SYNTH-13 · Guide
Athlete "types" are hypotheses, not diagnoses
The six-archetype model was this base's own speculative composite, validated by no entry, and is retired; what the evidence supports is a wide, partly inherited and poorly reproducible response to training, so profile only what an entry measures and let repeated blocks decide
The question it answers
"What type of athlete am I — sprinter, diesel, high responder, non-responder, recovery-sensitive — and should I train differently because of it?" It decides when a profile may steer training, what to do when a block did not work, and what to say about a genetic test.
Default (what to prescribe first)
Do not assign a type. Start from the goal and the stimulus it needs that recent weeks lacked (SYNTH-42, C+), describe the athlete with measured numbers rather than a label, and treat any pattern ("responds to volume", "needs more recovery") as a hypothesis for the next block — not from a study; the base's own practice, built on the finding that an individual's response to a repeated identical programme reproduced poorly while baseline values reproduced almost perfectly (REF-498, B+, 42 untrained middle-aged adults: ICC 0.19 for the VO2max response, 0.98 for baseline VO2max; transfer to trained athletes is untested).
Decision (situation → what to do)
The middle column is the base's own practice unless a refId sits in it; the right column is what the cited entries say.
| Situation | What to do | Source and how sure |
|---|---|---|
| "What type am I?" from a power curve, a metabolic test or an app's category | Give no type. Report the measurements against the goal's demands and choose the lever with SYNTH-42 | Power-profile classes and coaching "responder classes" are a practitioner heuristic with no peer-reviewed source on file (REF-436, C); the VLamax model figures are unverified practitioner numbers (REF-435, C+); "response clusters" from training data are a hypothesis with no source (REF-445, C) |
| Middle-distance runner (800-1500 m): speed type or endurance type? | Profile it — this is the one typing an entry supports — and follow SYNTH-60 for what the week then holds | In 19 elite men the speed reserve ratio (maximal sprint speed ÷ maximal aerobic speed) split 400-800 m, 800 m and 800-1500 m types, and sprint speed related to 800 m time at the same aerobic speed, r = −.74 (REF-888, B; bands unvalidated outside elite men — a direction, not a cut-off). Elite coaches place runners on a continuum from 400-800 m to 1500-5000 m types (REF-768, B+, narrative review) |
| "What should I train next?" | The stimulus the goal needs and recent weeks lacked, one lever per block, an outcome named in advance — SYNTH-42 | Between athletes VO2max explained 65-76% of the differences in threshold power or speed, economy 20-24%, fractional utilisation 4-11% (REF-746, B+, n=888, cross-sectional — between athletes, not within one) |
| A block was done and the marker did not move | In order: was it really done; is the change inside test noise; sleep, stress, fuelling; a second outcome besides VO2max; then repeat the block or change the stimulus. No label | Random within-subject variation can produce apparent responders and non-responders on its own (REF-430, B+, methodological). Measurement error cannot be averaged away in one person, and classification needs repeated tests and ideally repeated blocks (REF-441, B+, statistical perspective). The review of low responders lists the prescription, recovery (training status, sleep, stress, habitual activity), nutrition and measurement error (REF-431, B+, narrative). The leading resistance-training groups list similar first checks for the gym — adherence, whether overload progressed, and where possible sleep and diet — before intrinsic factors (REF-1143, B-, conference summary, muscle growth in mostly non-athletes) |
| "I am a non-responder" | Say that one block cannot show it. Repeat the same block, or change the stimulus — both are legitimate | REF-498: the same person could respond weakly in one 8-week block and well in the next identical one. Raising intensity at the same energy expenditure took VO2max responders from 16 of 31 to 13 of 15 in untrained adults (REF-995, B+, RCT; transfer to trained athletes is inference) |
| "I improve faster than my training partners" | Enjoy it; do not build the plan on it | In 481 sedentary adults on the same 20-week programme the VO2max gain averaged ~400 mL/min, ranged from little or none to more than 1.0 L/min, and clustered in families (maximal heritability estimate 47%) (REF-429, B+, single-arm, sedentary — it shows the spread, not shares of responder classes) |
