SYNTH-64 · Guide

Preparing for a marathon (and a half)

Grow the easy layer and the long run, keep the hard work roughly constant, load without depleting, fuel early, drink to thirst and start at or below goal pace; loading dose, long-run length and pacing are practice, not evidence

B-limited, mixed evidenceMostly from recreational to highly trained athletesWritten before October 2026, being re-checked

The question it answers

A marathon or half is 12-20 weeks away: the build, long run, marathon pace, the wall, loading, race fuel and drinking, pacing, shoes, taper, recovery. Basics: SYNTH-47; hours SYNTH-26; gym SYNTH-44; taper SYNTH-34; grams per hour SYNTH-02; durability SYNTH-32.

Default (what to prescribe first)

For a healthy Tier 1-3 runner running ≥3 days a week:

Grow easy volume; hold the hard work roughly constant. Across 119,452 marathoners' 16-week Strava logs, the fastest (120-150 min) ran over three times the volume of slower runners (overall mean 45.1 km/week), the extra almost all below the first threshold, Z2/Z3 time similar across levels, and >80% of the fastest pyramidal (REF-197, B+; cross-sectional — faster runners may simply tolerate more).

A running review of 16 studies (6 interventions) found pyramidal and polarized more effective than threshold-heavy training, though some of the best marathoners use threshold training (REF-773, B+).

Volume by hours — not from a study; the base's own practice: first marathon, finish goal: 4-5 runs, peaking ~4-6 h/week; recreational with a time goal: 5-6 runs, ~5-8 h; trained: 6+ runs, 8+ h. World-class marathoners run 160-220 km/week over 11-14 sessions (REF-089, B+, Tier 5 — shape, not target); quality sessions per hour band: SYNTH-26.

Build 16-20 weeks from a consistent base (practice). Keep week-to-week increases under 20%: in 261 recreational runners building for a half marathon, 20-60% jumps brought about 23 percentage points more injuries at 21 days, none by 8 weeks (REF-623, B+); a 10% and a 24% weekly increase carried the same risk (REF-565, B+, very limited evidence). Strong evidence names a long weekly distance (men) and a prior injury as risk factors (REF-564, A-).

One long run a week; its length is the lever. In 26 well-trained men matched on 10 km time, regular ≥90-min runners lost less economy over 90 min at threshold (+3.1% vs +6.0%), and economy durability tracked the weekly longest run (r = −0.67) (REF-160, B, cross-sectional).

One 6 × 10-min moderate-intensity session (intervals, not a long run — a transfer) drifts (HR, lactate, RPE rise) and leaves more next-morning fatigue than the same work split in two (REF-159, B, 14 men, acute only). Practice: add ~10-15 min every 1-2 weeks up to 2.5-3 h or ~30-32 km, whichever is first; cut back every 3rd-4th week; longest ~3 weeks out.

Marathon-pace work grows toward the race. World-class runners' race-pace volume rises as the race nears (REF-089, B+). Practice: from ~10 weeks out, marathon-pace blocks late in the long run (20-40 min, growing to 60-90 min), rehearsing race fuel and shoes. Other quality day: threshold/critical speed (SYNTH-27); a vVO2max session early in the build (SYNTH-47).

Goal check: in 16 world-class men (the 7 who held a steady VO2 at that speed), 2-h marathon pace needed 94 ± 3% of VO2peak, above their lactate turn-point (92 ± 3%) (REF-822, B); inference in that entry: a recreational goal pace above the runner's threshold pace is unrealistic.

Strength twice a week while building, then maintain. Heavy loading (≥80% 1RM) improved economy by a small amount (ES −0.266), combined methods by a moderate amount (ES −0.426), plyometrics only at ≤12 km/h, submaximal and isometric work not significantly (isometric: three small calf-only studies, pooled estimate favouring training but CI crossing zero); the heavy effect grew with speed and VO2max (REF-801, A-, 652 runners).

Swapping 32% of training for explosive work improved 5 km time and economy in 18 well-trained endurance athletes (REF-802, B; 5 km, not marathon). Dose: SYNTH-44. Inference: a 4-5 h marathoner runs below 12 km/h, where plyometrics helped.

