FAQs

When you look at the same route using different devices or software, the elevation readings vary significantly. There’s a mathematical reason for this. While one device’s altimeter may be more accurate than another’s, this doesn’t matter as much as you might think. Whether the device is currently calibrated also isn’t the main factor. Elevation gains are counted relatively—without calibration, only the start and end elevations are incorrect. The key factor is when an elevation gain gets counted!
If a route only ascends (each measurement point is higher than the previous one), it’s simple: elevation gain equals the highest minus the lowest measurement point. But what if a route is undulating, with measurement points sometimes a few centimeters below or above the last point? Depending on the device or software, there’s a threshold for when elevation gain counts at all—often around 3 meters. So if you ride 2.99m uphill then 2.99m downhill, this isn’t counted as elevation gain but as a flat route with 0 elevation gain. The same applies to 10cm uphill, 10cm downhill, 10cm uphill, 10cm downhill… This is often the logic used when Strava and Garmin Connect “recalculate” elevation gain to arrive at their known values.
This explains why values are very close on real climbs but far apart on undulating or flat routes. On relatively flat courses, the difference between centimeter-accurate elevation addition and smoothing can vary by a factor of 3—and this difference decreases with better altimeter quality.
Here are some comparisons based on a Challenge Roth course ridden with a Garmin bike computer (first lap plus descent: 93.6 km; second lap: just under 85 km), which illustrate this well:
The resulting GPX file has 11,236 individual points. Counting elevation to the nearest centimeter gives 1,397 meters of elevation gain (a). Calculating out all elevation changes less than 3m gives 701m (b).
The original route shows 803m with Garmin Connect’s altitude correction (c) and 709m without altimeter correction (i.e., barometric pressure-based) (d). We see that (b) and (d) are only 8 meters apart—this should be a realistic value.
icTrainer smooths the route in GPX mode using 15-second averages, resulting in 751 points with 804m to the centimeter (e) and 699m when calculating out all elevation changes less than 3m (f). This value from icTrainer corresponds almost exactly to the above values (b) and (d).
If you upload the route ridden in icTrainer to Garmin Connect, you get 901m with altitude correction (g) and 702m without altitude correction (h). Again, value (h) is very close to (d), (b), and (f).
You can see the wide range of results here. Challenge Roth specifies 1,200 meters of elevation for the entire route, which would mean 625 meters for the first lap plus descent and 575 meters for the second lap—significantly lower than some calculations.
Similar information can be found in this article on the same topic.