| VO2max is flat after a block | Check threshold power, efficiency or a submaximal marker before calling the block a failure | In trained cyclists training raised gross efficiency but not VO2max, and within riders the two changed inversely (REF-444, B+, n=14 and n=29); splitting athletes into "VO2max responders" and "efficiency responders" is an interpretation the paper does not report |
| A consumer genetic test (ACTN3, ACE and similar) | Do not change training on it | The ACTN3 RR genotype is modestly enriched in power athletes (OR 1.48 vs non-athletes; REF-433, A-) and ACE II in elite endurance athletes (OR 1.54 vs inactive controls; REF-434, A) — population enrichment, not individual prediction. Genetic testing cannot yet identify talent on its own (REF-440, B+, narrative review). For muscle growth, 2 of 500 polymorphisms tested in 1,300 young adults and 1 of over 315,000 tested in 119 young men related to the response, and the leading groups advise no undue weight on genetic tests (REF-1143, B-, second-hand) |
| "I recover slowly" / large day-to-day HRV swings | Not a type. Time the hard days from readiness — SYNTH-30 | HRV-timed hard days gave a similar or slightly better 3000-m result with fewer hard sessions (13.2 vs 17.7) in 40 recreational runners (REF-438, B+), and no significant between-group difference in 17 well-trained cyclists (REF-362, B). Neither trial classified individuals |
| "I fade after three hours" | A durability and fuelling question, not a personality — SYNTH-32 | SYNTH-32, B |
How to progress / periodise
Keep the same benchmark across blocks and judge the pattern, not one change — the REF-441 entry's practical compromise. Which test, and how to run it the same way each time, is SYNTH-53; a submaximal output at a fixed internal load had a typical error of 1.3-4.4% in its reliable stages (REF-752, B, cyclists), which is the size of change a single retest cannot resolve (inference).
Log baseline values; they are stable. Baseline VO2max and 15-min power reproduced with ICC 0.98 and 0.96 (REF-498) — the baseline is individual, the size of the change is not.
A pattern becomes usable after it repeats. Not from a study; the base's own practice: call something "how this athlete responds" only after it has shown on the same benchmark in two or three blocks, and re-check it about twice a year.
Vary the stimulus across the year. Inference from REF-498: because the response to any one stimulus is unstable within a person, a year that rotates stimuli is safer than one locked to "their" stimulus.
When NOT to use it / red flags
Do not score an athlete with a questionnaire into archetypes. The 35-item questionnaire, its scoring weights and the seven-step classification tree were base constructs with no validation and are withdrawn.
Do not read the power-duration curve as a physiological type (REF-436, C), and do not compute "utilisation" from it (SYNTH-42).
Do not let a genotype overrule demonstrated performance or training response (the REF-433 entry's own caution).
A flat marker together with lasting fatigue, low mood or illness is not a type question: handle readiness with SYNTH-30 and send health questions to a doctor.
What the evidence does and does not show
Shows. Response to the same programme varies widely between people and runs in families (REF-429, sedentary adults). Much apparent "responder" variation can be measurement noise (REF-430), and one person's response does not reliably repeat (REF-498, REF-441).
Pre-training phenotype predicts only a small part of the response; resting autonomic activity is the baseline variable with relatively stronger predictive potential for the VO2max response (REF-431). In elite male 800-1500 m runners a speed-versus-endurance profile separates subgroups (REF-888).
Does not show. That endurance athletes in general fall into six (or any number of) types (the one supported split is the middle-distance speed profile above); that a type predicts the best volume or intensity distribution; that matching training to a profile beats a goal-based default — SYNTH-42 states that no entry compares block-selection rules head to head.
The response entries are sedentary or untrained adults (REF-429, REF-498); how stable response is in trained athletes is unknown. The practitioner systems the old model leaned on have no verified source here (REF-435, REF-436, REF-445, REF-432).
What was retired on 2026-10-01 and why. The six archetypes (trainability champion, glycolytic, diesel, durable, worker, recovery-sensitive), their power-profile mapping, genotype signatures, per-type volume and intensity prescriptions, the multi-archetype combinations, the classification-inputs table, the decision tree, the questionnaire and its scoring.
