Where do I lose time?

Ask what race, then compare like with like. Division, Singles or Doubles, sex, age group, venue, and the athlete's earlier splits. Read runs 2-7 first: the rules let the first or last lap of a run vary, to be offset within that run, yet run 1's spread is about twice run 2's in PRO (REF-1496, B+) and one preprint calls run 1 often about 800 m (REF-1519, C). Not from a study; the base's own practice.

Own past races first, then a named group; never a verdict. Percentile tables by category exist in REF-901 (B+, 186,411 results to 2023/24), but their values are not in the base: quote none. The table below describes groups. Open has no split table in the base, so a first Open race has no comparison from here: it is the baseline. Not from a study; the base's own practice.

Say where the time went, as description. Not "running decides" and not "stations decide": both are part-whole associations in a fixed order, among finishers. A segment far behind the athlete's own others is a place to look, and SYNTH-71 decides what changes in training (inference; practice).

Run 1: at or a little under what you can hold for all eight. Not from a study; the base's own practice. In numbers: the mean of runs 2-7 of your last comparable race, or of the off-station runs in a rehearsal; with neither, do a partial simulation first (SYNTH-71).

That is slower than PRO finishers opened (REF-1496, REF-902; their run 1 may also be short): the fade is what they did, not a plan to copy, and no trial tests holding back. Even pacing in long events and the cost of a fast marathon start transfer as direction (REF-010, B+, narrative; REF-820, B+).

Run 1 comes before any station, so the triathlon test below does not speak to it.

Runs after stations: expect the fade, and take the pace from off-station running in rehearsal, not from fresh 1-km or 5-km pace. Not from a study; the base's own practice (REF-1518, B-, direction only, triathletes off a bike).

Opening a run off the bike at +5% lost 150 s and at −10% lost 59 s against −5%, about 15 and 12 s a percentage point (our arithmetic; REF-1508, B): at most an inference here, a floor on caution, not a target. Judge a run by how soon the pace settles. Hold a reserve for lunges and wall balls (practice; the peak is REF-510).

The runs are not rest: typical heart rate was above 90% of maximum from run 3 (REF-1497, B-).

Roxzone and stations: brisk and controlled. Moving through Roxzone is cheap time by REF-902's inference; compare it only within one venue (inference; layouts vary by venue, REF-1496, B+; layout was 66% of its raw variance in Open men, REF-1504, C). No study tests how hard to go at a station; even work with no all-out surge before a run is the base's practice. Learn the movement standards in the current rulebook.

Division changes the stations, not the running; Doubles changes both. Loads and rules are in the table below (the event's rules, dated; check the current rulebook). In Doubles agree who does how much at each station and rehearse the changeovers; the partners must run together, so the slower partner sets the pace. Not from a study; the base's own practice.

Race day by pointer, rehearsed in a simulation. Fuel by SYNTH-02's rows, drink by SYNTH-68, caffeine by SYNTH-67 after a training trial, no buffer by default. All transfer; no trial at HYROX length.

Test by repeating the same thing; predict no time. Repeat one partial simulation with the same stations, loads, order and rest, and optionally a fresh 5 km, under SYNTH-53's rule of one protocol. Not from a study; the base's own practice. A goal the athlete sets is a goal, not a prediction.

Running is about half. 47.9% to 50.3% in the faster half of PRO (our arithmetic; REF-1496, B+); ~50% over seven seasons of PRO and ELITE (REF-903, B+); 50.7% and 51.1% in Worlds top 10 by age group, stations 44.2% and 43.7%, Roxzone 5.1% and 5.2% (REF-1519, C); about 59% in 11 recreational athletes in a simulation, 51.2 of 86.5 min (our arithmetic; REF-510, B). None ranks what to train.

The gap by level. G2 to G5 PRO men: 22:25 overall, 9:19 in runs, 10:36 at stations, 2:30 left over; women 22:46, 9:49, 10:36, 2:21 (our arithmetic; REF-1496). Largest gaps in men: sled pull 2:26, wall balls 2:18, lunges 1:45, burpee broad jumps 1:36; SkiErg 0:20, row 0:30.

Running ranks differed little from overall ranks, strength stations reshuffled more and cost slower athletes more (REF-903). Men's and women's Open, top to bottom: runs 43.9% and 45.9% of the gap, wall balls, lunges and burpees 30.0% and 30.8%, sleds and carry 11.4% and 10.0%, ergs 3.5%, transitions 11.2% and 9.7% (REF-1505, C; a share of a gap, not of a training week).

