The limiter and the block thesis
What to train next to get faster or raise FTP: a profile is not shown to name the limiter, so pick the stimulus the athlete has been missing and the goal needs, pull one lever per block, and let the measured outcome decide the next block
The question it answers
What the next 3-6 weeks (a house default length) are FOR (one primary lever), which outcome will show whether it worked, when it is read, and what the reading changes. It does not answer "which determinant limits me": no entry shows that a profile can tell.
Default (what to prescribe first)
Not from a study; the base's own practice, in this order: (1) pass the consistency gate; (2) read the horizon; (3) choose the stimulus the athlete has been MISSING in the last 8-12 weeks of their log and that the goal event needs; (4) make it the one primary lever of the block; (5) write the block thesis; (6) measure the outcome itself; (7) keep what worked. Name the rule that decided the block.
Starting bet (the base's own practice, not a diagnosis): when the log shows no obvious gap and the goal rests on aerobic power, the first lever worth trying for most amateurs is work above threshold — VO2max-type intervals shaped by SYNTH-33 — on top of the easy volume the week can hold (SYNTH-26). What it rests on:
Its caveats — say them with the bet:
Why a profile does not choose the lever
Decision (who → what)
| Row | Rule | What it rests on |
|---|---|---|
| Consistency gate | House default: fewer than ~6 of the last 8 weeks with the athlete's usual session count → the lever is regularity plus one quality session; say it with THEIR week counts | No entry tests the number. Nearest: completing >80% of planned training weeks went with a 7-times higher likelihood of reaching the performance goal in 33 international track and field athletes (REF-1001, B, observational, Tier 4-5 — transfer to amateur endurance athletes is inference); a complete stop of ≥ 7 days went with a 5-8% slower marathon than the same runner's undisrupted one in 300,000 recreational runners (REF-941, B+, observational); marathon cohorts with more weekly distance and more runs per week finished faster (REF-1004, A-, between-cohort associations); sedentary subjects improved oxygen uptake at the threshold at every training intensity (REF-1005, A-, 1997 meta-analysis; sedentary people, not inconsistent athletes — transfer is inference) |
| Horizon > ~16 weeks to the A-goal | House default: the missing stimulus is the block's lever | No entry tests the bands |
| Horizon ~8-16 weeks | House default: the lever stays primary, race-specific work grows through the block | No entry tests it. Order is not the science: in 63 trained cyclists the order of interval formats over 12 weeks made no difference (REF-096, A-) |
| Horizon < ~8 weeks | House default: specificity to the event's demands; no new lever | One thin trial: in 20 male cyclists across five groups, six sessions in 3 weeks, the largest modelled 40-km gains were for 8 × 4 min at 85% of peak power (2.8%), which the authors read as close to race pace, and for 12 × 30 s at 175% (2.4%) (REF-1002, B, modelled, no control) |
| Horizon < ~3 weeks | Taper (pointer) | SYNTH-34 |
| No event; A, B and C races | The base's own practice: with no event the horizon is open — say so and build the block anyway. The bands count to the A-race; B and C races do not restart the block | No entry tests it; ranking races: SYNTH-76 |
| One lever per block | The base's own practice: one primary lever, one maintained quality, everything else at maintenance dose | Untested. Concentrating intensity into blocks gave no time-trial advantage in a 12-week load-matched trial of 30 trained cyclists (REF-098, B+); that is not one lever against several |
| What the goal needs | The base's own practice, not a tested mapping: aerobic power and events under about an hour → the starting bet; longer than about two hours → durability and fuelling SYNTH-32; off-road and surging races → SYNTH-49; running economy via strength → SYNTH-44; triathlon → the discipline that costs the most minutes; fuelling SYNTH-02, pacing SYNTH-10, heat SYNTH-04, position and aero SYNTH-48 are levers too | No entry tests the mapping |
| Threshold work as the lever | Inference: a fair choice when it is the missing stimulus, not because a ratio looked low; format SYNTH-27 | Conditioned subjects showed non-significant threshold gains (ES 0.63) and tended to respond better to high intensity (REF-1005; "conditioned" is undefined there; the outcome is oxygen uptake at the threshold); one 7-day block of six moderate-intensity interval sessions raised power at 4 mmol/L by 4.0% against −1.3%, and VO2max by 2.0% against 0.0%, in 30 well-trained male cyclists, while 15-min power did not change significantly (p = 0.14) (REF-137, B+, crossover) |
