Threshold work

Whether a given athlete's quality session should be threshold work at all, and if so in what shape — continuous reps, over-unders, or the lactate-guided double-threshold day — and against which boundary the target is set.

Use CP (critical power), not FTP, as the heavy/severe boundary when prescribing intervals (REF-024, B+, foundational conceptual review — one source).

What the paper says: CP is the asymptote of the hyperbolic power–duration curve, it marks the boundary between the domain where physiology can stabilise and the one where it cannot, and intensity should be prescribed relative to CP rather than to metabolically distant markers such as lactate threshold or VO2max.

Not from a study; the base's own reasoning: FTP is a 60-minute construct usually estimated from a 20-minute test with a fudge factor, and REF-024 does not mention FTP at all — the FTP-vs-CP contrast is our application of its principle.

Prescribing "threshold" reps at 95–100% of an over-estimated FTP puts the athlete in the severe domain, where the session becomes a bad VO2max workout instead of a good threshold one — and the athlete reports it as "threshold feels impossible lately" (inference).

We already estimate CP and W′ from stored streams. Where CP is available, prescribe against it and say so.

REF-303 (A-, RCT, n=48 well-trained, 9 weeks) is unambiguous: a polarized distribution outperforms chronic threshold work on the key endurance variables — POL raised VO2peak +11.7%, ramp time to exhaustion +17.4% and peak power +5.1%, while the threshold-dominant group made no further gain in any endurance-performance variable apart from a slight improvement in work economy.

This does not make threshold work bad — it makes a threshold-dominant week bad. The failure mode this maps onto (our reading; the trial assigned well-trained athletes to programmes) is the classic amateur one: every session lands in the middle because the middle feels productive.

Combine with REF-150 (B+, RCT, n=35 trained recreational cyclists): adaptation depends on the interaction of intensity and total work duration, not on intensity alone — 4×8 min at 90% HRpeak (32 min of work, lactate ~9.6) gave an 11.4% composite gain, beating both 4×4 at 94% HRpeak (16 min, +5.5%) and 4×16 at 88% HRpeak (64 min, +5.6%), the latter two no better than low-intensity riding alone (+4.2%).

Two short threshold reps are not a threshold session; the accumulated time at the intensity is the dose — but note the 4×8 winner ran at lactate ~9.6 mmol/L, which is a hard session, not a threshold one, and the format nearest to threshold in that trial (4×16 at lactate ~4.9) did not beat easy riding.

No trial in this base crowns one threshold protocol. What follows is a default the assistant can prescribe, the switches away from it, and how sure each row is. Every row inherits this card's honesty: the threshold evidence here is thinner than the VO2max evidence.

Default (Tier 2-3 cyclist or runner, 5-12 h/week, no lactate meter): one threshold session a week, as one of the two quality days, done as 4 × 10 min progressing to 3 × 15 and 2 × 20 min just BELOW threshold (heavy domain), "hard but manageable", RPE 14-15.

What supports it: reps of 10-14 min at RPE 14-15 (lactate ~2.8 mmol/L) raised power at 4 mmol/L in well-trained cyclists, almost all men — but tested as a one-week block of six sessions, so a once-a-week spread, and transfer below well-trained, are inference (REF-137, B+, crossover n=30; REF-421, B+, crossover n=22).

"A few percent below CP rather than at it" is REF-750's application (B, n=7 men). Not from a study; the base's own practice: the rep ladder, ~2 min easy between reps, and one session a week.

The middle column (who, when, when to switch) is the base's own practice throughout; the right column is what the cited entries say.

The intensity anchor (this card's doctrine, unchanged). CP first — or critical speed for runners. CP is a boundary, not a duration: "indefinitely sustainable" was a mathematical term, never an hour (REF-751, B+), and MLSS typically sits somewhat below CP (REF-168, B+).

A critical review finds support for CP/CS at the heavy-severe boundary and little for fixed percentages, and notes that prescription methods have not been compared on adaptation (REF-173, B+). With only FTP20, treat it as a ceiling for long reps, not a target — a Tier 1-2 rider's 60-min power usually sits below it (REF-747).

Lactate only where it is actually measured (the 2-4.5 mmol/L band is REF-301's). Heart rate from a field threshold, never a formula, and as a cross-check rather than the target in short reps or heat (SYNTH-38, B). Never mix markers: CP and lactate-threshold markers disagreed by ±22 to ±52 W in individuals (REF-732, B+, n=17 recreational men).

How many a week. At 5-8 h the week carries two quality sessions, at 8-12 h two to three (SYNTH-26, B — the bands are base practice); threshold takes one of them, two only above ~8 h (base practice). Three to four threshold sessions a week belongs to the elite-volume model only (REF-301).

Progression. Time at intensity first, target second (base practice; REF-150 shows intensity and duration interact, not an order). Raise the target only after a retest (base practice). Read a fade inside a rep as a target or rep length that is too much (REF-750's application).

Across a 16-week build, pyramidal-then-polarized improved 5-km time most in 60 well-trained male runners, while all four distributions worked (REF-509, A-): more threshold early, more V̇O₂max work near the race — applying that to cyclists is inference.

Runners vs cyclists. The format trials here are cycling (REF-137, REF-421, REF-150, REF-042); runners anchor on critical speed or a field threshold pace, and running CP models carry more measurement noise (REF-024 limitations). The Norwegian model itself is running (REF-301).

In 85 world-class male runners, easy runs and tempo runs explained performance variance while long intervals barely correlated (REF-090, B, correlational, Tier 4-5 — not a prescription for amateurs).