What is a hysteresis threshold?

The minimum sustained rise before a climb counts. Below it, an upward wobble is treated as noise rather than climbing.

Why not just add up every rise?

Because altitude readings are noisy. Standing at a traffic light for two minutes produces dozens of tiny rises, and adding all of them gives hundreds of metres of climbing without moving.

Is there a correct threshold?

No. There is a defensible range — most services sit between 2 and 10 metres — and the answer moves with the choice. That is why two honest services disagree.

What does a route print use?

The number in your file, when there is one. A figure that depends on an invisible setting is a poor thing to print.

Field notes 5 MIN READ

What elevation gain actually counts

A single hidden number decides whether your ride climbed 462 metres or 1 505. Here is what a threshold is and why every service picks a different one.

Words: Epic Efforts · August 2026

Two services, one ride, two totals. The reason is a number neither of them shows you.

The noise problem

An altitude series is never clean. A barometric altimeter responds to a door opening. A GPS altitude wanders by several metres while you stand still.

Add up every upward step in that series and a four-hour ride accumulates enormous phantom climbing — most of it recorded at traffic lights.

The fix, and its cost

Every implementation applies a threshold: a rise only counts once it has been sustained. Go up two metres and back down, and nothing is added. Go up two metres and keep going, and the climb starts counting.

This is usually called hysteresis, and it works. It also means the total is now a function of a setting.

Here is the same real ride computed three ways:

| Threshold | Reported gain | |---|---| | 0 m — count everything | 1 505 m | | 3 m | 1 140 m | | 20 m | 462 m |

The ride did not change. Only the definition did.

Drift, before any of that

There is a step before the threshold. A barometric series drifts over hours as weather moves through — the pressure falls, the altimeter reads it as climbing.

The correction is to fit a straight line through the whole series and subtract it, so the start and end sit at the same level unless you genuinely finished higher. The shape of every hill survives; the slow sag across the afternoon does not.

Get that wrong and you flatten the terrain. An early attempt of ours used a moving average instead of a straight-line fit and ate a real 80 m ridge, reporting it as 29 m. The mistake was that a moving average removes anything shorter than its window, and a mountain is shorter than an afternoon.

Which is why the file wins

If your activity states a gain, that number goes on the print.

We can compute one, and do when nothing else exists. But a figure that swings by a factor of three depending on a setting nobody can see does not belong on a wall next to a number you told people at the top of the climb.

Questions people actually ask

What is a hysteresis threshold?: The minimum sustained rise before a climb counts. Below it, an upward wobble is treated as noise rather than climbing.
Why not just add up every rise?: Because altitude readings are noisy. Standing at a traffic light for two minutes produces dozens of tiny rises, and adding all of them gives hundreds of metres of climbing without moving.
Is there a correct threshold?: No. There is a defensible range — most services sit between 2 and 10 metres — and the answer moves with the choice. That is why two honest services disagree.
What does a route print use?: The number in your file, when there is one. A figure that depends on an invisible setting is a poor thing to print.

YOURS, FROM A FILE, IN 90 SECONDS

One effort. One sheet of paper.

Bring the file and every choice is yours: treatment, colour, which numbers come along. Museum-grade 250 gsm matte, printed in Denmark, shipped in a tube with a nail taped inside.

Print it. FROM 349 kr

More from the journal

All articles →