Decision

Situation	Do this	Source and how sure
First marathon, <4 runs/week (Tier 1)	Regularity and the long run are the lever; one quality session at most; volume jumps under 20%	REF-623 B+, REF-565 B+; session cap = practice, gate SYNTH-42
Recreational, 4-6 runs (Tier 1-2)	Default above: pyramidal, two quality days, long run carrying marathon pace late	REF-197 B+ (association), REF-773 B+; doses = practice
Trained, 16-week build (Tier 2-3)	More threshold early, more polarized and race-specific late	REF-509 A- (60 well-trained men, 5 km outcome — marathon transfer is inference); REF-089 B+
Hit the wall or faded before	Check first-half pace and grams actually eaten before calling it fitness	REF-775 B+ (pacing records only, no fuelling measured); REF-787 B+ (marathoners averaged 35 g/h); SYNTH-32; the check itself is practice
Finish goal, first marathon, or legs that fail late	Run/walk from the start is an option: no time cost shown, less soreness, no cardiac benefit	REF-894 B+: 42 non-elite runners randomised, 4:14:25 vs 4:07:40 (p = 0.377), cardiac markers alike and back to baseline by day 4, less muscle pain and fatigue (p = 0.006); run:walk ratio not in the abstract — the ratio is practice
Hot race	Plan drinking from a sweat-rate test, never beyond losses	REF-225 C (editorial: plan in heat >60 min); REF-776 A-; plan = practice
History of GI trouble	Train the gut over weeks (practice; the timeline is not established, REF-008); expect less tolerance running than riding	REF-787 B+ (GI history tracked symptoms, r = 0.37-0.51); REF-525 A-; REF-008 B+

The wall, and what prevents it

Who and how often: in more than 4 million recreational race records, 28% of men and 17% of women hit the wall (a pacing-defined late slowdown); it lasted 9.6-10.7 km at a relative slowdown of 0.37-0.40 against 5-20 km pace; 36% hit it in the 3 years before a recent PB vs 23% earlier; minutes lost grew with ability (REF-775, B+). It measured no fuelling; onset after 20 miles is its definition, not a result.

Mechanism: muscle glycogen fell toward zero during work and capacity dropped once it was depleted (REF-452, B, 1967, normal subjects); higher starting glycogen was used faster (5.0-5.1 vs 3.1 mmol/kg per km) (REF-774, B).

Prevention — inference, not a tested protocol: start with full stores, eat carbohydrate early, and do not spend the store on a fast first half. A model (no participants, unvalidated) puts the wall's distance on pace as a fraction of aerobic capacity, glycogen stores and muscle mass distribution, so pace, loading and in-race fuel each move it (REF-823, C); its "more than two-fifths hit the wall" is framing — REF-775's race data are lower.

Race week and race day

Carbohydrate loading. In a narrative review, raising starting glycogen postponed fatigue by about 20% and improved set-distance performance 2-3% in events over 90 min, and did nothing for 60-90-min running or cycling (REF-819, B+) — which explains why, in trained runners, a no-depletion regimen (3 days at 50%, then 3 at 70% carbohydrate) matched the classic depletion diet on gastrocnemius glycogen (203 vs 207 mmol/kg, vs 159 unloaded) but neither made a 20.9-km run faster (REF-774, B, n not reported).In 8 trained men resting and eating 10 g/kg/day of high-glycaemic carbohydrate, glycogen rose 95 → 180 mmol/kg in one day, with no further rise over two more (REF-891, B, uncontrolled, no performance test). So: no depletion phase.Not from a study; the base's own practice: 1-3 days of high-carbohydrate eating inside the taper, with the last day easy; 10 g/kg is what reached maximal stores in one resting day, not a tested marathon dose.

Carbohydrate in the race. IOC review: longer events 30-60 g/h, events >2.5 h up to 90 g/h with multiple-transportable blends (REF-066, B+). The base's matrix: 90 min-3 h 60-90 g/h amateur / 80-100 sub-elite / 90-120 elite; 3-6 h 80-100 / 100-120 / 110-120 (SYNTH-02, B- — practice, built largely on cyclists and mountain runners).Marathoners in one field survey averaged 35 g/h (REF-787, B+). ≥2 h at 60% VO2max is where significant GI disturbance appears, and running and heat worsen it (REF-525, A-); the gut adapts to practice (REF-008, B+).Inference: for most amateur marathoners the target is 60-90 g/h (below SYNTH-02's 80-100 g/h amateur cell for 3-6 h events; the runner's discount is the base's own inference, not tested), built over weeks of long runs from today's intake; 120 g/h is a proposed trained-athlete ceiling, not a runner's default (REF-3490, B+).

