[HAMR] New challenger - Bruce Berkeley AKA Cycle_dr 1

[aoxomoxoa]

I think simonp's answer may be the right explanation, as a wheel-based sensor measures ground speed distance (i.e. includes vertical distance) whereas the GPS calculated distance assumes a flat surface.  See Strava's explanation: https://strava.zendesk.com/entries/21278088-How-Distance-is-Calculated.

Do we know if Bruce is in fact using a wheel-based sensor?

Likewise, if Kurt and Steve have been relying solely on GPS-calculated distance then have they in fact been under-reporting their distances? I guess this would be a slightly more significant issue for Steve as I think his routes over the year have been hillier. But for Kurt this could mean his mileage is even higher - though how this could be established in the absence of wheel sensor data I dunno.

[Legs]

There's something a little odd about the Strava figures for Bruce's rides so far (I'm not suggesting anything untoward, just an anomaly I can't make sense of).

When I looked at his total for days 1-7, my calculation based on the GPX data uploaded to Strava was about 49km short of the figure suggested by Strava itself. For only 7 days, this is a much greater discrepancy than I would expect even accounting for home-zone privacy settings etc. I have been working though my calculations looking for possible causes, but was not able to identify anything.

Then I noticed that if you take any ride, say this one on the 4th Jan, the headline figure on Strava says 337.1km (tallying with Bruce's own comment for the ride), but if you move the mouse over the elevation profile below the map, the distance only goes up to 325.7km. This lower figure matches exactly the one I had also calculated directly from the GPX file.

So which figure is the correct one, and why such a discrepancy? This is too big an error to ignore as ~10km errors for a single day will soon add up.

Any ideas?

Is this to do with the privacy zones that Bruce has set up in Strava?  Whenever I go out for a run, whatever portion of my run is done within a certain radius of my home is invisible to other Strava users, and doesn't show up on the elevation profile.

[red marley]

I think Simon and aoxomoxoa may have it. Bruce's Strava data include cadence and power so he must be using some wheel-based sensor. The discrepancy (about 2.1% on average so far) is too great for this to be solely due to elevation vs flat assumptions of distance, but it could be explained by a combination of elevation, the difference in calibration between wheel sensor and GPS and correction of GPS errors.

The problem becomes which figure should be used? A 2.1% difference would amount to over 1,500 miles for a 75,000 total, so this is significant, and could well be more than the difference between riders by the end of the year.

For the moment, in my visualizations, I will be going with the (lower) figure as stored in the GPX file, not because I necessarily think is the figure that should be officially recorded but because it will be a bit of a faff to rescale the data.

[Wowbagger]

I have always been of the view that a GPS under-records by about 2% compared to an accurately calibrated wheel sensor, on the basis that it is "joining the dots" with a series of straight lines. Having said that, for a record of this nature it isn't possible to rely in wheel sensor because all competitors' distances have to be recorded in a common manner.

[danridesbikes]

Bruce's cadence data might come from the stages power meter

[Greenbank]

I think simonp's answer may be the right explanation, as a wheel-based sensor measures ground speed distance (i.e. includes vertical distance) whereas the GPS calculated distance assumes a flat surface.  See Strava's explanation: https://strava.zendesk.com/entries/21278088-How-Distance-is-Calculated.

He didn't climb 49km though (the discrepency can't be any more than the distance climbed by pythagoras).

Do we know if Bruce is in fact using a wheel-based sensor?

Likewise, if Kurt and Steve have been relying solely on GPS-calculated distance then have they in fact been under-reporting their distances? I guess this would be a slightly more significant issue for Steve as I think his routes over the year have been hillier. But for Kurt this could mean his mileage is even higher - though how this could be established in the absence of wheel sensor data I dunno.

Does the GPX file that you can download from Strava contain any <DistanceMeters> tags? That's how the Garmin would be storing the distance from a cadence sensor[1], and what Strava might be using for the distance. Don't have time to check myself right now...

The privacy zones (as suggested by Legs) will also contribute, it just depends how big he defined the zone and how many times he want near it (or them).

1. This is a bit of a simplification, and it's also most likely that Bruce (and Steve) are uploading .fit files which store the data in a different way.

[Greenbank]

Not sure what UMCA will do with such a discrepancy as a miscalibrated wheel circumference could make it interesting.

[PAC]

Bruce's cadence data might come from the stages power meter

Yes, the Stages PM provides cadence data.

I you look closely under the left-hand chainstay in this photo of his bike he tweeted on 8th January, you can see a BBB speed sensor dangling there. https://pbs.twimg.com/media/CYL0lh-UwAA81oQ.jpg:large

That's a rapid looking bike

[Greenbank]

Does the GPX file that you can download from Strava contain any <DistanceMeters> tags? That's how the Garmin would be storing the distance from a cadence sensor[1], and what Strava might be using for the distance. Don't have time to check myself right now...

Just checked, no it doesn't, it's just lat/lon/ele. Doesn't rule it out though. To check for sure you'd need to get your hands on the original .fit file, which is unlikely.

[Karla]

I have always been of the view that a GPS under-records by about 2% compared to an accurately calibrated wheel sensor, on the basis that it is "joining the dots" with a series of straight lines. Having said that, for a record of this nature it isn't possible to rely in wheel sensor because all competitors' distances have to be recorded in a common manner.

Not necessarily.  As I think I've recounted before, my GPS over-recorded by about a mile on this year's 12 hour TT - which was highly annoying as I thought I'd broken 260 miles but the results sheet said I was a mile short!

Your GPS uses something called a Kalman filter, which works out a 'best guess' position based on a weighted average of your current measured position and a prediction based on previous measurements.   That means that GPS tracks often overshoot turns, based on their predictions, before being brought into line by new measurements.  That's one reason why tracks can be over- rather than under-distance.