Profiles form a continuum, not runner and station types, stable across a career, 0.87-0.92 men (REF-1504, C), 0.85 and 0.90 women (REF-1505); neither has training data.

Pacing shape. In PRO it was alike across levels (η²p ≤ .054) and between sexes (η²p about .005), with more fade at the end in slower bands: run 8 against the mean of runs 2-7 was +17 s in G2 men, +39 s in G5 men, +16 and +33 s in women (our arithmetic; REF-1496). 24 PRO men fell from 16.13 km/h on run 1 (just below VT2 speed, 16.41) to 13.58 on run 8, averaging 84.13% of VT2 speed; Roxzone 10.52 km/h (REF-902, B, one race; describes, tests nothing).

In a simulation men went ~13.2 to ~11.3 and women ~11.9 to ~10.3 km/h from km 1 to 5, then held (REF-1497, B-).

Roxzone. 5.1% and 5.2% in Worlds top 10 (REF-1519, C); 7.9% in Open men, 7.2% at the top decile and 9.1% at the bottom (REF-1504, C); 3:32 to 6:02 of the remainder in PRO men (our arithmetic; REF-1496). A first race cost about 2% more there in men and 1.4% in women, gone by the second (REF-1504, REF-1505). Nothing tests training it.

Late running is not read from early running. Across 61,541 records early splits left a late-run residual P90-P10 of 67.8 s and a final-run spread of 111.8 s (33.2 and 63.2 s in a second sample); all prior history still left 67.2 s (33.0 s) on run 8; the authors bar reading this as fatigue, a trait or a training threshold (REF-1502, B+, abstract).

The next run, on a first measure. Median +20 s after lunges, +14 s after the sled push, −4 s after the row; lunges were the largest in 43.4% of events, not all; the residual quartiles span zero for all but the row (REF-1498, C, preprint). One champion's runs after stations took 3:17-3:30 in a lab partial simulation and 3:28-3:42 in a race 10 days later: unlike order, course and work (REF-1507, C+, one man).

Pacing analogues, triathlon only. The first run kilometre at −5% of fresh 10-km pace gave a 2028-s 10 km against 2178 s at +5% and 2087 s at −10%; no self-paced arm; it follows a swim and a ride, and run 1 of a HYROX follows nothing (REF-1508, B, 10 men). So at most an inference for runs after stations, and a floor on caution, not a target.

Fresh running tests did not replicate running after an hour on the bike (REF-1518, B-, direction only). Riding 90% rather than 80% of FTP slowed men's 5-km run 35.1 s and still won 57.7 s on bike plus run (women 80.9 s): this tests cycling intensity, not first-kilometre pace (REF-1521, B+, abstract).

In long events more even pacing may help (REF-010, B+, narrative); a fast marathon start went with a slower finish (REF-820, B+).

Going Pro; the same load in women. Within athlete, the five loaded stations took about 34% longer in men (1,005 movers, REF-1504) and about 40% in women (4,216, REF-1505) against the unchanged segments, whatever the profile within ±2% per SD; both C, self-selected.

Pro women and Open men lift the same kilograms and ran alike (+0.3%), yet women took 33.3% longer on the sled push, 27.0% on the pull, 24.6% on the carry and 2.5% on lunges: populations, not people (REF-1505).

Doubles: one preprint pair. Partners match on ability (r = 0.394 men, 0.386 women); complementary profiles went with +2.5 percentile points per SD in men, none in women (−0.14); run-to-run variation fell 15% and 11.7% against the same athletes' singles (REF-1504, REF-1505, C). REF-901 holds Doubles and relay percentiles, unread.

Age, one race. Top 10 at 45-49 took about 10% longer than at 30-34, and at 55-59 29.5% (men) and 30.7% (women) longer: running 24.0% and 28.1%, stations 34.6% and 34.0%; wall balls fell most, ergs least (REF-1519, C). Fields thin with age, so selection differs.

Fuel and caffeine in shorter mixed work. Carbohydrate before or during: g = −0.130 (95% CI −0.652 to 0.391; 3 crossovers, n = 40). Caffeine: g = 0.371, not significant in the results although the abstract says significant, 0.144 after trim-and-fill. CrossFit trials, none described at HYROX length (REF-1517, B).