In August 2026 the runners of the Edmonton marathon covered not 42.195 km but roughly 674 metres more. The reason is mundane: a detour around building works that was not compensated by shortening another section. Some runners' watches showed around 800 extra metres.

What followed was painful. The organisers apologised but refused to give the money back. And the Boston Athletic Association answered with a formulation that leaves no options: adjusted results are not accepted if the distance was longer or shorter than 42.195 km — it is a long-standing policy, and there will be no exceptions.

The cost of this is easy to work out. At the pace built into the qualifying standards, an extra 674.7 metres costs around 2 minutes 48 seconds for men aged 18–34 and around 3 minutes 26 seconds for women aged 40–44. In the selection for Boston 2026, 8,887 people who had met the standard were cut — the effective bar came in 4:34 faster than the official one. So three extra minutes is not “nearly made it”, it is missing out by a year.

The story is a good one because it forces a question that normally bothers nobody: where does the confidence that a course is exactly 42.195 come from at all?

The Jones counter

Certified courses are measured neither with a tape measure, nor with a measuring wheel, nor with GPS. They are measured with a bicycle fitted with a Jones counter — a mechanical gear drive on the front wheel hub that counts fractions of a revolution at high resolution.

The logic is this. The absolute circumference of the wheel is unknown and constantly drifting: tyre pressure, temperature, the measurer's weight. So the counter is not converted into metres directly; instead it is calibrated on a reference section.

The calibration course is straight, level, asphalted, usually 300–400 metres, measured with a steel tape.
The measurer rides it at least four times before measuring the course and at least four times after.
The “after” calibration is mandatory, because over several hours the tyre will lose pressure and part of the measurement will turn out to have been made with a different wheel.

The shortest possible route

The second principle is less obvious, and it explains half of the misunderstandings with watches.

A course is measured along the shortest possible route — SPR. The measurer rides strictly along the inside of every bend, cuts the tangents, and goes along the kerb wherever that is physically permitted. Nobody runs like that in an actual race.

Every course is measured twice, independently, and the two results must agree to within 0.08%. Of the two, the larger is taken.

One metre per kilometre

And finally the third principle — the short course prevention factor.

One metre per kilometre is added to every measurement, i.e. a multiplier of 1.001. A marathon is laid out at measurement as 42,237 metres, not 42,195.

The point is the asymmetry of the error. A course longer than advertised is annoying. A course that is short is a catastrophe: records are annulled, qualifications are cancelled, the whole results sheet is devalued. So a margin in favour of “slightly longer” is deliberately built into the system.

From which follows a conclusion worth accepting once and never arguing about again: any certified course is knowingly slightly longer than its nominal distance. By about 42 metres in a marathon. That is not an error, it is by design.

Why your watch always shows more

Now it is clear why a Garmin draws 42.4–42.7 km at a marathon finish, and why that does not mean the course was measured badly.

Three things add up.

The 1.001 margin — plus 42 metres.

You do not run the SPR. Overtaking, aid stations, a wide arc through a bend, an attempt to get away from the crowd. A metre or two extra on twenty bends adds up to one or two hundred metres, and on a crowded city marathon to more.

The GPS error itself. According to measurements at the Two Oceans ultramarathon (56 km, 255 runners, eight device categories), the median relative error of the watches ranged from 0.6 ± 0.3% to 1.9 ± 1.5%. Over the marathon distance, 1% is 420 metres. The error grows with tall buildings, bridges, tunnels, dense foliage and frequent turns: the receiver either cuts the arc of a bend into straight segments or, conversely, adds “noise” while you are standing still.

This is exactly why GPS is no good for certification — not because the watches are bad, but because the required accuracy of 0.08% is unattainable for them.

What a runner should do

Build your splits on the kilometre markers on the course, not on your watch. The markers are placed along the measured distance; your watch runs on its own version of reality. If by the thirtieth kilometre your watch is 300 metres “ahead” of the markings, you are running slightly slower than you think, and your pacing plan will be wrong.
Set the auto-lap to manual, or at least check against the markers. An automatic GPS kilometre shifts your pace by 3–5 seconds from the real one.
Run the tangents. It is the only legal way to get close to 42.195. On a twisting city course the saving reaches a couple of hundred metres — that is, a minute.
Check the course certificate if you are travelling for a qualifying time. An AIMS/World Athletics or national federation certificate is what separates a result that will be accepted from a result that will remain a pleasant memory.
Do not expect a recalculation. The Edmonton case spelled out the position of the majors' organisers in plain language: only a correctly measured distance is subject to recalculation.

The bottom line

The 2026 Edmonton marathon came out 674.4 metres longer because of an uncompensated detour; Boston refused to accept adjusted results.
Courses are measured with a bicycle fitted with a Jones counter, calibrated on a 300–400 m reference section at least four times before and four times after the measurement.
They are measured along the shortest possible route, twice independently, with a discrepancy of no more than 0.08%; the larger result is the one that counts.
1 metre per kilometre is added to everything (a factor of 1.001) — a marathon is measured out as 42,237 m.
A certified course is therefore always slightly longer than nominal, by about 42 metres.
The error of sports GPS watches, from measurements over 56 km, runs from 0.6% to 1.9%, i.e. up to 800 metres over a marathon.
Your reference on the course is the kilometre markers, not the reading on your watch.

Sources: Marathon Handbook, “Edmonton Marathon Defends Course Runners Say Ran Long”, 2026 · Boston Athletic Association, comment on its policy for accepting results · World Athletics / AIMS, course measurement rules and the short course prevention factor · Johansson R.E., Adolph S.T., Swart J., Lambert M.I. “Accuracy of GPS sport watches in measuring distance in an ultramarathon running race”. International Journal of Sports Science & Coaching, 2020. DOI: 10.1177/1747954119899880