Should load rise across my blocks before a peak?
In 46 elite sprint swimmers (105 seasons), the fastest pattern had load rising into the second macrocycle and a peak ~5 weeks out; flat, falling load went slowest. Observational.
Hellard and colleagues analysed the training records of 127 elite swimmers (60 men, 67 women; 46 sprint, 81 middle-distance) followed for a median of 3 seasons between 1992 and 2012, 289 swimmer-seasons in all.
For each, they took the 25 weeks before the season's best performance, scored each week's swimming by lactate intensity and its dryland work as a percentage of that swimmer-season's heaviest week in the same category, averaged the five normalised categories into a weekly total training load, and grouped the load curves with latent class mixed models.
Across the cohort 86-90% of swimming was at or below 4 mmol/L (40-44% below 2 and 44-46% between 2 and 4), 6-9.5% at 4-6 mmol/L and 3.5-4.5% above 6 mmol/L.
Among sprinters, three patterns emerged: Long (53% of swimmer-seasons) — two macrocycles, load rising from the first to the second, moderate week-to-week variation (6 ± 3%) and more training at ≤ 4 and > 6 mmol/L — swam fastest (6.0 ± 3.1% off the mean of the year's ten best world times), ahead of Balanced (6.8 ± 3.5%, P = 0.31) and Stable Flat (7.4 ± 2.9%, P < 0.05), a flat, high-volume pattern with load falling between the halves and the most 4-6 mmol/L work.
Among middle-distance swimmers, four patterns differed in performance by amounts that did not reach significance. The authors recommend macrocycles of about 15 weeks with progression between them, a load peak about 3-5 weeks before the major competition and little 4-6 mmol/L work.
Our reading: a large, real-world data set that supports "build the second macrocycle on the first, do not let it sag", but the patterns also tracked the Olympic cycle in which they were recorded, most swimmers changed pattern from one season to the next, only one performance contrast reached significance, and the paper gives several different numbers for the macrocycle length and the peak.
Protocol
Participants: n = 127 (60 men, 67 women), aged 15-30; sprinters 46 (4 at 50 m, 42 at 100 m), middle-distance 81 (52 at 200 m, 29 at 400 m and longer); two national training centres; followed a median of 3 years (range 1-11); 7 ± 4 competitions a season. Consent from swimmers and from parents of those under 16; institutional review board of INSEP.
Exclusions: chronic illness or injury needing treatment, 4 weeks or more of missed training, medication affecting immune function or inflammation; and swimmer-seasons "with very slow performances or noticeably irregular training programs".
Window: the 25 weeks before each swimmer's best performance of the season (the longest period common to all); 47% of best performances came in the first macrocycle (national selection) and 53% in the second (the major international meet).
Performance: best time relative to the mean of the 10 best world performances that year for the same sex, stroke and distance (Pr, %; lower is better).
Intensity: a 6 × 200 m incremental lactate test at the start of each season, repeated and adjusted 4 times a season; swimming classed as I1 < 2, I2 2-4, I3 > 4-6, I4 > 6 mmol/L and I5 maximal speed, grouped as moderate-to-heavy (≤ 4), severe (4-6) and extreme (> 6); dryland as general conditioning and maximal strength (1-6 reps, 80-100% 1RM), in minutes.
Total training load (TTL): each week's volume at each intensity scaled 0-100% of that swimmer-season's maximum at the same intensity, averaged across the five categories; variation = mean difference in TTL between consecutive weeks.
Analysis: latent class mixed models (lcmm, hlme) per distance class, number of classes chosen by BIC and interpretation; ANOVA or Kruskal-Wallis between patterns; paired tests between the first 13 and last 12 weeks; MANOVA on log-ratio-transformed proportions; significance 0.05, no multiplicity correction ("exploratory").
Funding and interests: partly funded by INSEP and the French Ministry in charge of sports (grant 14r21); no commercial or financial conflict declared. The first author provided the funding and collected the data.
