Basic, Strong, Snap to Road or Reference: Which GPS Smoothing Method to Use

By Roy Lachica on October 22, 2026. Last updated October 22, 2026.

Smoothing a GPS track is not one operation, it is four different ones with genuinely different consequences, and picking the wrong one is how people end up with a beautifully smooth track that is 800 metres shorter than the route they ran. This is a guide to choosing, with what each method does to your data and what it costs you.

The four methods side by side

Method	What it does	Best for	Effect on distance
Basic	Finds track points that produce unrealistic speeds or tiny back-and-forth movements and repositions them. Leaves the rest of the track alone.	A handful of scattered bad points in an otherwise sound recording.	Shortens slightly.
Snap to road	Takes anchor points roughly every 500 metres, routes between them on open-source road and trail data, and repositions the points in between along that geometry. Timestamps are preserved.	Road and path activities where the shape is right but the track sits off the road.	Changes to match the real road length, up or down.
Strong	Replaces every coordinate with the average of its 15 nearest neighbours. An aggressive moving average applied to the whole track.	Recordings where the signal was poor from start to finish.	Shortens, sometimes substantially.
Smooth toward reference	Compares your track against a reference file and pulls points that are more than 10 metres off back toward it, snaps speed spikes onto the reference line, then removes residual jitter.	A known course you have a clean file for, races especially.	Converges toward the reference route length.

Basic

This is the default and the one to reach for first. It is selective: it looks for the points that are doing damage, the ones that imply impossible speeds or micro loops, and moves those. Everything else stays exactly where your device put it. Because it does not touch the bulk of the track, it is the method least likely to surprise you, and the one that preserves the character of the recording. If your file has a few bad moments rather than a bad hour, stop here. For the same problem viewed from the spike side, see removing GPS spikes.

Snap to road

Snap to road is categorically different from the others: it does not average anything, it moves your points onto a real road or trail using routing data. That makes it the strongest option when your track is the right shape in the wrong place, drifting into gardens, across buildings, or on the wrong side of the street. It is also the method most likely to restore segment matching, because matching cares about where the line actually runs.

The trade-offs are real. It needs an internet connection, since it queries routing data. It assumes you were on a mapped road or path, so it is wrong for cross-country running, open fell, beaches, ski touring and anything off the network, where it may route you along a road you never used. And it changes your measured distance to the length of the routed line. Full details in the snap to road guide.

Strong

Strong smoothing averages every point against its 15 nearest neighbours. It is the right answer for a track that is jittery from beginning to end, the kind of recording you get from a phone in a pocket under heavy tree cover, and the kind that produces saw-pattern tracks. It will produce a visibly clean line.

It is also the bluntest instrument available. Averaging rounds off tight corners, so a track with switchbacks, hairpins or sharp turns loses detail that was genuinely there. And because jitter adds length, removing all of it reduces your total distance, sometimes by more than you expect. Use it when the alternative is an unusable file, not as a default.

Smooth toward reference

If you have a clean file of the same course, a race organiser route or a previous recording of the same loop, this is the most accurate option available, because it is the only one with outside information about where you actually were. Points that sit more than 10 metres off the reference are pulled back toward it, speed spikes are snapped onto the reference line, and a final pass clears any remaining jitter.

It needs a reference file that genuinely matches your route, GPX, FIT or TCX. A reference for a different course, or a different lap configuration, will drag your track somewhere it never went. See repairing a workout using a reference GPX file.

Choosing in practice

A few spikes, the rest is fine, use Basic, or Fix GPS issues.
Right shape, wrong position, on roads or paths, use Snap to road.
Jittery everywhere, no reference available, use Strong, and check the distance afterwards.
You have a clean file of the same course, use Smooth toward reference, it beats all of the above.
The track is in completely the wrong place, smoothing is not the tool, see manually fixing a track that drifted far off course.
Points are missing entirely, smoothing cannot invent them, see restoring missing GPS data.

Scenario: a trail marathon recorded under dense forest. The track wanders five to fifteen metres off the trail almost everywhere, there are three large spikes, and the distance reads 43.8 km for a 42.2 km course.
Workflow:

Start with Basic to deal with the three spikes without touching the rest.
Check the distance, if it is still long because of all-over jitter, consider Strong, but expect the corners on the switchback descent to soften.
Better, if the organiser published a course GPX, use Smooth toward reference instead, which fixes position and distance at once without flattening the corners.
Snap to road is the wrong choice here, the trail is probably not fully mapped and you risk being routed onto a forest road.

What smoothing always costs

Every smoothing method changes your measured distance, and in most cases shortens it, because GPS noise inflates distance by adding wiggle. That is worth saying plainly: a smoothed track usually reports less distance than the raw one, and the smoothed number is normally the more honest of the two. If the distance is what you care about, read why your activity distance is wrong and how to change the distance of a workout before you decide how much smoothing you want. And some files are beyond repair, which is covered in why some workouts cannot be fixed.

Frequently asked questions

Can I apply more than one method?

You can run operations in sequence, and Basic followed by a reference pass is a reasonable combination. Stacking Strong on top of other smoothing is rarely a good idea, you lose detail twice.

Does smoothing change my heart rate or timestamps?

No. These methods move coordinates. Timestamps are preserved, including by snap to road, and sensor channels are untouched. Pace and speed change only because the distances between points change.

Which method is least likely to look edited?

Basic, because it leaves most of the track exactly as recorded. Heavy-handed smoothing is one of the things that makes a file look manipulated, which matters if you care about activities being flagged.

Why did my distance get longer after snap to road?

Because the routed road line is genuinely longer than the drifting shortcut your track was taking. Snap to road matches the real route rather than minimising length.

Not sure which one your file needs? Upload it to GpxFix, look at the map and the speed chart, and start with the least invasive method that could explain what you see. Related reading: understanding GPS spike zones and resampling and simplifying a GPX file.