Each was the base's own construction; the old entry's own limitation read "no published study has validated the full 6-archetype model as a whole". A framework the assistant cites must not hand an athlete a label that no entry can back.
Evidence gaps
No trial of type-matched or profile-matched training against a goal-based default, in any sport.
No reproducibility study of training response in trained athletes (Tier 2+); REF-498 is untrained middle-aged adults. For resistance training a large repeat-validation trial in Jyväskylä has run 3 years and is due to end within 1-2 years, in adults of unstated training status (preliminary data published) (REF-1143).
No validated speed-versus-endurance profiling for cyclists; the running bands come from 19 elite men (REF-888).
No peer-reviewed validation in the base of power-profile typologies, VLamax-based typing or data-driven response clusters (REF-436, REF-435, REF-445).
No source in trained athletes for "higher intensity rescues non-responders": the verified trial is in untrained adults (REF-995); REF-432 stays an unverified composite.
No response-variability entry in trained women or in masters athletes.
How this entry was checked and changed
Why this level
(the retained rules rest on B+ methodological and observational entries — REF-430, REF-441, REF-498, REF-431, REF-429 — none of which is a trial of individualised prescription; the six-archetype model itself was a speculative composite of C/C+ practitioner entries that no entry validates, and is not carried forward)
Content correction 2026-09-27
Archetype 4 "high fat oxidation rate" attributed to Valenzuela/REF-196 -> REF-196's abstract measured no fat oxidation and found no association between the −2.9% power decay and VT, peak PO, VO₂max or training load; fat oxidation moved to labelled inference (REF-142 narrative).
Archetype 4 "not predicted by fresh VO₂max/FTP/efficiency (per REF-423-427)" -> holds for amateurs (REF-427) and pros after a 4-h submaximal ride (REF-196); in REF-424 (Spragg 2022, pros, 5 × 8 min above CP) ΔCP was strongly positively correlated with relative VO₂max, peak power, threshold power and gross efficiency (abstract-checked; the REF-424 entry text is pending correction)."High mitochondrial density" removed (REF-425 reports no enzyme-profile correlation). "2000 kJ" replaced by the protocols actually used (40 kJ/kg, REF-196; 1000 kJ, REF-427). CHO 90-100 g/h -> the ranges the cited entries carry (80-100, REF-142/143; 90-120, REF-141).
Archetype 1 ">20-25% VO₂max gain in 12 weeks" -> HERITAGE (REF-429) figures: 20 weeks, mean ~400 mL/min, range little/none to >1.0 L/min, 2.5× between-family variance, heritability 47%; the top-15% cut-off labelled house practice. "Hopker 2018" -> Hopker 2012 (REF-444), responder split marked interpretive. "80/20 polarized (REF-085)" -> REF-303 (duplicate of REF-085; POL split was ~68/6/26, 80/20 by sessions is REF-087's descriptive pattern).
Archetype 1 "2-3 HIT/wk in HIT-blocks" -> REF-136 tested 5 HIT in the loading week then 1/wk (+8.8 vs +3.7% VO₂max); house version labelled practice. All volume and session-frequency prescriptions labelled "Not from a study; the base's own practice".
Archetype 3 ACE II -> REF-434 reports the II association for runners, triathletes and swimmers, not cyclists; AMPD1 marked not actionable (REF-448). MIT-block given REF-137's numbers (+4.0 vs −1.3% PO4mmol, n=30).
Archetype 5 "~15-20% of athletes" -> ~10-15% (REF-436 ~15%, REF-431 ~15%, REF-445 ~10%); REF-429 gives no percentage. Added REF-432 to the rescue protocol and the REF-441/REF-430 measurement-noise check before labelling.
Archetype 6: added REF-438's actual result (3000-m +2.1% vs +1.1% ns, 13.2 vs 17.7 hard sessions, VO₂max improved in both groups) and REF-361's corrected reading (falling HRV in elites can be saturation or freshness).
Header "REF-360-365 (HRV monitoring)" -> REF-361-364; questionnaire cross-reference REF-365 (ACWR — Impellizzeri 2020 finds no evidence supporting ACWR, level C) replaced by REF-361/REF-363. Confidence column, questionnaire bands, ≥4-item dominance threshold and "~6 months re-test" labelled practice/inference.
