The Edmonton Half, Measured — Part 1: What Happened, and How We'll Actually Find Out

On the morning of Sunday, August 16, 2026, several thousand runners set off from the Edmonton Convention Centre for the 35th Servus Edmonton Marathon and its half. The half doubled as a national-level championship, so the field ran from first-timers up to Olympians. Within a few hours, a pattern showed up in the finish-line chatter and then in the local press: a lot of watches had read the half distinctly long — some by the better part of a kilometre.¹²

We have an unusual reason to care about this one precisely. Brown Family Sports is a running-coaching practice, our founder ran the Edmonton half that morning, and analysing GPS tracks against a plan is something we do every week. So rather than add one more watch screenshot to the pile, we're going to do the thing the situation actually calls for — measure it properly — and show our work.

What everyone agrees on

It's worth starting with the parts that aren't in dispute, because there are more of them than the argument suggests:

• The certified distance of a half marathon is 21.0975 km, and of the full, 42.195 km. Those numbers are fixed by World Athletics and are not what anyone is contesting.⁴
• Many runners' watches recorded the half meaningfully over that figure. That the watches read long is agreed; what it means is the whole question.¹
• A certified course is measured along the shortest possible route a runner could legally take. Anyone who doesn't hold perfect tangents covers more ground than the certificate says — by running, not by error.³
• The stakes are real and time-bound: Edmonton is a Boston qualifier, and a course that runs long inflates official times just as 2027 Boston registration approaches.⁶

The two accounts, stated fairly

An investigation that only quotes one side isn't one. Here are both, as close to their own words as we can get them.

Both of these can be sincere. A measurer can certify a course correctly and a race-day detour can add distance that the certificate never covered; a runner can read long and be reading mostly the ordinary over-measurement every watch produces. The disagreement is genuine, and it's exactly the kind that a big enough pile of honest data can resolve.

The winner's own account

Two days after the race, the man who won the half — Olympian Rory Linkletter — posted his own explanation, and it's a model of how to raise this without heat. He praises the race team, frames the problem as an honest mistake, and is clear it didn't cost him anything personally; his concern is for the runners who came in chasing a specific time. It "deserves an honest apology to ALL participants," he wrote.⁹

He also gives the most concrete mechanism anyone has offered. As he describes it: a construction detour added roughly 710 m of extra distance each way, which was meant to be cancelled out by moving the out-and-back turnaround about 690 m closer each way. But the turnaround appears to have been moved only about that much in total — roughly half of what was needed — leaving the course long by on the order of ~700 m. He says a crew will physically wheel-measure the turnaround to settle it. It's a claim to be verified, not yet a proven number — but it is checkable, which is exactly the right way to argue this.

One more reason his account matters: it lines up with the independent measurement. A ~700 m error is right where the runner GPS data is pointing — a single de-noised track in this piece read about 900 m long, and the wider sample we're gathering is clustering in the same range. When a course-geometry argument and a pile of watches converge on the same number from opposite directions, that's the kind of agreement a coincidence rarely produces.

Why one GPS watch can't settle it — and why we'll say so loudly

This is the part most of the online argument skips, so we'll put it first and in plain terms: GPS over-measures. Everyone's watch reads long, on every course. Satellite-position noise, tree cover, and tall buildings all add phantom distance, and a downtown river-valley course is close to the worst case for it. The typical honest over-read is about 0.5–1.5% — on a half, that's roughly 100–320 m before anything is actually wrong with the course.³⁸

So a long watch reading is the expected outcome and proves nothing by itself. The organizers' point that "GPS devices record distances differently" is correct as far as it goes. What it doesn't address is the size and the shape of what runners saw. Those only become visible in aggregate. Four things carry weight where one screenshot can't:

• the distribution across hundreds of tracks, not any one file;
• how far the median sits beyond the 0.5–1.5% you'd expect anyway;
• whether the recorded route geometry matches the certified map turn for turn — which catches a mis-set turnaround that a distance total alone would hide;
• median, spread, and sample size published together, so the claim can be attacked on the merits.⁸

A worked example: our own track, honestly

Here is what "measure it properly" looks like on a single file — ours — with every caveat attached. Our founder's half was recorded at one-second sampling, 5,913 GPS points. We ran it two ways: the raw point-to-point distance, and a noise-filtered distance that removes sub-jitter wander (segments too small to be real running), then compared both to the certified 21.0975 km.

