Where the results come from
Every result is read from the official HYROX results service and stored as published. Nothing is estimated, and no time is ever invented to fill a gap — where a value is missing you will see a dash rather than a number.
Results are published provisional and settle over the following days. An unofficial time can still change. Each race page says which state it is holding, and provisional races are re-checked until they settle.
Dates
The results service publishes no event dates at all — not on any page, not in its headers. Dates here come from a public season calendar and are accepted only when the order the results service itself lists that season in agrees with them. A date more than two months from where that ordering places it is refused rather than shown.
That check exists because it caught real errors: a calendar will happily give you a city's next edition when you asked about a previous one, which is how six events were once dated five to seven months wrong. A few events still carry no date. They say so.
Places on a leaderboard
The place shown on an event's leaderboard is ours, not the source's. A HYROX event runs the same division across several race days, and the results service ranks each day separately — so its 1st place is the fastest of one day, not of the event.
The boards here rank by finish time across every race day of the event, within one division and one gender. Equal times share a place, and the next place skips accordingly. A DNF holds no place at all and sorts last: there is no time to rank it by, and a nil time is worth nothing at the top of a board.
Ranks, and what they can be compared to
The rank printed on a race page — 5th of 277 — is the source's own, and it
counts one race day and one gender. Two such ranks from different fields compare
nothing at all: on real data, 5th at one event has been ten seconds slower than 31st
at another. Where a comparison cannot establish that both sides raced the same field,
it shows no ranks rather than a misleading pair.
The denominator is counted from the results held rather than taken from the source's own field size, which is frequently exactly double the real number of finishers.
Percentiles
A percentile says what share of the field a time beat. It is computed from the distribution of every finisher in the same contest and the same gender, and is only built where that cohort has at least five people in it — below that the figure would describe a handful of individuals rather than a field.
Roxzone
Roxzone is the time spent moving between the running lane and the stations. Where the results service publishes it, that value is shown. Where it does not, it is derived as the finish time minus the runs minus the stations, and the page says the value is derived.
A derived roxzone is a few seconds out, because every component it is built from is published rounded to the second. It is never shown as a negative: the source occasionally publishes splits that add up to more than the finish time, and no roxzone at all is more honest than an impossible one.
Run laps
The published best run lap excludes the first run. Running 1 is timed from the start to the first timing mat and is systematically quicker than a full lap, so it is not comparable to the other seven and is never presented as though it were.
Comparing two races
The point of putting two races side by side is to separate you were slower from the race was slower. The field-adjusted column does that by asking how everyone who raced both events moved between them: the median change per segment, across that shared group, within the same division and the same gender.
It is a median rather than an average because sled times have brutal outliers. It is hidden entirely below fifteen shared athletes, rather than shown with a caveat — a number built on four people is noise wearing a decimal point. And it bundles two things it cannot separate: how hard the race was, and how much the field trained in between. Where the gap between the two events is known, it is stated alongside.
Which athlete is which
The results service publishes names, not identities — the same person appears at three events as three unrelated rows. Linking them is guesswork done carefully: names are matched on their sorted parts so Löber, Samuel and Samuel Löber meet, with gender, nationality and age bracket used to confirm or reject.
The bias is deliberately towards splitting rather than merging. A false split costs one data point. A false merge invents an improvement that never happened — which is exactly the number the comparison exists to report. Two people with the same name in the same field are always treated as two people, because nobody races the same contest twice.
What does not add up, and why
Run total, stations and roxzone will not sum exactly to the finish time. Each is published rounded to the second, so the parts drift a few seconds from the whole. That gap is the source's rounding, not an error here, and nothing is quietly adjusted to make the arithmetic look tidier than it is.
Corrections
If a time is wrong, it is wrong at the results service, and it is fixed here by being fixed there — the next re-check picks it up. If something on this site is wrong in a way the source is not, that is a bug and worth reporting: mail@ftiersch.de.