Key Results
| Pattern (sprint, 105 swimmer-seasons) | Share | Performance, % off world top-10 | Week-to-week variation | First to second half | Peak in second macrocycle (Table 1) |
|---|---|---|---|---|---|
| Long | 53% (56) | 6.0 ± 3.1 (fastest) | 6 ± 3% | progressive | 5 weeks before |
| Balanced | 24% (25) | 6.8 ± 3.5 (P = 0.31 vs Long) | 8 ± 6% | degressive | 10 weeks before |
| Stable Flat | 23% (24) | 7.4 ± 2.9 (P < 0.05 vs Long) | 4 ± 3% | degressive | 8 weeks before |
| Pattern (middle-distance, 184 swimmer-seasons) | Share | Performance, % off world top-10 | Week-to-week variation | First to second half | Peak in second macrocycle (Table 3) |
|---|---|---|---|---|---|
| Balanced | 25% (47) | 5.8 ± 3.4 (fastest) | 6 (6)% | progressive | 9 weeks before |
| Stable Flat | 20% (36) | 5.9 ± 3.4 | 6 (5)% | degressive | 3 weeks before |
| Long | 36% (66) | 6.3 ± 3.0 | 6 (5)% | progressive | 6 weeks before |
| Short | 19% (35) | 7.3 ± 4.7 (P < 0.10, 0.13 and 0.18 vs Balanced, Stable Flat, Long) | 8 (8)% | degressive | 6 weeks before |
Intensity distribution, whole cohort: 86-90% at ≤ 4 mmol/L (40-44% below 2, 44-46% at 2-4), 6-9.5% at 4-6, 3.5-4.5% above 6 mmol/L. The fastest groups did more at ≤ 4 and > 6 mmol/L and less at 4-6.
Absolute sprint volumes (moderate-to-heavy): Balanced 26,300, Long 30,600, Stable Flat 34,010 m a week; severe 2,550, 2,590 and 3,350 m a week.
Patterns are seasons, not swimmers: 80% of sprinters and 87% of middle-distance swimmers followed for more than one season did not repeat their pattern.
Patterns follow the era: among sprinters the Long pattern came mostly from 2000-2012 seasons, Stable Flat mostly from 1996-2000 (62%) and Balanced mostly from 1992-1996 (52%) (P < 0.001); Stable Flat swimmers were the youngest (18.8 ± 2.6 y).
Numbers that do not add up: (1) the abstract describes the fastest sprint pattern as "two long ∼14–15-week macrocycles", which cannot both fit in a 25-week window; the body gives the second macrocycle as about 14-15 weeks, "15–16 weeks" elsewhere in the Discussion, and the Results text says it was the Stable Flat group that had the longer second cycle (about 16 vs 13 weeks) while Table 1 gives "longer 2nd cycle" to the Long group.(2) The peak before the best performance in the fastest sprint group is given as 5 weeks (Results, Table 1; Fig. 3's "week 6", like its "week 22" against the Results' 21, is one week later throughout, probably week numbering), 4 weeks (Discussion), "4–6 weeks" (abstract), "∼3–5 weeks" (recommendation) and "7, 6, and 5 weeks" (take-home message); the first-macrocycle peak as 21 weeks (Results) and week 22 (Fig. 3).(3) The Long sprint group's moderate-to-heavy volume is printed as 30,600 ± 91 m a week, an implausibly small SD beside ± 9,600 for the other groups.(4) In Table 4 the specialty rows do not add to the group totals — Long 35 of 66 (freestyle 11 printed as 46%), Stable 38 of 36, Unstable 38 of 35, and the Results attribute to the Balanced group the most pre-Olympic seasons, the fewest Olympic seasons and the most new entrants, while in Table 4 the Balanced group has the fewest pre-Olympic (13%) and the most Olympic seasons (38%), and the fewest Olympic seasons and most new entrants are in the Short ("Unstable") group.
Abstract and body: the abstract says the window preceded "the major competition of the season"; the methods use the season's best performance, which for 47% of swimmers came in the first macrocycle.
Mechanisms
Measured: training load, its variation and its distribution across lactate zones; no physiological outcome.
Authors' interpretation: progressive overload keeps the stimulus growing (they cite Selye's general adaptation syndrome); long cycles allow cumulative adaptations, which, citing others, begin in 2-3 weeks, culminate between 4 and 7 weeks and continue to 12 weeks or more; swimmers may train more at 2-4 mmol/L than land athletes partly because technique and efficiency are best near those speeds.
Inference: the pattern differences are confounded with era, coaches and the Olympic calendar (the authors themselves read the era shift as changing coaching practice); and a swimmer whose season goes well can keep raising the load, while illness or staleness makes it fall — so a falling load may be a consequence of a poor season as much as a cause.
Inference: "progressive" and "medium load" are relative to the swimmer's own heaviest week in that season; the Stable Flat sprinters had the largest absolute volumes, so the finding is about the shape of the load curve, not about doing less.
