Method
Most of this site is arithmetic on other people's timing data. This page says exactly which parts are measured, which parts are estimated, how wrong the estimates get, and what we refuse to guess about at all.
Badwater, Hardrock and Western States hand runners no GPS tracker. Nothing to carry, nothing to charge at mile 90. So at the races people most want to follow, every tracking product on the market shows nothing at all.
What those races do publish is a time at each checkpoint. That is the entire input. We take the official splits and rebuild the race in between them.
A runner's dot moves because we assume they keep travelling at the pace their own splits imply, adjusted for the terrain ahead. That is a model, not a sighting. The moment they cross the next mat, the guess is replaced by a fact.
We measured this rather than asserting it. Across 11,084 archived runner-hours we compared where the model said someone was against where their next confirmed split proved they had been. Error grows with time since their last checkpoint:
| Time since last checkpoint | Typical error | 9 times in 10, within |
|---|---|---|
| 30 minutes | 0.17 mi | 0.50 mi |
| 1 hour | 0.35 mi | 0.98 mi |
| 2 hours | 0.70 mi | 2.01 mi |
| 4 hours | 1.38 mi | 3.97 mi |
| 8 hours | 2.39 mi | 6.64 mi |
Measured across archived races. Loaded live from the same file the tracker itself uses.
This is why a followed runner's card tells you when they were last confirmed and how long ago that was. A dot four hours old is a worse dot, and the page should say so rather than look equally confident at every moment.
The next-checkpoint estimate is unbiased overall, but individual legs vary a great deal: runners stop, eat, sleep, and fall apart. Measured across the same archive, the true leg time lands between 0.82x and 1.27x of the prediction about 80% of the time.
So an hour-long estimate honestly means somewhere around 49 to 76 minutes, and we show it that way. A crew plans around a window. A single confident minute is the thing that makes people stop trusting a tracker.
We got this wrong once, publicly, and it changed how the site works. When a race's advertised cutoff passed, our software marked every runner the timing crew had not yet posted as "did not finish". At one race that was 108 people who had not quit at all. Their results simply had not been published yet.
Now only the timing company can declare a DNF. When someone has splits and no posted result, we label them no official result and say plainly that we do not know whether they finished. A race does not get archived while results are still outstanding, because archiving is itself a claim that the result is final.
Races get cut short for wildfire smoke, lightning and heat, and courses get rerouted when land closes. When that happens the race page carries a notice in the organiser's own framing with a link to their announcement, and the runners who were brought off the course are not called DNFs. They did not stop by choice.
Every race here is timed by somebody else. Their timing partner is credited on the race page, and the race's own results remain the official record. If our numbers and theirs ever disagree, theirs are right and ours are the bug.
Race names and marks belong to their organisers. We are an independent project and are not affiliated with any race.
Tell us and we will fix it and say what changed: contact us. Corrections to a person's own result are the highest priority thing we do.
Back to the races · The all-time boards · Data for broadcasters