| Threshold-to-ceiling ratio | The base's own reasoning: a note, not a decision axis. Read it only from a MEASURED maximal aerobic power, only against the athlete's own earlier value on the same protocol, never from the power-duration curve (CP over 5-min power reads W′) | No entry compares a field-ramp ratio with a laboratory one. Laboratory ratios averaged close to 80% at every level (REF-987) |
| The block thesis | The base's own practice: written before the block — goal · horizon · lever and why (missing from the log, needed by the goal) · the outcome and its baseline · the retest session · the review date. One sentence: "the next N weeks are for X because Y; we will know from Z on <date>" | Untested. Review after 3-6 weeks is a house default; by inference it fits a 40-km gain being measurable after 4 weeks of added intervals in highly trained cyclists (REF-999) |
| No baseline | The base's own practice: if the outcome the block will be judged on was never measured, the first step is the cheapest test that gives a baseline, scheduled — not a guess | Which test: SYNTH-53 |
| The outcome | The base's own practice: power or pace at the durations the goal needs — the same key session compared rep for rep, or a planned maximal retest at the end. Field bests from efforts that were not maximal do not count | No entry tests the rule; repeating a test: SYNTH-53 |
| In-block marker: `output_at_hr` | The base's own practice: output at a fixed heart rate on steady rides and flat easy runs, last four weeks against the four before, read with RPE; a change inside its noise band is no change | Power in heart-rate-clamped stages had a typical error of 1.3-4.4% in 17 well-trained male cyclists (REF-752, B, reliability of one standardised test); reading it across ordinary sessions, and for running pace, was not tested there. Lower heart rate with higher RPE is not fitness: SYNTH-41 |
How to progress: reading the block
Was it done? Not from a study; the base's own practice: if a large share of the planned key sessions was missed, the block says nothing about the lever — keep the lever and fix load, life or fuelling.
Done, and the outcome moved beyond its noise band: keep what worked — extend the lever while it pays, or hold it at maintenance and start again from the gate (the base's own practice).
Done and flat: two exits, both shown only in adults who were untrained or not described as trained — transfer to trained athletes is inference.
Raise the intensity: after 10 weeks of moderate training 16 of 31 untrained adults were VO2max responders; 16 more weeks at the same intensity added none, while energy-matched training at rising intensity brought the count to 13 of 15 (REF-995, B+, RCT).
Or add dose: non-response in maximal power was 69%, 40%, 29%, 0% and 0% with one to five 60-min sessions a week, and disappeared in every non-responder after six more weeks with two extra sessions (REF-996, B+, 78 healthy adults not described as trained; second period uncontrolled).
A flat response to one stimulus does not predict the next. In 21 recreationally active adults, individual changes after three weeks of endurance training were not positively related to changes after three weeks of sprint intervals (REF-997, B+, crossover; no mode was repeated, so a true difference cannot be separated from noise; transfer to trained athletes is inference).
In 37 recreational runners, baseline nocturnal HRV related to the gain in peak treadmill speed negatively with added volume (r = −0.74) and positively with added intensity (r = 0.63) (REF-1000, B, correlations in two small groups; nobody was assigned by the marker).
One block is one noisy point. When 42 untrained adults did the same 8-week programme twice, the individual VO2max response barely reproduced (ICC 0.19) (REF-498, B+; transfer to trained athletes is inference). Not from a study; the base's own practice: judge a lever over 2-3 blocks before calling the athlete a non-responder to it.
Day-to-day adjustment is a different question. HRV-guided training had a significant effect on submaximal physiological markers (g = 0.296) and small, non-significant effects on performance and VO2peak across 8 studies with 198 participants; most guided arms did fewer moderate- or high-intensity sessions, so the two are confounded (REF-994, A-). It tunes the load, not the lever: SYNTH-30.
When NOT to use it / red flags
What the evidence does not show
That choosing the missing stimulus beats any other rule — no entry compares block-selection rules head to head; that the horizon bands, the "6 of 8 weeks" gate, the written thesis or the 3-6 week review improve anything; that a VO2peak gain pays off in performance within one block (REF-989).
Evidence gaps
How this entry was checked and changed
the procedure (missing stimulus, horizon bands, one lever, thesis, review, consistency gate) is the base's own practice and no entry tests it; the findings under it are C+ to A- but indirect: group means, between-person data or untrained adults. Regraded 2026-10-01 from B-
output_at_hr marker are kept as labelled practice; "beyond about 2.5 h" aligned with SYNTH-32 (about two hours). Level B- -> C+.