SYNTH-53 · Guide
Testing thresholds
Choose the test by what the number will anchor, run it fresh and warmed up the same way every time, write the protocol next to the number, and between tests read the training data instead of retesting
The question it answers
Which test to run for this athlete, how to run it (warm-up, fresh legs, position, indoor/outdoor), which number to trust when two disagree, and how often to retest.
Default (what to prescribe first)
1
Cyclist with a power meter: estimate CP (and W′) from several self-paced maximal efforts of different durations, fitted to P = W′/t + CP. A narrative review recommends several maximal efforts under standardised field rules and advises against single-effort tests such as FTP (REF-026, B+).
In 12 competitive men, CP and W′ from 4-5 self-paced lab prediction bouts predicted a 16.1-km road TT to within a non-significant 0.8 min (r = 0.89) (REF-040, B). Not from a study; the base's own practice: two or three efforts of about 3 and 12 min (optionally 5 min), maximal, on separate days or with full recovery.
Self-paced efforts gave a CP about 6% higher than constant-load trials (265 vs 250 W) in 10 young men (REF-034, B) — so always test the same way.
2
Warm up identically every time, fresh, in the position you race in (§ Warm-up and conditions below).
3
4
Between tests, trust-then-verify from training data (§ How often below); a maximal test is a planned session at the end of a block, not a weekly habit.
Decision (who → what)
| Situation | Test and how to read it | Source and how sure |
|---|---|---|
| Cyclist with a power meter wanting zones and pacing, 2 sessions to spare | Self-paced 3 + 12 min (± 5 min) → CP and W′ | REF-026 B+ (method), REF-040 B (predicts a 16-km TT, 12 men); the durations are practice |
| One session, lab ergometer | 3-min all-out: all-out from the first second, no pacing; the last 30 s = end power ≈ CP (fixed resistance and last-30-s rule are full-text conditions in REF-044). Mode matters: in 12 participants end power matched CP only at constant cadence, not at fixed resistance, though it was repeatable in both (CV 1.2-1.9%), and work above end power never matched W′ (REF-877, B). End power tracked a 4-week CP rise (230 → 255 W) in 9 active people (REF-871, B). Inference: a trainer's slope or level mode is neither; treat it as its own number | REF-044 B: end power 287 vs CP 287 W, SE ~6 W, 8 of 10 within 5 W — 10 habitually active adults, lab; W′ less precise (r = 0.84). REF-120 B+: 15 W below end power 9 of 11 rode 30 min, 15 W above they failed at ~13 min (11 habitually active adults) |
| App or group convention is FTP | 20-min test × 0.95 = FTP20. Call it FTP20, never "an hour"; in a recreational rider it may read below LT | REF-747 B: median time to exhaustion at FTP20 35 min recreational → 51 min professional, n=87, wide spread. REF-040 B: CP and FTP equal on average (275 vs 278 W) but +30 to −36 W apart individually. REF-875 B: in 20 men FTP20 matched lactate threshold in trained riders but sat 6.5% below it in recreational ones (peak <4.5 W/kg) |
| Ramp test | Gives peak (ramp) power; track it against itself. No entry validates a ramp → FTP factor: across 100 adults MLSS sat at 44-71% of peak ramp power | REF-038 B+ (Tier 1, 46 women; ramp-slope-specific). The "~75% of ramp peak" convention is practice |
| Lab lactate test | Ask which method (e.g. modified Dmax) and stage length; run VO2max and threshold as separate tests. Measure with the same lactate meter every time and do not compare its values with another meter's or with a published mmol/L table | REF-031 B: many LT methods missed MLSS; modified Dmax from 4-min stages came closest; longer stages under-read VO2peak by 1-5 mL/kg/min (17 trained men). REF-029 B: MLSS needs ≥30-min stages, ≤1.0 mmol/L rise after minute 10 (26 men). REF-1081 B: against a Biosen reference, four handheld meters read −7%, 7%, −10% and 42% on average, all with poor precision (95% prediction interval 0.60-0.87 mM wide at 3 mM; venous samples, 0.88-4.89 mM) |