Drinking: to thirst. The international consensus attributes exercise-associated hyponatraemia (EAH) — and all known deaths from it — to drinking more than is lost (Grade 1A); drinking to thirst is the recommended prevention (1C); sodium drinks cannot prevent EAH in someone who keeps overdrinking (1C); gaining or holding weight is the warning pattern, and <0.75 kg lost over a standard marathon has been associated with asymptomatic EAH (REF-776, A-, 2015).A model found sodium replacement unnecessary in an elite marathon (REF-223, C). Practice: no "drink at every station" rule.

Pacing. Performance in events over 2 min may improve with more even pacing; the optimum beyond 4 h is unknown (REF-010, B+, 2008 review).In 1.7 million recreational marathoners, fast starts went with poorer finish times and so did fast finishes (read as over-caution earlier); women paced more evenly but were sometimes too cautious (REF-820, B+, observational, no effect sizes in the abstract).Not from a study; the base's own practice: first half at goal pace, never faster (a few seconds per km slower only for a first marathon or a hot day), then even effort (let HR drift, REF-159); a big late surge means the start was too cautious. Walls were commoner in the years around a PB (REF-775).

Super shoes. Plated shoes lowered metabolic demand ~2-3% (mean −2.75%) in 14 crossover trials, with no performance or injury outcome (REF-720, A); in world-class male marathons race speed was 1.0% faster (~79 s) (REF-721, B+, observational, Tier 5); the saving held at 7.5-12 km/h in 14 moderately trained runners (REF-723, B) and was smaller uphill (REF-722, B+).Injury effect unmeasured (REF-724, C+). Practice: break them in on long runs and marathon-pace sessions, not on race day.

Taper. Two weeks, volume −41-60%, intensity and frequency kept (REF-107, A); every band up to 21 days improved time trials, ≥22 days was not significant (REF-3492, A).In over 158,000 recreational marathoners, strict tapers (weekly distance cut further each week) beat relaxed ones and longer tapers up to 3 weeks beat shorter ones; a strict 3-week taper went with a median 5 min 32.4 s (2.6%) faster finish than a minimal taper, more in women (REF-821, B+, observational, distance only).Elite British marathoners reported a median 14-day taper, continuous volume to 53% of normal, peak interval speed 114% of race speed (REF-818, B, survey); world-class runners start 7-10 days out (REF-089, B+). Practice: 2-3 weeks, each week lighter than the last, hard sessions kept but shortened (SYNTH-34).

Half marathon: what changes

Half-marathon entries: an injury ramp (REF-623), championship pacing (REF-824) and one cross-section of women (REF-825); no training trial — the rest interpolates SYNTH-47 and the marathon evidence.

Training (practice): long run ~90-120 min (the ≥90-min habit of REF-160); threshold/critical-speed work is the main quality session; half-marathon-pace reps grow over the last 6-8 weeks. In 42 recreational women, faster mean training speed went with a faster half (r = −0.68, significant after adjusting for body weight and skinfolds), weekly volume did not hold up (REF-825, B, cross-sectional — faster runners may simply train faster).

Loading: not needed for a half finished inside ~90 min (REF-819 B+; REF-774 B found no gain at 20.9 km); optional for slower runners.

Pacing: in six World Half Marathon Championships (491 men, 347 women), the best held an even speed from 5 to 15 km while slower runners faded from 5 km, and those in a pack throughout slowed least (REF-824, B+, elite, association). Practice: even from the early kilometres, in a group at goal pace.

Fuelling: <90 min 30-60 g/h; 90 min-3 h 60-90 g/h amateur (SYNTH-02, B-); many finish under the 2-h GI threshold of REF-525 (inference).

Drinking: thirst still; inference from REF-776: EAH risk grows with duration and overdrinking, so lower here, not zero.

Taper (practice): 10-14 days. Wall: no half-marathon data in the base.

After the race

In 64 recreational finishers (54 men, 10 women), three easy 40-min runs at 48, 96 and 144 h (95-105% of HR at VT1) did not slow CK or LDH recovery vs rest, and runners jumped better at 96 h (108.29% vs 100.58% of pre-race) (REF-761, B+).

In 86 runners, troponin was back to baseline by 96 h (but differed again at 192 h), CK by 144 h, LDH by 192 h, and CRP was still up at 192 h (REF-762, B, uncontrolled). Practice: easy running from day 2-3, no hard session for ~2 weeks; mention the race to anyone reading a blood test that week.

How to progress / periodise

Practice, on SYNTH-42's horizon bands: >16 weeks out, volume and the engine; 16-8 weeks, both quality days, long run growing; last 8-10 weeks, marathon pace in the long run, more polarized elsewhere (REF-509 for direction); last 2-3 weeks, taper. Practice: check with a half marathon or a marathon-pace long run 4-6 weeks out.

