Optimizing Segment Creation: how to create good segments
posted this on February 27, 2012 03:57 PM
There are certain rules to follow when creating segments that can minimize the amount of problems your segment will experience.
It is possible to decrease the risk of producing false-positives, missing, and inaccurate matches, and more. For detailed information about how to optimize your segment creation, please see the information below detailing the types of segments to avoid, and the specific problems to which those segments can be vulnerable.
Zoom in! The closer you can see the data that you're working with, the better the end result will be. Switch away from Terrain Maps (at the top right of your map display) to either Standard or Satellite maps to enable you to zoom in further. Keep in mind that when you're at the segment creation page, the "Move Start/End Point" buttons are dynamic; when you are zoomed out, they will move the endpoints of your segment in large increments - as you zoom in, they will make finer adjustments.
Very short Segments
Segments that are very short - for example, sprints that take less than 15 seconds to complete - will be much more prone to matching errors than longer segments. There are two severely limiting factors for segments like this.
First is the sampling rate of your GPS device; some devices only record a GPS point every 5 to 10 seconds - meaning that the margin of error of your segment match accounts for a very high percentage of the total segment time, since the segment is so short. In the example below, you can see that there is no GPS point that falls exactly on the start and end points for the segment, and that the entire segment match is comprised of only four GPS points
Furthermore, since the segment is so short - in terms of distance this time - the start and end points will be very close together. There is a degree of leniency necessary in our segment matching system to account for inaccuracies in GPS data, and it's possible that both the start and end points of a very short segment can fall within the 'buffer' zone intended to allow for slight GPS drift. This can cause our system to match the segment when there is hardly any overlap between the segment and the ride.
Segments with bad start or end points
Segments created with start points in less than ideal positions can also be very problematic. For example, if you create your segment in such a way that it starts too early, some users may have trouble matching that segment; see below left for an illustrated example of this.
Additionally, placing your segment start point too close to an intersection might lead to inaccurate segment times. The reason for this is that if there is a stop sign or other traffic regulation for the intersection at which the segment starts, users might trigger the start of their segment match prior to stopping - meaning that their time while stopped at the intersection will be included in their time for the segment. See above right for an example of this.
The reverse of both of these examples is also true; placing the end point of your segment too close to an intersection can lead to stopped time being included in a segment effort. This is very common at the top of climbs; when a segment finishes too close to an intersection, users will often stop to rest after pushing themselves up the climb - causing their resting time to be counted into their segment time. Likewise, placing your segment start point after an intersection can cause some users not to match at all, if they don't turn the same direction that the segment does.
Segments that start or end on a tight switchback turn can cause bad matches. Because our system allows for some small margin of error to account for GPS drift, it is possible that you can trigger the start or end of a segment before or after you actually should have - if the road or trail makes a very tight turn and turns away from, then returns close to the true endpoint. We are hard at work on improving this situation, and hope to handle these segments better in the future. In the meantime, the best way to increase your chances of a good match to segments like the one shown below is to extend the segment as far from its own 'tail' as possible (while conforming to the other guidelines discussed on this page).
Trails that parallel roads
Because our system allows for some small margin of error to account for GPS drift, if a trail segment runs parallel to a road segment and the two are close enough together, it may not be possible for Strava to provide mutually exclusive results. This means that athletes riding or running on the road may match (and receive unrealistic times) on the trail segment, and vice versa. Unfortunately if the two paths are close enough together for this to be an issue, the only way to improve your chances for accurate matches is to move the segment endpoints further away.
Very long segments
Longer segments are generally more accurate than shorter ones - however only to a certain extent. If you are completing a 30-mile long segment but make even a very small deviation from the route that the segment follows, you won't be able to match the segment. Even though a one-block discrepancy might have little effect on your time for the segment, any deviation from the segment's route will prevent it from matching your activity. See below for an example of this.
Very complex segments
Certain segments are just so complex - composed of so many tight twists and turns - that we are unable to give completely accurate results. This is sometimes evident on cyclocross courses, tightly-packed trail networks, etc. This is exacerbated by the fact that segments like this are most often found off-road - with tree cover frequently negatively affecting the quality of both the segment and the activity GPS signal. Once again, there unfortunately isn't a real solution to this problem other than to use the best available GPS data, and whenever possible follow the other guidelines described here.
Any time your route takes you through a tunnel of any significant length - as are common in mountainous areas - your GPS signal will be lost, and it is very likely that the distance you travel while without a GPS signal will be long enough to trigger our 'Gap Threshold.' This means that segments that incorporate tunnels will be unlikely to produce accurate matches to your activities. Unfortunately there isn't a great workaround for this - the best solution is to simply avoid creating segments that include tunnels longer than around 500 meters.
Things to keep in mind
Use good quality data!
Quality GPS data is key to accurate segment recognition. Since we use the GPS data from your activity to create segments, it's crucial that segments are created from activities with high quality GPS data - and that, whenever possible, are recorded on a high quality GPS device. Not all GPS devices are created equal but even for the most accurate devices, trees, tall buildings, and even clouds can wreak havoc on GPS signal. Look closely at your activity before choosing it as the basis from which you create your segment; if the GPS data looks jumpy or doesn't follow your route closely, try using another activity.
Recording intervals vary between devices - for example, the Strava mobile app records every second while Garmin devices use either 1-second intervals or a smart recording which has a varied recording interval. Segment matching works the same on each GPS dataset, but depending on the device's recording interval, can yield different results. Segment matching uses the GPS points in the data closest to the start and endpoints of the segment, and as this can vary with each activity, timing on a segment can vary slightly because of this. At the present time, we don't interpolate or extrapolate GPS data to normalize for the exact start and end positions of the segment.
Are there duplicates?
Check to see if the segment already exists. After you specify a start and end point and click "Next," you may see a list of similar segments. Review them to see if the new segment you're proposing is truly new. Sometimes the segment exists, but poor quality location data prevents Strava from matching them. In these cases, see Segment Matching Issues for instructions on what to do. Keep in mind that if there is a segment found that is similar to the one you're creating, selecting to "Use" that segment won't automatically add it to your activity.
For more information on Creating Segments, see here.
Re-Routed off-road trails
If you are attempting to create a segment to account for an off-road trail that has been re-routed in response to environmental changes, there unfortunately isn't a way to prevent users on the old 'version' of the trail from matching the new one - or vice versa - assuming the two trails are similar enough to match regardless of which route was followed. The best option is to filter segment results by 'This Month,' 'This Year,' etc.