Full-base sync (after the ~500-entry pass): Level parenthetical "phenotype and genetics primaries are A/B+" -> the practitioner/composite phenotype entries are now C/C+ (REF-435, REF-442, REF-443 C+; REF-436, REF-445 C); peer-reviewed primaries A-/B+. Level B kept (the archetypes were always labelled the base's construction).
Archetype 1/5 "~30% polarized-responder cluster" and "ML non-responder cluster ~10%" (REF-445) -> REF-445 now has no source and no cluster table; removed. "Mann's low/non-responders ~15% (REF-431)" -> REF-431 gives no percentages; the only share left is Couzens' unchecked ~15% (REF-436, C)."Often older (REF-431 modulators)" -> REF-431's factors are recovery (training status, sleep, stress, habitual activity), genetics, baseline phenotype, autonomic activity, programme, nutrition, measurement error — not age.
"Hecksteden found only 11/20 subjects consistently classified (REF-441)" and "≥3 baseline tests" -> not in the corrected entry (a statistical perspective with no cohort); replaced by its actual argument: n-of-1 measurement error, within-person variation in response, repeated tests and repeated blocks.
"REF-432: responder rate 53% vs 87%" -> REF-432 is a C+ composite with unverified sources and says its percentages are not quotable; numbers removed, claims marked as its working position.
Archetype 4 "long Z1 sessions of 3-5 h (REF-141)", "90-120 g/h (REF-141)", "80-100 g/h (REF-142/REF-143)", "REF-141: low-volume HIT reduces drift" -> none are in the corrected entries; replaced by REF-141's three abstract suggestions (years of consistent volume, prolonged sessions with race-pace or progressive intensity, heavy strength/plyometrics); CHO intake relabelled house practice without a number.
W' bands ">25 kJ" / "<15 kJ" (REF-442) -> removed from the entry as untraceable; now qualitative. REF-435 "B+" -> C+, VLamax figures marked unchecked INSCYD practitioner figures. REF-443 (PPARGC1A) and REF-448 (AMPD1) and REF-437 (DFA α1) marked C+ composites with unverified sources. REF-433 given its ORs (1.48 / 1.27; RX commonest).
REF-136 "same total HIT" -> "similar" (n=15). REF-438: between-group effect small (ES 0.42) added. REF-363 "rolling mean" -> only the CV is from the paper. REF-498: population (untrained middle-aged, n=42) added; "fibre type from biopsy" as a reproducible baseline -> not in the abstract (baseline ICCs are VO₂max 0.98, 15-min power 0.96, "most other outcomes").
Removed two "pending correction" remarks: REF-196 and REF-424 have since been corrected against their abstracts (REF-196 carries no fat-oxidation correlation; REF-424 reports strong positive correlations).
REF-425 is now sourced (Hamilton 2024, EJAP, DOI 10.1007/s00421-024-05459-6, attached with the owner's OK 2026-09-27): n=13 well-trained cyclists/triathletes; VT1 durability tracked the 5-min TT loss; citrate synthase, complex I and carnosine did not predict durability. References to it here updated.
Resync 2026-09-28: Archetype 3 "ACE II ... cyclists are not among the significant groups, so this marker is weak for cycling" -> REF-434 reviewed cycling but reports no cycling-specific estimate; the entry warns against calling it null for cyclists.
Resync 2026-09-28: Archetype 6 "gains more when HIT is timed to HRV (REF-438)" -> REF-438: similar or slightly better 3000-m result with ~25% fewer hard sessions, between-group ES 0.42, VO₂max up in both groups — "same fitness for less hard work, not more fitness"; the recovery-sensitive link is the base's hypothesis; "often older / masters" labelled inference.
Resync 2026-09-28: stale remarks about entry text removed — REF-423 no longer carries "~1 h near LT2" / "VO₂max/LT2 did not predict" (it says the predictor question was not tested; r = 0.91 GE-5-min link added); REF-428 no longer carries per-cent gains (added: LOW-Load beat control on Wingate mean power).