When NOT to use it / red flags

Worsening headache, nausea or vomiting, confusion or unusual drowsiness during or after a long race in someone who drank a lot: medical help now, no more water or sports drink until assessed. The consensus says symptoms run from mild to severe encephalopathy and that a hyponatraemic athlete gets no hypotonic or isotonic fluid until urination (1C) (REF-776); the symptom list is the base's own practice. Never advise drinking more for these.

Do not add a third hard session to make up for missed long runs (SYNTH-26).

Localised bony pain that worsens on runs, or pain at rest: referral, not a training decision (practice).

What the evidence does and does not show

Shows: the cited findings above, most of them associations or single trials; the strongest are the taper meta-analyses, the shoe metabolic meta-analysis, the EAH cause (Grade 1A), and very large observational datasets on recreational pacing and taper (REF-820, REF-821).

Does not show: marathon-length loading or in-race fuelling trials in runners; long-run or marathon-pace dose; a tested pacing plan or taper (both observational, self-selected); a run:walk ratio (REF-894); amateur volume targets; causation behind REF-197.

Evidence gaps

A marathon-distance loading trial with a performance outcome (REF-819 is a narrative review; REF-891 measured glycogen only, at rest).

A randomised test of a pacing plan or of taper length in recreational marathoners (REF-820, REF-821 are observational); non-elite half-marathon pacing (REF-824 is championship racing).

Marathon-pace physiology in recreational runners and women (REF-822 is 16 world-class men).

In-race carbohydrate trials in road marathoners; long-run and marathon-pace dose trials; any half-marathon training study; women or masters marathon training.

How this entry was checked and changed

Why this level

(volume and distribution: B+ cross-sectional REF-197 and B+ review REF-773; the wall: B+ observational REF-775; drinking: A- consensus REF-776; loading: a B+ review REF-819 and B studies REF-774 REF-891; pacing and taper: B+ observational data on 1.7 million and 158,000 recreational runners REF-820 REF-821; run/walk: B+ RCT REF-894; shoes: A metabolic meta-analysis REF-720; recovery: B+ RCT REF-761. Long-run length, marathon-pace dose, in-race grams for runners and the pacing plan are house practice)

Content correction 2026-09-27

Revision 2026-09-29: folded in REF-819 (Hawley 1997: loading pays past ~90 min) and REF-891 (Bussau 2002: one resting day at 10 g/kg maximised glycogen) — house loading changed from 2-3 days to 1-3 days; REF-820 (Smyth 2018: fast starts and fast finishes both slower) — house pacing changed from "at or a few seconds slower" to "at goal pace, never faster; slower only for a first marathon or heat"; REF-821 (Smyth & Lawlor 2021) and REF-818 (Spilsbury 2014) — house taper changed from 2 (up to 3) weeks to 2-3 weeks, each lighter than the last; REF-894 (Hottenrott 2016: run/walk) as a new row; REF-822 (Jones 2021) as a goal check; REF-823 (Rapoport 2010, C) at the wall; REF-824 (Hanley 2014) and REF-825 (Knechtle 2011) in the half.Four candidate gaps closed. Level unchanged (B).

Verification 2026-10-01: 44 cited REF entries opened in the KB.md snapshot (all have DOI or PMID, none NOT IN SEARCH/UNVERIFIED); numbers, n, design and level labels match for 53 of 58 ref-bearing sentences/cells. No WRONG, no LEVEL mismatches.

Verification 2026-10-01: SYNTH-02, SYNTH-34, SYNTH-44, SYNTH-26, SYNTH-68 opened for the lines SYNTH-64 attributes to them; matrix figures, 'twice a week' strength and 'no third hard session' are quoted correctly. The hot-race row (plan from a sweat-rate test) agrees with SYNTH-68 but not with SYNTH-65's hot-race row ('thirst still guides') — logged under SYNTH-65.

Verification 2026-10-01: the Level pair is a judgment call (B → B-); apply only if the owner agrees. The Revision log line 'Level unchanged (B)' was left untouched.

2026-10-01: level tags next to cited frameworks updated to their current levels after the independent audit.

Revision 2026-10-16: REF-801's submaximal and isometric economy results reworded to "not significantly" (isometric pool: three small calf-only studies, ES −0.269, CI −0.79 to 0.252; see SYNTH-88). Level unchanged.