REF-1505 · Figueroa, 2026b · SportRxiv — preprint, version 2, final production version dated 23 August 2026, posted 5 September · Paper
Is the same sled harder for a woman?
PREPRINT — not peer reviewed: Pro women ran like Open men (+0.3%) but took 33.3% longer on the same 152-kg sled push, 2.5% on the same lunges. Populations, not people.
Main caveat: PREPRINT — not peer reviewed. On SportRxiv; Crossref showed no published version on 2026-10-30. kb-watch checks monthly for one (relation is-preprint-of).
Figueroa sealed thirteen predictions from his men's paper (REF-1504) before extracting any women's split, then replicated the design in the women's Open division: census 125,079 athletes and 182,432 results, 54,889 races with full splits over five corpora, including 575 athletes with at least 5 valid races over at least 3 years and 4,216 athletes moving between Open and Pro.
The profile plane matched the men's (Tucker 0.985/0.983), and stability was 0.85 and 0.90 corrected. The Open-to-Pro penalty was +0.338 log (about 40%), equal across profiles within ±2% per SD. Elasticities ran from 1.33 on the sled push (the same as in men) to 0.35 on lunges.
Because Pro women and Open men move identical masses at the sleds, carry and lunges, and run almost the same (+0.3%), the author compared them within 179 event-years and found women's excess largest on the sled push and smallest on the lunges. He reads this as load being relative to the athlete, labelled suggestive.
The gap between top and bottom had shares within two percentage points of the men's. Four differences survived his checks, all suggestive. Our reading: the same strengths and limits as the companion paper, with a well-used comparison; the groups compared are not the same people, and Pro women are a selected group.
Protocol
Data: public results service hyresult.com; primary vendor site not accessed; pseudonymised. Valid women's cross-section 2021-2026; 2018-2020 races carry an old format and fail validation.
Census: women's Open 125,079 athletes and 182,432 results, 2018-2026.
Full-split corpora (division, sex, unit): n = 54889 unique races, all women. Longitudinal: at least 5 valid races over at least 3 years, 668 targeted, 575 with age, 3,903 in-cohort races. Pro transition: 4,216 athletes, 22,387 valid races, 9,630 Open-Pro pairs within 1 year. Doubles: 8,351 partners, 2,517 dyads, 3,359 analysable team races. Cross-section: 29 full-field events 2024-25, 16,294 athletes. Debutant retention panel: 13,173 in the index.
Repeated athletes: followed on purpose; overlaps between corpora itemised (2,221 races serve more than one).
Pre-registration: thirteen predictions (P1 to P13) sealed on 14 August 2026; replication means same direction with the female 95% CI containing or exceeding the male effect. A matched-structure test used 200 male sub-cohorts of 575 athletes.
Design gate: power 0.21 to detect half the male ageing effect, so all longitudinal analyses in women are capped at exploratory by the author's rule.
Cross-sex bridge: Pro-division women vs Open-division men, same event and year, 179 events with at least 10 finishers each side; loads 152 kg, 103 kg, 2 × 24 kg, 20 kg and a 6-kg wall ball, with target height 2.70 m vs 3.00 m, so wall balls are reported apart. Anchored on the eight runs.
Funding and interests: no external funding; no conflicts declared. Sole author; data pipelines, analyses, figures and drafting produced with the assistance of an AI system (Claude, Anthropic) operating under his direction, as declared; design decisions, data collection, verification and final content are stated as his responsibility.
Key Results
| Same mass (Pro women vs Open men) | Excess over the running anchor (95% CI) |
|---|---|
| Sled push, 152 kg | +33.3% (+31.9 to +34.7) |
| Sled pull, 103 kg | +27.0% (+25.6 to +28.3) |
| Farmers carry, 2 × 24 kg | +24.6% (+23.6 to +25.5) |
| Sandbag lunges, 20 kg | +2.5% (+1.6 to +3.4) |
| Wall balls, 6 kg; target 2.70 vs 3.00 m | −1.0% (−2.0 to −0.0) |
The anchor itself: Pro women +0.3% vs Open men on the eight runs.
| Top-to-bottom gap, share | Women | Men |
|---|---|---|
| Runs | 45.9% | 43.9% |
| Ergometers | 3.5% | 3.5% |
| Strength (sleds, farmers) | 10.0% | 11.4% |
| Power (wall balls, lunges, burpees) | 30.8% | 30.0% |
| Transitions | 9.7% | 11.2% |
A share of a gap, within a fixed order; not what to train.
