Seventy four numbers didn't make the book.
Every figure in HYROX PREP was rated before it was printed. Seventy four failed and were cut. This is the standard they failed, and the reason each one failed.
Most of what is written about this race is confident and unsourced.
That isn't a complaint about anyone in particular. It's the ordinary condition of fitness writing. A number gets published, it gets quoted, the quote gets quoted, and four links later it's a fact with no parent. It's usually round. It's usually about something nobody has actually measured.
HYROX PREP runs to 12,178 words and it's free, so the only thing it can be judged on is whether the numbers in it are real. Seventy four of them were not. Those got cut before the book shipped rather than after somebody caught them.
I'd rather show you the cutting room floor than ask you to take the rest on faith.
Every number carries a label.
Before a figure goes into anything I publish it gets one of six labels. The label decides how the number is allowed to be written.
| Label | What it means | How it may be printed |
|---|---|---|
| PRIMARY | The official rulebook, or an official HYROX publication | Stated plainly |
| STRONG | Peer reviewed, or a large public dataset with a named, checkable sample | Stated plainly |
| MODERATE | A large dataset, self published, internally consistent, no methodology given | Attributed, never asserted |
| WEAK | One coach, one blog, one small sample, or a method the author disclaims | Not printed as fact |
| ARITHMETIC | Derived here, self verifying | Printed as arithmetic, never as a finding |
| ABSENCE | Verified that no defensible public number exists | The absence itself is printed |
The two doing the real work are the last two.
ARITHMETIC keeps me honest about the difference between a measurement and a division. If a number is eighty meters divided by your jump length, it's arithmetic, and dressing it up as research would be a lie about where it came from.
ABSENCE is the one almost nobody publishes. Sometimes the honest finding is that the number doesn't exist. That's worth more to you than a plausible invention, because you can act on a known gap and you can't act on a wrong answer you trust.
The six ways a number failed.
Grouped by how it broke, because the pattern is the useful part. Learn these and you can audit anybody, including me.
An average wearing a median's clothes
The most common failure by a distance. A figure gets reported as the typical time when it's the mean, and means in this sport run several percent above medians because slow finishers have a long tail.
One table would have printed a median sled pull of 5:41. The actual median is 5:18. Another would have printed station medians summing to 40:20 when the median column sums to 37:51. Both are the kind of error a reader with a calculator catches in about a minute, which is exactly why they can't be printed.
Numbers that exist in no source
Some figures survive purely by repetition. Chase them to the bottom and there's nothing there.
Cut for this reason: a claimed cost in seconds for shoe slip on the sleds, a per stop momentum cost on the sled push, and a Roxzone time penalty for walking. Also a no rep rate for wall balls, which was checked against the rulebook, the sport's official science report, and nine independent searches. There is no such published rate. If you see one quoted as a percentage, it was invented.
Precision the sample can't support
A number with a decimal point in it implies a measurement that earns the decimal.
A widely repeated intensity figure reads 79.5 percent. The underlying study has eleven participants and a standard deviation of twenty one points. At that spread, 79.5 isn't precision, it's decoration. The book says "most of the race" instead, which is what the data actually supports.
Same reason killed a Roxzone quartile gap stated as exactly 2:16, when the source itself only committed to "over 7:42." That makes 7:42 a floor, and the subtraction a fiction.
One event, one gym, one athlete
A single race, a single finish bracket, or one person's Tuesday is not a population.
Cut on this rule: sandbag lunge step targets traceable to one event and one six minute finish bracket, and a sled loading adjustment from one person in one gym who said in the same breath that it varies gym to gym. A sled push pause interval survived, but only because it changed jobs. It was published by an athlete who wrote explicitly that it was the best approach for him at his loads, so it appears in the book as a disagreement between athletes rather than as an instruction.
Internally impossible
The most instructive failure, because you can catch it with nothing but the source itself.
One widely shared pacing table prescribes the run splits for a sub ninety minute finish. Take its own paces, add published station and transition averages, and the total lands under eighty minutes. The table can't produce the race it promises. Nothing on that page could be used, including the parts that looked fine.
Right number, wrong claim
The number is real. The sentence built on it is not.
A body of research on running economy after eccentric damage was going to be cited as the mechanism behind a pacing rule. That literature measures economy twenty four to ninety six hours after downhill running. The book is talking about thirty seconds after a hundred meters of lunges. Same subject, wrong timescale, so the citation was doing rhetorical work rather than evidentiary work.
Three of them were mine.
The seventy four weren't all other people's numbers. Three of them were the argument I started with, and the research killed them.
I planned to write that the Roxzone is worth three minutes or more at zero fitness cost. It isn't. The gap between fast and slow finishers in there is partly a fitness gap, which the source I was leaning on concedes a few paragraphs after making the claim. The analyst's own estimate is one to two minutes. The book says one to two minutes, and says why.
I planned to write that athletes slow eighteen to twenty two percent by Run 8. That figure has no methodology, no sample size, and sits on a page whose own split table contradicts the story it tells. Cut entirely.
I planned to write that athletes open about ten seconds per kilometer too fast. The real published answer is worse and far more useful: measured against their own eight run average, age group men open somewhere between seventy three and ninety four seconds per kilometer too fast. The true number is roughly eight times the one I was going to print, and it turned into the most actionable fact in the book.
If a standard only ever catches other people's errors, it isn't a standard. It's a marketing position.
Four things nobody has measured.
These are the ABSENCE findings. Each was searched for and confirmed missing from the public record. If someone quotes you a number for any of them, ask where it came from.
- How your 5k time predicts your HYROX pace. No public dataset pairs individual 5k times with individual HYROX splits. The pacing target in the book is a starting point you correct against your own benchmarks, and it's labeled as one.
- How many real sled sessions a first timer needs. No defensible number exists. The book gives a rule instead: touch a real sled at least once before race day, and touch it tired.
- The no rep rate on wall balls. Not published anywhere.
- What a wall ball converts to in dumbbell thrusters. No load equivalence exists in the public record. Worth adding that the absence itself rests on a thin base, so I wouldn't let it harden into a biomechanical fact either.
Every one of those is a place where the honest move was to print the gap and give you a rule you can run without the number.
What survived.
The finding the whole book is built on cleared the bar, and you can check it in about two minutes.
Across 395,452 solo results, the four machines a garage athlete can't own matter least, in two different ways. The two sleds cost the median athlete under a tenth of his race by time. The two ergs separate the field less than any other station in it: between the fastest tenth of the field and the slowest tenth, the SkiErg accounts for 0:57 and the row for 1:12.
Wall balls account for 6:28. The widest spread in the sport is a wall ball, the second widest is a burpee broad jump, and three of the four widest need nothing but a floor, a wall and a pair of dumbbells.
Station times and spreads are from HyroxDataLab, Seasons 7 and 8. That's a large, self published dataset with a stated sample and no published methodology, so it's labeled MODERATE, and it's attributed everywhere it appears. It is not official HYROX data, and the book says so four times.
None of this makes the book right.
It makes it checkable, which is the only honest thing a free plan can offer.
If you find a number in it that fails one of these tests, email me. I'll fix it and say what changed.
Read HYROX PREP.