Practical Application
Should load rise across my blocks before a peak? In 46 elite sprint swimmers (105 seasons), the fastest pattern had load rising into the second macrocycle and a peak ~5 weeks out; flat, falling load went slowest. Observational.
Do not let the second half of a build sag below the first (B, observational, elite swimmers): the slowest sprint pattern and the slowest middle-distance pattern both had load falling between the halves of the 25 weeks. The second-fastest middle-distance pattern also fell, so this is a tendency, not a rule.
Place the heaviest weeks about a month to six weeks before the target, then taper (our reading: the paper places this peak anywhere from 3 to 7 weeks out in different sections, 5 weeks in Table 1, followed by a 2-4-week progressive decrease); the taper itself is SYNTH-34.
Vary the weeks moderately. A sprint pattern with about 6% change week to week was faster than a flat one with about 4%; an association, not a tested prescription (SYNTH-29 for load spikes).
Block lengths inside a macrocycle are the authors' advice, not their data: 2-3 mesocycles of 3-5 weeks with 1-2 recovery weeks (Discussion), or 4-6 weeks (Conclusion), within about 15 weeks. That sits with the base's house default of 3-6-week blocks (SYNTH-42); mesocycles were not analysed here.
Transfer: swimmers with a threshold-heavy distribution and loads scaled to their own maxima; how this shape carries to cyclists or runners with 6-12 h a week is untested.
Limitations
Observational and exploratory. Associations between load patterns and best times; no multiplicity correction; the authors say causality cannot be inferred.
Weak performance contrasts. Only Long vs Stable Flat in sprinters reached P < 0.05; the middle-distance patterns did not differ significantly; SDs were large in every group.
Confounding. Era, coaches, Olympic quadrennial, age and experience differed between patterns; swimmer-seasons with very slow times or irregular training were excluded.
Relative, lactate-only load. TTL is scaled to each swimmer's maximum and built from lactate-zone volumes and dryland minutes, which the authors call a partial view subject to measurement error.
Ends in 2012. The authors note training practice has changed since.
Reporting. Macrocycle lengths, peak timings, one SD and parts of Table 4 do not agree with each other (see Key Results).
Funding: INSEP and the French Ministry in charge of sports; the first author is from the French Swimming Federation's research department; no commercial conflict declared.
How this entry was checked and changed
Why this level
observational cohort below the rubric's 200-participant mark for B+ (127 swimmers; the sprint analysis rests on 105 swimmer-seasons from 46 swimmers), exploratory, with no correction for multiple comparisons; read from the full text (open access, CC BY 4.0)
Full text (checked 2026-10-26)
Elite male (n = 60) and female (n = 67) swimmers (46 sprint, 81 middle-distance) over 20 competitive seasons (1992–2012); 289 individual swimmer-seasons from 127 swimmers followed for a median of 3 years (range 1–11) (full text)
Sprinters: the Long group showed the best performance (6.0 ± 3.1%) compared to the Balanced (6.8 ± 3.5%, P = 0.31) and Stable Flat (7.4 ± 2.9%, P < 0.05) groups (full text)
Sprint week-to-week variation: 4 ± 3% Stable Flat, 6 ± 3% Long, 8 ± 6% Balanced; 53, 24 and 23% of swimmer-seasons in the Long, Balanced and Stable Flat groups (full text)
Among swimmers followed for more than one season, 80% (sprint) and 87% (middle-distance) did not train with the same profile (full text)
Middle-distance: Short 7.3 ± 4.7% vs Balanced 5.8 ± 3.4% (P < 0.10), Stable Flat 5.9 ± 3.4% (P = 0.13) and Long 6.3 ± 3.0% (P = 0.18) (full text)
About 86–90% of training at [La]b ≤ 4 mmol⋅L-1: 40–44% below 2 and 44–46% at 2 to ≤4, leaving 6–9.5% at >4–6 and 3.5–4.5% at >6 mmol⋅L-1 (full text)
Fastest sprint group: two load peaks 21 and 5 weeks before the best performance (Results); peak load in week 22 and week 6 (Fig. 3 caption); a load peak 4 weeks before (Discussion); peaks at 7, 6, and 5 weeks (take-home message) (full text)
47% of swimmers achieved their best performance in the first macrocycle and 53% in the second (full text)
Association tests used a significance level of 0.05 without multiplicity corrections; partly funded by INSEP and the French Ministry in charge of sports, grant no. 14r21 (full text)