GPS Accuracy: What Multi-Band Actually Fixes

A modern GPS watch is trustworthy in open sky and unreliable everywhere the sky is blocked. Multi-band (dual frequency) GNSS is a genuine improvement, but it fixes one specific problem, reflected signals in forest, canyons and cities, and does nothing about the other reasons your distance looks wrong. If you run mostly on open roads and trails, accuracy is already a solved problem and spending more on it buys you very little.

What the watch is actually doing

Your watch listens for timing signals from satellites and works out where it must be for those signals to arrive when they did. The calculation is only as good as its assumption: that each signal travelled in a straight line to your wrist. With clear sky and satellites spread widely overhead, that holds, and the recorded track sits neatly on the road.

Two things break it. The first is simply not seeing enough satellites, which happens in a deep gully or a stand of conifers. The second, and the more damaging one, is multipath: the signal bounces off a rock wall, a wet canopy or a glass office block and arrives late, having travelled further than the straight line. The watch cannot tell it took a detour, so it trusts the delayed arrival and places you somewhere you were not. That is why bad traces are rarely just fuzzy. They are confidently wrong, sitting a street over or drifting through a building.

What to expect from your surroundings

Environment	What to expect from the track
Open road, moorland, beach, track	Reliable. Repeat runs of the same route agree closely.
Open forest, scattered trees	Mostly fine. Occasional wobble that inflates distance a little.
Dense wet canopy, conifer plantation	Noisy. Straights get zigzagged and distance usually creeps up.
Narrow canyon, gorge, cliff base	Poor. Few satellites plus reflections off rock, so the trace jumps.
Streets between tall buildings	Often the worst case. The track can sit on the wrong street.
Switchbacks, tight singletrack	Corners cut short, so distance often reads low.
First minute of a run	Unsettled while the fix stabilises. Wait for the ready signal.

What multi-band actually changes

Older watches listened on a single frequency. Multi-band models listen on two at once, and because the two frequencies are distorted differently on their way through the atmosphere and behave differently when reflected, the watch has more information to work out which signals to trust and which look like they took a detour. The practical result is that reflection-heavy environments improve most: city blocks, canyons, cliff traverses and heavy tree cover.

Where it does not help is just as important. Multi-band does not create satellites that are hidden behind a mountain, so a deep, narrow gorge is still hard. It does not fix a corner that was cut because your recording interval was too coarse. It does not improve elevation much, since altitude from satellites is inherently weaker than horizontal position and most good watches lean on a barometric altimeter instead. And it costs battery, which is the trade-off covered on battery life. On an open road run, multi-band and single-band traces are hard to tell apart.

If your running is forest, mountain or urban, the accuracy argument for a flagship is real, and it arrives bundled with mapping and battery that probably mattered more anyway.

Why the watch says you ran further than you did

Every recorded point carries a little error. The watch joins those points into a line and adds up the segments, and because the error scatters points to either side of your true path, that line is longer than the path you took. Noise adds distance. It is why a wooded run reads long, and why the effect grows with interference rather than with your speed.

The opposite happens too. On tight switchbacks, or with a coarse recording setting, the watch records a point before and after the bend but nothing in between, so it draws the chord instead of the arc and distance reads short. Most runs contain some of both.

Why two watches on the same run disagree

They received different signals through different antennas and made different decisions about which satellites to believe. Every manufacturer also filters the raw positions before showing them to you, and those filters are proprietary: some prioritise a clean-looking track, others follow the raw data closely. Two honest watches can produce two different numbers from the same run, and neither is lying. Compare a watch against itself over time.

Recording rate and smart recording

Most watches offer per-second recording or a smart mode that saves points only when direction or pace changes meaningfully. Smart recording saves battery and file size, and on straight roads costs you nothing. On twisting terrain it is the setting most likely to shave distance off your run, because corners are exactly what it decides not to record.

If your route twists, use per-second recording and accept the battery cost. If you are pacing a long event where the watch must survive to the finish, smart recording plus a lower GNSS mode is a sensible trade. Deciding which you are is most of the work in choosing a sports watch.

Measured distance beats trail distance

A certified track or a measured road course is a far better reference than the distance printed on a trailhead sign or listed on a route-sharing site. Trail distances are often rounded, inherited from an old survey, or taken from someone else's GPS trace with the noise baked in. When your watch and a trail sign disagree, the sign is not automatically right.

How to test your own watch honestly

Run several laps of an athletics track in lane one and compare, more than once. Then run a familiar loop repeatedly and look at how consistent the totals are, because repeatability matters more than any single result. Finally, open the map of a run through your worst local conditions and judge the shape of the trace rather than the number: if the line sits on the path, the watch is doing its job.

Common mistakes

Starting before the watch has a stable fix, then blaming the first mile. Treating a friend's different total as proof of a fault. Buying multi-band for open ground. Leaving smart recording on for technical trail and wondering where the distance went. Trusting instant pace in a forest, when a lap or rolling average is far steadier. And treating satellite elevation as truth when the barometric reading, drift and all, is usually the better guide.

Questions people ask

Should I pay extra for multi-band?

Only if you regularly run somewhere signals reflect: tall city blocks, rock walls, deep forest. On open ground it is a feature you will pay for in money and battery and never see. Work backwards from where you run, which is what the watch finder asks first.

Is my watch broken if a run reads long?

Almost certainly not, especially if it was under trees. Noise inflates distance by its nature. Suspect a fault only when a run on open ground, on a route you know, is repeatedly and substantially off.

Does phone assistance make the watch more accurate?

It makes it faster. Cached satellite data helps the watch find a fix quickly, but once locked on, accuracy comes from the receiver and the sky.

Why is my pace jumpy when distance looks fine?

Pace is calculated from small position changes over a short window, so it magnifies exactly the errors that mostly cancel out over a whole run. Use lap pace or average pace for decisions and treat instant pace as a rough hint. The same "is this measured or modelled" question applies to the rest of your screens, which is the subject of which metrics matter.