these formats in RUNNING, where the same shapes are far harder to execute; the transfer is assumed, not shown.
The 30/15 numbers (118% / 60% of 40-min power) come from a specific protocol in well-trained cyclists; treat them as a starting point to be adjusted, not as a target to be defended.
How this entry was checked and changed
Why this level
(the principle rests on one single-arm study, REF-041 (B); the shapes on acute crossovers (B+/B); the training evidence on REF-150 (B+), REF-045 (B) and REF-005 (B); the A/A- reviews cited are in non-athletes or support no prescription)
Content correction 2026-09-27
"REF-005 (B): 3/wk produced no gain and overtraining markers" -> 3/wk in the following four weeks changed nothing and raised end-of-test noradrenaline; n=8 highly trained runners, uncontrolled sequential design; not overtraining syndrome.
"85-95 rpm preserves neuromuscular function (REF-255, A)" -> not in the review's abstract, which says high cadence at high intensity impairs more and the studies are too heterogeneous for a prescription; relabelled house practice.
"REF-421 (B+): VO2max-fraction principle HIT-specific; responders vary" -> abstract: both blocks improved PO15min (4.9% vs 2.8%, ns), MIT > HIT for PO4mmol, no VO2max difference; the HIT-specific regression and responder claims are the entry's exploratory analysis, labelled as such.
Study details added from corrected entries: REF-041 n=22, R² values; REF-046 n=19, 1123/879/917 s; REF-130 312/182/179 s; REF-150 percentages; REF-257 population; REF-696 16 RCTs, 318 healthy young, ~8%, not different from endurance controls. "Works universally" softened to "no phenotyping needed".
Residual sync (REF-696 re-checked): level parenthetical "(B+" -> "(A-", matching the regraded entry (meta-analysis of 16 mostly small RCTs); "explicitly not a full replacement" -> labelled as the base's inference (the corrected entry carries it as inference, not as the paper's finding).
Level parentheticals synced to the cited entries' current Level lines: REF-041 A -> B; REF-046 A -> B+; REF-130 A -> B; REF-194 B+ -> B; REF-150 A -> B+; REF-005 A -> B; REF-255 A -> A-; REF-421 A -> B+.
Level parenthetical "several A-level trials — the best-evidenced area in this base" -> two A- reviews plus B+/B trials.
Resync 2026-09-28: "REF-421's exploratory analysis: % VO2max and time ≥90% predicted the VO2max gain in HIT but not MIT — read as the entry's account" -> the corrected entry has deleted that analysis as not in the abstract; the sentence is replaced by the responder spread the entry does carry (PO15min +4.9 (8.7)%), and the Limitations note now points to those SDs.
Resync 2026-09-28: Tier "2-5 — the format evidence spans up to elite" -> 2-3: every cited trial is Tier 2 or 3 in its current entry (REF-700 was regraded from Tier 4-5 to 3; REF-257 is Tier 2), and REF-696 is Tier 0-1.
Revision 2026-09-29: added "### Choosing the format — who gets which": an evidence table per format (30/15, 15/15, 40/20, even long reps, hard-start/over-under, ladder, SIT) separating acute time-near-VO2max data from training outcomes, with level and population; a who-gets-which table (Tier 1, Tier 2, trained 6-12 h, elite, runners/cyclists, time-crunched, masters, women, XCO); sessions per week, block length, progression; and how to tell it is working (power over HR, linking SYNTH-16).Owner asked for 'the most effective VO2max workouts'; the answer is a default per athlete, not one protocol.
Revision 2026-09-29: new evidence brought in — REF-045 (elite 30/15 vs 4×5 training trial), REF-048, REF-136, REF-749 (A, dose-response in inactive adults), REF-243, REF-050, REF-742, REF-090, REF-595, REF-118, REF-035, REF-468, REF-498, REF-431; all doses, rep schemes and hour bands without a named source are labelled house practice.
Revision 2026-09-29: added "### Evidence gaps" (no sub-elite 30/15-vs-long training trial, no running format comparison, no 40/20 entry, acute-only evidence for shaped reps, no masters/women format trial, no interval meta-analysis in trained athletes) with candidate studies, none added.
Revision 2026-09-29: Practical Application default line scoped to 'a trained cyclist' and pointed to the new section; Cross-References extended. Level unchanged at B+ — the new rows rest on B/B+ trials and on A/A- meta-analyses in non-athletes, which do not raise the core.
Verification 2026-10-01: every number checked against the entries matches — REF-041 (R² 0.54/0.44/0.25, n=22), REF-046 (1123/879/917 s, 11 women + 8 men), REF-257, REF-130 (312/182/179 s), REF-700, REF-194 (87.6 vs 86.1, 98.2 vs 95.4, RPE 16.1 vs 16.5), REF-150 (11.4/5.5/5.6/4.2%), REF-005, REF-421, REF-696, REF-045, REF-048, REF-050, REF-742, REF-749, REF-243, REF-136, REF-090, REF-595, REF-118, REF-035, REF-498.Level/design/n parentheticals all match the head blocks. No NUMBER or WRONG finding.
Verification 2026-10-01: the problems are structural, not factual — the older half of the card (lead, Practical Application, What NOT to do) still speaks in rules ('cap at 2', 'never three', 'flat is not the default', 'shape matters more than length') that the newer 'Choosing the format' section, added 2026-09-29, contradicts with its own tested defaults (4×8 held evenly for Tier 2; 3/wk and 5-session blocks for trained/elite).
Verification 2026-10-01: the 'rotate the format' pair replaces the earlier unlabelled 'rotate formats' sentence in §3 by labelling the Practical Application line; the §3 sentence itself ('Treat the cap as coaching convention ... and rotate formats') should also be read as convention.
Verification 2026-10-01: Evidence gaps names four external candidate papers (Rønnestad 2015, Helgerud 2007, Milanović 2015, Rosenblat 2020) with no refId; they are presented as candidates, not as support, so not flagged.
Verification 2026-10-01: rows counted = format table (7) + who-gets-which (9). Evidence: 30/15, 15/15, even long reps, hard-start/over-under, ladder, SIT, Women, XCO. Practice: 40/20, Masters. Mixed (evidence in another population or acute only, plus house session counts): Tier 1, Tier 2, Trained 6-12 h, Elite, Runners, Time-crunched.
Verification 2026-10-01: pointers to SYNTH-16 (B), SYNTH-26 (2 quality sessions at 5-8 h, house practice) and SYNTH-35 (B; 48-72 h house default) match those cards.
2026-10-01: level tags next to cited frameworks updated to their current levels after the independent audit.
Revision 2026-10-04: added Olympiatoppen's Intensity Scale v2 (2024) REF-1077 as practice lines labelled "(Olympiatoppen guideline, level C — expert practice, not a study)"; no default, decision row or level changed.
Verification 2026-10-04: the 6 × 6 min model is labelled 'I-sone 4-5' in the national-team table; the guideline says longer rest 'in some cases, for example prior to competitions, it may be beneficial', not as a rule. Other formats, %FTP values and rest ratios checked OK against olt.txt and okt_sykkel.png. No default, decision row or level changed.