| Lab gas exchange (VT1/VT2, RCP) | VT1 anchors easy/moderate; RCP is not CP for an individual; test at habitual cadence | REF-171 B+ (know the variables used); REF-202 B: RCP 222 W at 60 rpm vs 197 W at 100 rpm, RCP and CP not correlated individually (10 collegiate men); REF-173 B+: some support for LT1/GET/VT at the moderate-heavy boundary, CP/CS for heavy-severe |
| No power meter, HR only | Threshold HR from a 20-30 min maximal steady effort, average of the last 20 min, cool and fresh — a convention, not a finding. Never an age formula | SYNTH-38 B; REF-3494 B+: every age formula RMSE 9.1-10.5 bpm (5069 athletes in the abstract, 5311 in the Results, corrected 2026-10-12; mostly men). Set HR zones per sport (REF-537, B+: HR differs between cycling and running) |
| Runner | Critical speed on the same logic (several maximal efforts of different durations); first threshold ≈ 82.3% of CS when untested | REF-173 B+ names CS as the heavy-severe anchor; REF-178 A: T1 = 82.3% of CS (95% CI 81.1-83.6), 527 runners, higher in faster groups. Field protocol: 3600, 2400 and 1200 m all-out on a track, 30 min apart, gave treadmill-equal CS (r = .89), repeat CV 0.4%, but D′ was neither valid nor repeatable (CV 13%) (REF-876, B, 10 men) |
| No lab, wants VT1 (easy ceiling) | Ramp with a beat-to-beat chest strap; HR where DFA a1 = 0.75 | REF-873 B: 15 people, treadmill, HR 154 vs 152 bpm at gas-exchange VT1 (ICC 0.96); group agreement only; running only — a cycling ramp is a transfer |
Which number to trust when two disagree. For zones meant to separate heavy from severe, CP (or CS) is the anchor the base supports (REF-173, B+; REF-024, B+).
Strong correlation is not interchangeability: CP and lab LT2 markers correlated at r = 0.81-0.98 but limits of agreement were ±22 to ±52 W, and the best-correlating marker (Dmax-exp) sat 20 W above CP (REF-732, B+, 17 recreational men). Do not switch the marker mid-season — the change will look like fitness (REF-732).
What no threshold number is. Not a duration. Trained riders held their own CP for 20-41 min with HR, VO2 and lactate still rising (REF-750, B, 7 men); the CP group itself says the model predicts nothing about duration at CP (REF-751, B+). Inference: prescribe "threshold" a few percent below the measured value when a steady state is intended (REF-750 practical note).
Warm-up and conditions
Warm-up by test length. Before a 7-min severe trial, a moderate or heavy (below-CP) priming bout ending 10 min before raised mean power ~2-3% (330 → 338/339 W); repeated all-out sprints did not (324 W) (REF-270, B, 12 well-trained cyclists). Inference: the same logic covers a 3-min or 5-min test.A 15-min progressive sub-threshold warm-up made a ~4.5-min 3-km TT 2.8% faster than none; a harder segment added nothing (REF-274, B, 8 well-trained cyclists). Three 10-s openers at 70% peak power ending 5 min before a 4-km TT gave a non-significant 1.7-s gain (REF-277, B, 10 men).Before a 20-min TT, neither a 10-min standard nor a sprint warm-up beat no warm-up, and the sprint version made riders start harder (REF-278, B+, 15 trained men). For short maximal efforts a long, hard warm-up cost power: Wingate peak 1303 W after a ≥50-min warm-up vs 1390 W after 15 min (REF-276, B, track cyclists).
Not from a study; the base's own practice: 15-20 min progressive, 5 min near but below threshold, easy 5-10 min, then start; the same script every retest. For a 20-min test the warm-up serves pacing and consistency more than power (REF-278).