Resync 2026-09-28: Level parenthetical "phenotype and genetics primaries are A/B+" -> peer-reviewed primaries A to B (REF-434 A, REF-433 A-, most B+/B); practitioner/composite phenotype entries C+/C. Level B unchanged.
2026-10-01: rewritten in the decision-framework format and cut from six archetypes to the sourced core. Title and level now state that the archetype model was the base's own speculative composite validated by no entry (Level B -> B-).The default changed from "classify the athlete into 2-3 archetypes and tailor volume and intensity per archetype" to "do not assign a type; start from the goal and limiter (SYNTH-42), profile only what an entry measures (SYNTH-60 for 800-1500 m runners), and let repeated blocks decide".Removed: the six archetype blocks with their per-type hours and session counts, WKO5 mapping, genotype signatures, multi-archetype combinations, classification-inputs table, decision tree, 35-item questionnaire, scoring logic and "How to use".This resolves the conflict with SYNTH-42 ("do not label the athlete a type") and removes session counts and weekly hours that SYNTH-26 owns; HRV-guided training now points to SYNTH-30, durability to SYNTH-32, testing to SYNTH-53.Kept and re-read live: REF-429, REF-430, REF-441, REF-498, REF-431, REF-444, REF-438, REF-433, REF-434, REF-440; added REF-746, REF-888, REF-768, REF-752, REF-362.No longer cited: the durability, block-periodisation, torque, HRV-interpretation and unverified genotype entries the archetype blocks leaned on (among them REF-136, REF-137, REF-196, REF-423 to REF-428, REF-141 to REF-143, REF-361, REF-363, REF-437, REF-442, REF-443, REF-446 to REF-448).
Verification 2026-10-01: 20 cited REF entries read in full. Every number (ICC 0.19/0.98/0.96, ~400 mL/min, >1.0 L/min, 47%, r = −.74, 65-76/20-24/4-11%, OR 1.48, OR 1.54, 13.2 vs 17.7, n=17, 1.3-4.4%), population and level tag matches its entry. No STRETCH, NUMBER, WRONG or LEVEL finding.
Verification 2026-10-01: Level B- is honest — the middle column is declared practice by the table header, the right column is B+ methodological/observational evidence, and the Level line says none is a trial of individualised prescription. Transfer from untrained adults (REF-498, REF-429) is stated in the Default, Tier line and gaps.
Verification 2026-10-01: quotes attributed to SYNTH-42 check out ('No entry in this base compares "train the diagnosed limiter" against "train the default" head to head'; 'Do not compute "utilisation" from CP over the best 3- or 5-min power'). SYNTH-32 and SYNTH-42 are level B as tagged.
REF-498 has a PMID but no DOI line (the entry says so); not flagged as SOURCE. REF-888's speed typing rests on n=19 elite men with aerobic speed predicted from race times, which the row discloses. Row split counts the Default and 10 decision rows; row 1 and the SYNTH-32 pointer row are counted as practice.
2026-10-01: level tags next to cited frameworks updated to their current levels after the independent audit.
2026-10-02: aligned with the rewritten SYNTH-42 (missing stimulus, one lever per block, level C+): lead, Default and the "What should I train next?" row no longer say "the limiter of a goal" — SYNTH-42 now holds that a profile is not shown to name the limiter; level written next to it B- -> C+; the quote "train the diagnosed limiter vs train the default" replaced with SYNTH-42's current wording (no entry compares block-selection rules head to head).
2026-10-02: "higher intensity rescues non-response" repointed from the unverified composite REF-432 to REF-995 (Reuter 2023, RCT, 31 untrained adults, responders 16 of 31 -> 13 of 15 with energy-matched higher intensity); the evidence gap narrowed to trained athletes. REF-432 stays in the list of unverified entries.
Revision 2026-10-16: folded REF-1143 (Roberts 2026, B-, conference summary of resistance-training response heterogeneity, full text) into the flat-block row (similar first checks for the gym, muscle-growth evidence), the genetic-test row (2 of 500 and 1 of over 315,000 polymorphisms related to muscle growth; no undue weight on genetic tests) and the reproducibility gap (a resistance-training repeat-validation trial is running, training status unstated); cross-reference added.Defaults, decision rows and level unchanged.