scam_watcheroo wrote:
If my theory about a bicycle/vehicle being used is correct, I may be able to find other people with GPS watch data that rode on a bike and compare graphs to see if the data looks similar.
I had a look at some of my bike recordings and the data quality is much the same as it is for running. If I hid the axes so you couldn't see speed and distance you would not be able to tell which is which. I don't have cadence measurement so I can't comment on that. I'm not sure what could increase noise on a bike - interference from the metal frame is about the only thing I can think of. I had a look between recordings I have for a steel framed bike and an aluminium bike but could see no difference. And again, there is no difference to my running data. To get cadence on a bike I assume you would have to have a foot pod as the watch is not going to be moving at all.
I did do a quick sanity check on the data and I was a bit surprised by the results. I looked at Recording 13 as it has the clearest distinction between the two zones. I then calculated the average speed using your values in Columns D and E . In the high cadence zone I took the values from 1.5 to 2.9 hours into the run and for the low cadence zone I took the values from 3.4 to 7.1 hours. The two average speeds were 4.18 mph and 6.83 mph. I had thought the two values would be closer. I don't know how the other recordings compare but visually they look consistent. I still believe the data is corrupted, the noise is too great and the cadence data is way too bizarre to determine anything from it. But I am curious as to the difference.
If you want to push the cheating argument further I think you need some secondary evidence. You have access to the routes he ran and you said he ran multiple loops . Physically where are the transition points from running to "cycling"? If they always occur at the same location - right in front of his house - that would be significant. If they are randomly scattered around it would prove nothing, though people will speculate.
Also, there are two points in Recording 13 where the speed reaches very high values - 35 mph at 4.42 hrs and 30 mph at 6.31 hrs. (your smoothed curve lowers these values but they still stand out). Topographically where do these points occur. They would need to be on a steep down hill to achieve those speeds on a bike. If he used this loop for other runs does the same phenomena occur at this location?
As I said I feel the data is crappy and it is only good for generalities rather than hard facts. If any conclusion is going to be drawn from it ancillary data will need to be used. I said in an earlier post hardware doesn't lie. In this case we could say hills and doorsteps don't lie.
Sorry this post is a bit long but I've spent my career digging through engineering data so it becomes something of an obsession.