| Finding, women | Number | Author's label or note |
|---|---|---|
| Stability | 0.847 (0.744-0.939) and 0.897 (0.789-0.991) | Replicates |
| Open to Pro | +0.338 log (0.334-0.343), about 40%; 9,630 pairs, 4,216 athletes; TOST ±2%/SD p = 1.5e-4; strength +0.8%/SD (p = 0.02) | Athletes who chose to move |
| Elasticity | 1.33 sled push to 0.35 lunges; r = 0.95 with men's values; r = −0.26 with added load | Only these two printed |
| Roxzone | Stable 0.785 (0.646-0.902); 9.7% pooled and 10.1% median across 169 events of the gap; debut +1.4%, gone by race two; Open to Pro +10.3% (9.2-11.4); R² with level and plane 0.307 vs 0.525 in men | Suggestive sex difference |
| Doubles | On ability r = 0.386 (0.369-0.403); complementary −0.14 percentile points per SD (−0.88 to +0.60) vs men +2.5; run-to-run CV −11.7% | Complementary premium absent |
| Movers | 5.7% (33/575); lose 17.6 percentile points more than non-movers (7.8 to 27.3) | Exploratory |
| Age | Power +0.028 log per decade (0.018-0.039), cross-sectional | Within-athlete fit discarded |
| Retention | 50+ return 46.5% vs under 30 38.9% | Survivors, by the author's note |
| "Dead" candidates | Set-point reversion, weaker assortment, specialisation helping speed | Died under the author's own checks |
Run after a station: not analysed. Roxzone: the top decile sped up across transitions late in the race and the bottom decile slowed (interaction p = 4e-39).
Mechanisms
Measured: split times and their within-race composition; ranks; loads from three sources stated as concordant.
Author's interpretation: performance cost tracks load relative to the athlete, ordered by each station's mechanical regime, so women's apparent offset on the shared plane mostly reflects division calibration; labelled suggestive.
Inference: the bridge matches populations on running, and that match is an accident. Pro women are selected fast runners relative to their sex; Open men are everyone. A person-matched comparison was not possible. Body mass was unavailable, so a load relative to the athlete is reasoned, not measured.
Practical Application
Is the same sled harder for a woman? PREPRINT — not peer reviewed: Pro women ran like Open men (+0.3%) but took 33.3% longer on the same 152-kg sled push, 2.5% on the same lunges. Populations, not people.
Choosing Pro as a woman: on average the loaded stations took about 40% longer against the unchanged segments, regardless of profile (one preprint, self-selected movers); per-station costs differ, so this is not one athlete's forecast. Sled push responds most to load, lunges least, in this analysis.
Women's doubles: partners match on ability; no complementary premium here, against a suggestive one in men. One preprint each.
Do not use it to set a women's training split: a share of the gap is not a share of the training week, and the replication says the sport's shape is alike in both sexes, not what to do about it. Not from this paper; the base's own practice is in SYNTH-71.
Limitations
PREPRINT — not peer reviewed. On SportRxiv; Crossref showed no published version on 2026-10-30. kb-watch checks monthly for one (relation is-preprint-of). Level capped at C. Its version note names "panel rounds" that appear to be the author's process (one panel is called pre-submission), not journal review.
Single author, AI-assisted, same dataset family. A replication by the same person with the same code on a different sex, with disclosed AI assistance; not an independent one. A Zenodo deposit exists; not rerun by us.
Observational, finishers only, one ecosystem. The author says no intervention question is answered here.
Underpowered strata. Longitudinal analyses are exploratory (575 athletes, power 0.21); 14 movers for hysteresis, 18 exits, partner-switch not run.
Bridge. Populations, not persons; running match accidental; no body mass; wall-ball geometry differs.
Fixed order. Stations are confounded with race position.
Extraction losses. Five 2025 events without usable splits, about 1.5-2.5% per corpus, not random.
"Sex difference" labels. Four candidates remain, suggestive and pending a 2027 replication; three others died under the author's checks.
How this entry was checked and changed
Why this level
read from the full text (SportRxiv PDF, CC BY 4.0). A large cohort would earn B+ by design; the preprint rule sets one step below and caps at C
Full text (checked 2026-10-30)
Thirteen predictions sealed before any women's split; 54,889 races with full splits across five corpora; census 125,079 athletes (full text)
Tucker congruence 0.985/0.983; disattenuated stability 0.847 and 0.897 (full text)
Load penalty +0.338 log, about 40%; 9,630 pairs from 4,216 athletes; TOST ±2%/SD p = 1.5e-4; elasticities 1.33 sled push, 0.35 sandbag lunges; r = 0.95 with male values (full text)
Bridge: running anchor +0.3%; sled push +33.3%, sled pull +27.0%, farmers +24.6%, lunges +2.5%, wall balls −1.0%; 179 events (full text)
Gap: runs 45.9% vs 43.9%, ergometers 3.5% vs 3.5%, strength 10.0% vs 11.4%, power 30.8% vs 30.0%, transitions 9.7% vs 11.2% (full text)
Complementarity −0.14 pp/SD (−0.88 to +0.60), n = 3,359 team races (full text)
No external funding; no conflicts declared; AI assistance (Claude) declared (full text)