The point of this example isn't the headline number — it's the decomposition. When we strip out the GPS jitter, the path is still about 906 m long. In other words, on this one track, only ~120 m of the excess is sensor noise; the rest is real ground the body actually covered. That's already more than the 0.5–1.5% a clean run should show. Our per-500 m breakdown also picks up the half's out-and-back turnaround cleanly, near the 11 km mark.

The method, and how you can attack it

Everything above is why the rest of this series is a data-collection project, not an argument. We've stood up two hardened upload endpoints — one for the half, one for the full — that take a runner's GPS track, measure it the same careful way for everyone, and record both the raw and the noise-filtered distance so we can't quietly publish only the flattering one.⁷ We also capture each track's sampling rate, because a one-second recording and a "smart-recording" watch are not directly comparable and pooling them naively would smear the result.

When Part 2 publishes, it will lead with the numbers that let you argue back: median over-read, the spread, the sample size, stratified by sampling rate — and a route-geometry overlay against the certified map. If the median lands inside the ordinary 0.5–1.5%, that is a finding, and we'll report it plainly: the watches were doing what watches do. If it lands well beyond, that's a different finding, and it will be just as visible. We're not here to win an argument; we're here to publish a measurement.

One last note on tone, because it matters. The people who organize the Servus Edmonton Marathon put on a 35-year-old event that thousands of runners love, and the measurer who certified the course is a professional doing careful work. Being wrong about a race-day detour, if that's what happened, is not the same as being dishonest. This series will hold the question to a high standard of evidence, extend the same fairness to the organizers that we'd want extended to us, and follow the data wherever it lands — including to "the course was fine" if that's where the median goes.

Next in the series — Part 2: The data. The distribution of every track we've collected, measured the same way, with median, spread, sample size, and a route-geometry overlay against the certified course. Add yours above and you're in it.

Sources

1. Madeleine Cummings, CBC News (Aug 17, 2026), "Organizer says Edmonton Marathon routes were accurate. Elite, amateur runners disagree" — the anchor report, carrying both the runners' accounts (including the half winner and a two-time Olympian) and the organizer / Athletics Canada responses.
2. Global News (Aug 17, 2026), "Runners question Edmonton Marathon course length" — broadcast segment: GPS data suggested the course may have been about a kilometre longer than advertised.
3. Marathon Handbook (Aug 2026), "Edmonton Marathon Defends Course; Runners Say It Ran Long" — the measurement analysis: certified courses are measured along the shortest possible route, the tolerance figures at issue, and the construction / turnaround details.
4. World Athletics — Servus Edmonton Half Marathon results and the certified reference distance (half 21.0975 km; full 42.195 km).
5. Athletics Canada — national governing body; road-race sanctioning and course-certification policy (including the 2026 requirement that a course altered year-to-year be re-measured and re-certified unless the change is slight).
6. Boston Athletic Association — Boston Marathon qualifying standards and the 2027 registration window; context for why an inflated official time is costly and time-sensitive.
7. Brown Family Sports — GPX intake for the Edmonton half and full: the browser-side measurement, the raw-plus-filtered distance recording, and the sampling-rate capture described above.
8. AIMS / World Athletics course-measurement resources — background on why hand-held GPS over-measures relative to a calibrated-bicycle certified measurement, and the typical magnitude of that over-read.
9. Rory Linkletter, OLY (@rory_linkletter), Instagram (Aug 18, 2026) — the half winner's statement on the course: an "honest mistake that deserves an honest apology to ALL participants," the detour-vs-turnaround explanation, and the plan to physically wheel-measure the turnaround.