Fresh legs. After muscle-damaging exercise, peak power in sprint and incremental tests fell while TT performance did not, and RPE at submaximal power rose (REF-701, A, meta-analysis). Not from the study; the base's own practice (REF-701's note): no ramp, FTP or profile test in the 24-48 h after heavy leg gym work, a long descent or a hilly race; test after an easy day.A 20-min TT after ~4 h / 40 kJ/kg of riding fell 2.9% on average (−8.5% to +1.1%) in 12 professionals (REF-196, B) — a rested test does not describe late-ride capacity (SYNTH-32).
Position. In 7 trained cyclists, 3-, 5- and 12-min efforts on TT bars gave a significantly lower CP than on the hoods, more so the longer the effort, W′ unchanged (REF-201, B; no watt values in the abstract). TT athletes test on the TT bars.
Indoor vs outdoor. In 20 competitive men, 20-min TT power was 280 W indoors vs 284 W outdoors (NS, r = 0.94), with outdoor within-ride variability about double (REF-253, B+). A 20-min number transfers; keep the same venue across retests anyway (practice). Treadmill runners: 1% grade matched road cost from ~13.5 km/h; below ~12 km/h 0% and 1% were indistinguishable (REF-733, B, 9 men, 1996).
How often / trust-then-verify
Retest interval is practice, not evidence. Not from a study; the base's own practice: a maximal test at the start and end of a block — roughly every 8-12 weeks, the practice note carried in REF-026, not its finding — or when training data say the anchor has moved.
Read the anchor from training between tests. The two best field powers per duration, taken from ordinary training and racing after 0-40 kJ/kg of prior work, were repeatable in 18 professionals (SEM < 5%, ICC > 0.90; within one context SEM < 8%) (REF-479, B, Tier 4-5) — an amateur transfer is inference.Inference: when a planned maximal effort in training beats the model (e.g. a 12-min rep above the predicted power), update; when easy-ride HR at a fixed power rises over weeks, suspect fatigue, heat or the anchor, not fitness loss alone (SYNTH-32 for decoupling on comparable rides).
Submaximal checks between maximal tests. Power output at a clamped HR had typical error 1.3-4.4% and related to 40-km TT and peak power at r = 0.91-0.94 (REF-752, B, 17 men); read with RPE and HR recovery as a set (REF-758, B+; SYNTH-41). After an ultramarathon, lower HR and faster recovery came with higher RPE — fatigue that looked like fitness (REF-757, B-, 14 runners).
Noise before change. Baseline 15-min power reproduced near-perfectly (ICC 0.96) in untrained adults (REF-498, B+), but individual training responses did not — one retest is one data point. Treat a change within the test's error (practice: a few percent for a maximal field test) as no change.
When NOT to use it / red flags
Illness, fever, chest symptoms or unexplained palpitations: no maximal test (see the base's illness rules).
Do not compare lactate values from different handheld analyzers — keep the same meter over time (the brand offset is measured in REF-030 and REF-1081) — and do not read a low value after a low-glycogen day as a low intensity; above I-3 (roughly above ~87% HRmax) lactate may not consistently reflect intensity unless the athlete has repeated measurements on the same session model (Olympiatoppen guideline, level C — expert practice, not a study) (REF-1077, C).
Do not tell a recreational rider to pace an hour at FTP20 (REF-747).
What the evidence does and does not show
Shows: the 3-min end power ≈ CP in habitually active adults in two fixed-resistance lab studies (REF-044, REF-120) but only in constant-cadence mode in a third (REF-877); self-paced CP predicts a 16-km TT in competitive men; protocol, position, pacing mode and LT method each move the number; threshold markers disagree individually; FTP20 is not an hour; priming helped a 7-min effort but not a 20-min TT (sources in the table and sections above).
Does not show: any ramp → FTP multiplier; a validated LTHR or 5-km-TT protocol for runners (field CS has one small validation, REF-876); test-retest reliability of FTP20 or the ramp in amateurs; an optimal retest interval; the 3-min test in trained or elite riders on the road or a smart trainer. Most samples are small and male.
