Methodology

Where the data comes from

Timing mats and finish times, as the race organizer published them. Nothing here is collected from runners, bought, or sourced from a social network. If a number is on this site, the organizer had already published it.

The archive currently holds 71 races, marathons and half marathons, and 1,401,460 finishers.

Race maps are drawn from OpenStreetMap: the route itself comes from the organizer's own GPX, and the city around it — the coastline, the rivers and the arterial roads — comes from OpenStreetMap contributors, under the ODbL licence. Like everything else here, that data is prepared once and written to a file: your device never talks to a map server, and no third party learns which race you opened.

Every aggregate metric is computed once, when a race is added, and written to a file. No query about you is ever run against a database. Your own analysis is computed on your device, from the file the page already loaded.

How hitting the wall is defined

Your cruising pace is your measured pace between km 5 and km 20. We take it from there because the opening stretch is crowded, and a whole-race average already contains the slowdown we are trying to measure.

Hitting the wall means running at least 5 km of the second half at 25% or more above that cruising pace, and not coming back. The rule is Smyth's (2021), from an analysis of over 4 million marathon performances.

The 25% is not a number we chose. When the body switches from carbohydrate to fat as fuel, sustainable output falls by 15 to 25%, because fat yields less energy per unit of oxygen (Hearris et al., 2018). The threshold sits at the top of that range so the measure under-calls rather than over-calls.

Why you are never compared with the whole field

Slowdown tracks ability. Faster runners fade less, almost everywhere, in almost every race. So a single field-wide median would measure you against a distribution you were never part of, and would make a slower field look like a harder race.

Every comparison on this site is made inside a band: runners of your sex who finished this same race near your time. Where a band holds too few runners to be meaningful, it is grouped with other bands. Our site tells you when that happens.

The small-group floor

No group of fewer than 20 runners is ever shown. Below that, an aggregate stops describing a group and starts describing individuals. Anyone who knows two people in a category of three can work backwards to the third.

A suppressed group is drawn as an explained gap, never hidden.

Race-day conditions

Hourly temperature, humidity and wind for the race location, from open-meteo's historical archive, formatted in the race's own time zone. Your own exposure is worked out from your start time and your finish time against that curve. It is not a row stored anywhere.

Humidity is always shown paired with the temperature it was measured at, never as a range and never as a trend. Relative humidity falls through a warming morning while conditions get harder, so a humidity range invites exactly the wrong conclusion.

The ideal temperature for a marathon sits between 4°C and 10°C. Above that, pace falls away, and it falls further the longer you are out there. Slower runners meet the hottest hours of the morning (El Helou et al., 2012, across 1.8 million runners; Ely et al., 2007 and 2008). Conditions are described, never used to restate your time: a weather-adjusted finish time is a counterfactual, and there is no consensus among the models that compute one.

Age grading

The WMA (World Masters Athletics) road running tables, the same standard most official results sites use. It is a reference standard built from the best times in the world at each age, not a scientific study. What the science shows is that performance declines with age in a predictable way (Tanaka and Seals, 2008; Lepers and Cattagni, 2012).

In most races we hold your age band and not your exact age, so an age-graded time usually comes as a range. The figure your organizer publishes, from your exact age, falls inside it.

The names, and what each one demands

At the end of your analysis you may be given a name: Clockwork, Ageless, The Phoenix. It is not a horoscope and not decoration: each has a written rule, and each arrives with the number that backs it. Where the number does not exist for your race, the sentence goes, and the name goes with it.

Almost every rule compares you against runners of your sex who finished this race near your time, never against the whole field. Pace falls away more for slower runners, so a single field-wide yardstick would hand every name to the fast.

One exception matters: where a name asserts something in absolute terms, a relative yardstick is not enough. Clockwork asks that you be among the steadiest 25% of your band AND that you never varied more than 4% around your own average pace. Without that ceiling the steadiest quarter of a slow band would include runners swinging by 28s per km. Calling that clockwork is a claim no coach would accept.

One number is estimated, and the page says so. The count of younger runners you finished ahead of comes from the finish-time distribution by sex and age band that the race already publishes, interpolated inside wide bins. That is why the text says “about” and rounds: for the runners who receive the name, the estimate sits within 1% of the exact count on average.

And some runners get no name at all. That is an answer rather than a failure: about a quarter of a field clears none of the thresholds, and inventing a generic name to fill the gap would cheapen every other one.

What this cannot tell you

We measure your splits, not your body. We can see the shape of a collapse, not its cause. Heat, hills, fuelling, a bad night, a toilet stop and easing off on purpose all look identical here.

Roughly 3 in 10 runners who hit the wall never slow enough for a timing mat to notice. If you felt it and we didn't find it, you are not wrong. (Buman et al., 2008, from 315 runners' own accounts.)

We measure the collapse in fixed 5 km blocks. A bad patch falling across two blocks can be missed, about 1 runner in 30.

Every comparison here is against people who finished the race. The results lists we read rarely publish the splits of those who stopped, so nobody who dropped out is in any denominator on this site.

The half marathon: what can be said, and what cannot

A half is measured on the same yardstick as the marathon, at half the scale: 2.5 km blocks instead of 5, cruising pace between km 2.5 and km 10, and the slowdown compared against runners of your sex who finished near your time, exactly as in the marathon. Conditions, the people you passed, your age-group position, the climb and your pace by stretch all work the same way.

What does not exist is the wall. Over 21 km glycogen lasts the whole race, so a slowdown here is pacing, not the fuel switch the 25% rule was built to detect. That is why a half's analysis never issues a wall verdict, never awards The Phoenix, which requires a wall and a way back from it, and asks a different question: not did you break, but did you hold.

In a half, running the second half faster is common rather than rare: in several races it is more than a third of the field. The analysis shows your race's share instead of treating the feat as an exception.

Half-marathon age grading uses the distance's own table, from the same source and the same year as the marathon table (WMA/USATF, Alan Jones, 2025). A half time is never converted into a marathon equivalent.

Many halves have few mats, sometimes only 10 km and the finish. Where there is no mat at halfway, the 10.5 km split is estimated from its neighbours and the page says so; where there is no intermediate mat at all, the cards that depend on one are withheld rather than guessed. Where there is no mat before 10 km, the cruising pace is the average of the first 10 km, start included.

This site investigates no one

Race Auditor's course-cutting analysis does not run here. None of its verdicts are held on this site and none are shown. No runner carries a mark, there is no list of suspects, and there is no count of how many were flagged at a race.

The wall analysis measures pace, and pace is not an accusation. Fading in the second half is the most ordinary thing that happens in a marathon. The method on this page cannot detect a cut course, does not try, and will never put that on your result.

Your lookup does not feed an investigation. Your bib is sent once to find your row and is not stored. What remains is daily totals, and a total points at nobody.

The investigations do exist, and they are published separately, case by case, as an editorial decision. They are never a computed field on a race page or on a runner's result.

What this site holds about you

No names. Not hidden, not hashed. Nothing Race Auditor holds has a name column at all, on this site or behind it, so there is nothing here to greet you by and nothing to leak.

No minors. Runners under 18 are removed before the data is ever loaded. Where an age band cannot be established, the runner is left out.

Your bib number is sent once, to find your row, and is never stored in the site's usage counts, never displayed on the page, never put in a link, and never printed on a share card. Your finish time is shown back to you instead, because you typed it and it confirms we found the right race.

No one's race analysis will ever appear in a search for their name.

Removing your data

There is no button here to delete your data, and adding one would be theater: nothing on this site identifies you, so a button would delete on the strength of a bib and a finish time, both of them public. There is an address instead, with a person reading it. The privacy page says what to send.