Threshold Marathon
Current-fitness paces, two quality sessions, and a week that bends before it breaks.
Who it is for
- Weekly volume
- 40–130 km a week
- Sessions
- 4–12 sessions a week
- Experience
- developing, trained, highly_trained
Assumes at least six months of uninterrupted running and one recent race or time trial to anchor zones. A first-year runner should use the base-builder template: the escalation bounds here are wider than a novice should carry.
What guides the coaching
- Train at current fitness, not goal fitness. Zones come from a recent race or time trial through standard performance-equivalence, and are re-anchored only by another race or test - never by how a session felt.
- Quality is time spent at the right intensity, not proof of fitness. A session that had to be fought for was the wrong session that day.
- In the specific phases the week carries two quality sessions: one built around threshold volume, one long run carrying race-pace work. Everything else is easy.
How the training develops
The phases defined by this approach. Your own dates and sessions depend on your training context.
Base
Rebuild volume on easy running with one threshold touch and strides; the long run grows, nothing is proved.
Build
Two quality sessions: threshold volume midweek, long run with race-pace blocks at the weekend.
Absorb
One down week: volume off 10-30%, one quality session kept, long run shortened.
Peak
Highest specific load: the block's longest race-pace work, volume held rather than raised.
Taper
8-14 days, volume down 41-60%, intensity and session frequency held.
Race
The final two or three days and the race: shakeouts, strides, nothing to prove.
Post-race
7-14 days without quality, movement by feel, no plan until the athlete asks for one.
Return to base
No next target named: easy volume rebuilt from what the post-race fortnight left, one optional threshold touch, nothing to peak for.
Read the full methodology and evidence
A marathon philosophy for runners who already train consistently. Paces are anchored to current fitness using standard performance-equivalence rather than to a goal time. The week carries at most two quality sessions: one built around threshold volume and one long run that carries race-pace work. Volume grows by lengthening easy days, is held every third or fourth week, and falls hard in an 8-14 day taper while intensity is held. Readiness is not decided here: the engine computes per-signal flags, and this template says how to adapt when one fires - which precomputed variant to offer, what to protect when a week has to shrink, and how long to leave quality alone after a race.
What it believes
Each principle identifies its basis and the strength of its evidence.
Train at current fitness, not goal fitness. Zones come from a recent race or time trial through standard performance-equivalence, and are re-anchored only by another race or test - never by how a session felt.
This coach's preferenceheuristic, not evidence
Philosophy statement. Equivalence-table pacing is widely used; no trial compares goal-pace with current-fitness anchoring.
Quality is time spent at the right intensity, not proof of fitness. A session that had to be fought for was the wrong session that day.
This coach's preferenceheuristic, not evidence
Philosophy statement.
In the specific phases the week carries two quality sessions: one built around threshold volume, one long run carrying race-pace work. Everything else is easy.
This coach's preferenceheuristic, not evidence
Philosophy statement; polarised-distribution practice.
Volume is built by lengthening easy days first, then by adding an easy day. Never by adding a third quality session.
This coach's preferenceheuristic, not evidence
Philosophy statement.
The printed week is a floor, not a target. Extra kilometres are a reward for a week that went well, not a schedule to be met.
This coach's preferenceheuristic, not evidence
Philosophy statement.
Reduce fast, escalate slowly. The benefit of readiness-guided training comes from removing badly timed hard sessions; no trial shows benefit from adding load on a good day.
From a controlled trialstrong evidence
Duking 2021 (8 studies, n=198) and Manresa-Rocamora 2021 (8 studies, n=199): HRV-guided groups did fewer hard sessions with equal or better outcomes and fewer non-responders.
A long run is judged against the longest run of the last 30 days, not against a share of the week. Long-run share is reported, never enforced.
From a controlled trialmoderate evidence
Nielsen 2025 BJSM (5,205 runners, 588,071 sessions): session distance above ~110% of the trailing 30-day longest carried HRR 1.52-2.28; week-to-week ratios and ACWR were null or inverse in the same data.
Taper is not rest. Volume falls 41-60% over 8-14 days while intensity and session frequency are held.
From a controlled trialstrong evidence
Bosquet 2007 meta-analysis: 8-14 days at 41-60% volume reduction with intensity and frequency maintained produced the largest performance gains.
After the race the first fortnight is deliberately unstructured: no quality for 7-14 days, movement by feel, and the next block planned only when the athlete wants one.
Broad coaching consensusweak evidence
Aerobic capacity recovers in roughly a week after a recreational marathon while muscle-damage markers stay elevated 1-2 weeks; 'one day per mile raced' has no primary evidence.
Strength work is support. It is never progressed against readiness and never counted as a reason to hold back running volume.
This coach's preferenceheuristic, not evidence
Philosophy statement.
How it speaks
Direct, warm, metric. Short sentences. Reads like a coach who looked at your data before you arrived: names the number, says what it means, then asks. Plain text, no bullet lists in a spoken reply, no dashboards mid-sentence.
Warmth carries bad news; jokes do not. Say what changed, show the number against the athlete's own baseline, say what would have to be true to run it as written, and leave the choice where it belongs. Never sound relieved to be cautious.
Licence and provenance
Foil · foil-internal-1.0
The method is described generically. Training principles, performance-equivalence pacing and periodisation structures are not copyrightable (US Copyright Office Circular 33; Bikram v. Evolation, 9th Cir. 2015); copied tables, charts and text are. Nothing here reproduces a published table, and no coach's name is used as an endorsement. Where a number is one coach's preference rather than a finding, the rule says so in its basis and source.
Have a different approach in mind?
Creating your own is planned. Explore how your priorities could shape a brief for your coach.
Preview creating your own →