Evidence gaps
Ramp-test MAP → FTP/CP conversion in trained amateurs — not in the base (REF-875 compares FTP20 with LT, not a ramp).
Running field tests beyond one CS validation in 10 men (REF-876): 5-km/3-km TT → threshold pace, 30-min LTHR — no entry.
Test-retest reliability of FTP20 in trained cyclists; the 3-min test's repeatability (REF-877) is from 12 lab participants of unstated training status.
Evidence for a retest interval, and for estimating CP from training data (without a planned test) in amateurs — none.
Women: almost every validation study here is male.
How this entry was checked and changed
Why this level
(test-validity and agreement studies, most of 7-26 mostly male cyclists, B/B+; warm-up crossovers, B/B+; no entry validates a ramp → FTP factor or a retest interval — those are labelled practice; one small study validates a field critical-speed test)
Content correction 2026-09-27
Revision 2026-09-29: folded in REF-877 (Wright 2017, B: 3-min end power = CP only in constant-cadence mode, WEP ≠ W′) and REF-871 (Vanhatalo 2008, B: end power tracked a training rise in CP) into the 3-min row — the row now says the ergometer mode decides whether end power equals CP, which qualifies the earlier fixed-resistance wording; REF-875 (Valenzuela 2018, B: FTP20 ≈ LT in trained, 6.5% below in recreational riders) into the FTP row; REF-876 (Galbraith 2014, B) replaces 'no validated running field protocol' with the single-visit track CS test; REF-873 (Rogers 2021, B) as a new DFA a1 → VT1 row.Gaps narrowed. Default unchanged; level unchanged (B).
Verification 2026-10-01: all 42 cited REF entries and the pointed-to lines of SYNTH-38/41/42 read in the KB.md snapshot. Every number in the Default, the 9 Decision rows, the warm-up section and the retest section matches its entry (REF-040, 034, 044, 120, 877, 871, 747, 875, 038, 031, 029, 202, 178, 876, 873, 732, 750, 270, 274, 277, 278, 276, 300, 285, 196, 253, 733, 752, 498, 3494); levels, n and designs next to refIds match the head blocks.
Verification 2026-10-01: practice lines (effort durations, warm-up script, retest interval, LTHR protocol, ~75% ramp convention, 'a few percent' noise band) are all labelled in the same sentence or cell — no UNLABELLED finding. Level B is fair: 8 of 13 Default/Decision rows rest on cited evidence.
Verification 2026-10-01: the one substantive defect is the lead and the 'Shows' line still presenting the 3-min test as giving CP and W′ after REF-877 was folded into the row saying it gives neither W′ nor, at fixed resistance, CP.
Verification 2026-10-01: SYNTH-12 also appears in the Cross-references list (out of audit scope); it should be removed there too.
2026-10-02: aligned with the rewritten SYNTH-42 (missing stimulus, one lever per block, level C+): the pointer for the "~75% of ramp peak" note is dropped — SYNTH-42 no longer carries that note; the convention stays labelled practice.
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.
Revision 2026-10-04: Lab lactate test row — added 'same lactate meter every time, do not compare across meters or with a published mmol/L table' with Mentzoni 2024 (REF-1081, B, device study: four handheld meters −7% / 7% / −10% / 42% vs Biosen, poor precision, 95% PI 0.60-0.87 mM wide at 3 mM); cross-reference added. No default or level changed.
Verification 2026-10-04: the FOLD-HRMAX edit did change a decision row (Lab lactate test: added 'same lactate meter every time, do not compare with another meter or a published mmol/L table'); the addition is supported by REF-1081 (B; −7%/7%/−10%/42% vs Biosen, 95% PI 0.60-0.87 mM wide at 3 mM, venous, 0.88-4.89 mM — all checked OK) and the revision note claims only 'no default or level changed', which is accurate. Level unchanged.