Segments are often being labelled with activity types that would generally conflict with the speed i.e. cycling at 50mph or walking at 15 mph.
This is already how the system works 😉
The activity type ML models take speed into account, along with many other factors (I think over 20 factors in all).
Keep in mind that the speed you see for a trip is only the average speed of the trip. The majority of the samples within the trip could have been much slower than that speed, with only a small handful of samples being faster, thus pushing up the average.
When the classifier picks an activity type that doesn’t seem to match the average speed, it will be the case that that type does match the most common speed of the trip.
@matt that may be how it works in theory, but in practice it still very frequently guesses “walking” or “cycling” at impossible speeds, and this is after almost two years of training it. Probably happens 3 or 4 times a week that it guesses cycling when my avg speed is like 89kmph (I can share many screenshots of this happening if you want). Is there no way to hardcore avg speed limits for each activity type?