[Ham]

A Calman filter - what a great idea!

[wajcgac]

For a 2% discrepancy, he must have the calibration out of whack. I've found the auto-wheel size calculation that the Garmin Edge does rather inaccurate, and you get better results overriding that and entering a manual value. Doing this, and running two Garmins - one off GPS, one off a speed sensor - over 400km and 600km distances I've seen differences of around 0.2%, an order of magnitude better than what Jo's finding from Bruce's tracks.

It would be interesting to manually set the wheel calibration to say double what it really is rather than use the auto-wheel size. Go out for a ride on a bike that also has an old style non GPS computer on it as well (set to the correct wheel size).

Then upload to Strava and see what it actually reports the distance to be compared to the non GPS computer . I'd have thought Strava would calculate your distance on the GPS positioning data and not that recorded by the incorrectly set GPS wheel sensor.

Taking it a step further if it took the speed data as correct from the wheel sensor then it would make getting KOMs much easier but that won't happen as it calculates your speed from the time between GPS coordinates at the start and finish of the segment.

Does this make sense and is it worth trying it?

[TimC]

Using an Edge 1000 with a speed sensor (and cadence from Stages, so very similar to Bruce's set up) and a Vivosmart recording the same rides up to about 100km, I've found the difference to be in the order of 0.5% with the Edge recording a slightly longer distance. It doesn't bother me, but it could be very significant in the context of HAM'R, and it never occurred to me before - I wonder who else it hasn't occurred to?

[simonp]

I've always found some discrepancy between GPS tracks and reality. No-one has mentioned WGS84 yet either.

When Kurt broke the record and celebrated I had a little voice at the back of my mind about this. Not to pour water on his achievement, but different methods give different results.

[wajcgac]

Out of curiosity I've been out an done a little experiment.

Cateye Enduro 8 V Garmin Edge 500 set to use manual wheel circumference.

The Garmin was set for a manually entered wheel circumference of 4192mm - double the Auto wheel circumference it was originally set for which was 2096 mm.

The Cateye was set to the std wheel circumference for a 700/23c wheel of 209 cm.

According to the Cateye I cycled a distance of 1.18 km (on a flat route)

As expected whilst cycling along the Garmin showed my speed and distance travelled to be twice that recorded by the Cateye.

When uploaded to Strava, contrary to my thoughts earlier the reported distance was still twice that of the Cateye at 2.3 km, max speed and average speed were both double as well.

When looking at the analysis of the ride it starts at 0km and finishes as 2.3 km as I don't have privacy zones set (I crop at start and finish of ride instead)

Therefore it would seem that if you took the Strava upload at face value there would be no indication that there was something wrong with it and you could easily believe someone had cycled further than they actually had.

I then exported the GPX file from Strava and uploaded it to Ridgewithgps - where the distance ridden was shown as 1.2 km (I think this is similar to how Jo discovered the initial problem)

When uploaded to Garmin Connect the distance is shown as 1.23 km.

So it would initially seem a problem with Strava as other websites aren't fooled by playing around with the wheel sensor circumference data.

(I know Kurts Garmin connect data was always under distance compared to his Strava data by as much as 1.5km per day but mostly less than this so I'm sure his distance data is as good as you can get with GPS.)

Here is Bruce's Garmin connect page

https://connect.garmin.com/profile/cycle_dr1

Based on what I had seen above I was expecting to see a difference between this and Strava but there isn't any so that would appear to dispel the wrong wheel circumference theory.

I can't seem to be able to look at the Garmin Connect analysis of Bruces ride as the website comes back with a timeout error - may be because I'm not 'connected' with him.

All very confusing.

Jo, how are you able to get the GPX data for Bruces ride on the 4th - are you a Strava Premium Member?

 

[TimC]

I had no problem getting into Bruce's ride analysis on Connect, so I guess it's a local problem for you. The GPX of anyone you're following is certainly available to Strava Premium members, not sure about standard members.

[wajcgac]

Thanks Tim, as you say it looks like a temporary local problem for me as it's working now

Just had a look at the analysis on Garmin connect for the ride on 9th Jan.

https://connect.garmin.com/modern/activity/1008396301

https://www.strava.com/activities/466219804

Garmin has no privacy zones so I compared it with the ride shown on Strava.

Garmin shows that he rode about 2.3 km before he emerged from his privacy zone on Strava and he rode nearly 6km at the end of his ride whilst within his Strava privacy zone.

Bruce seems to do a few small loops at the end of the ride - I guess just to get the distance up a bit.

In total about 8.3km extra can be seen on Garmin Connect for this ride.

Looking at the ride on the 4th Jan it's nearly 12km extra seen on Garmin connect as he does small loops around the finish again.

https://connect.garmin.com/modern/activity/1002922268

https://www.strava.com/activities/462734868

I'm guessing Jo that if you are downloading the GPX file on Strava, even as a Premium Member and somebody following Bruce it is not giving you the full data to observe Bruce's privacy settings.

The Strava privacy setting seems to work in a slightly odd way - in that it seems to crop the ride to the point where you finally leave or enter the zone. You can pass through the zone and an observer will see the track - the bit you can't see is before he leaves the zone at the start and from the last time he enters the zone before the ride finishes.

One thing to learn from this is for those who have rides uploaded to Strava and Garmin is the need to set your privacy settings on Garmin as well. Garmin is done differently in that you can restrict access to your rides to certain people - yourself only, your connections, connections and groups or everyone.

[red marley]

Thanks wajcgac for your work on investigating this. I think your observation that Bruce, unlike the other riders, is riding significant distances within his Strava privacy zone either at the start or end of the day explains the discrepancy for all days. It doesn't look like there is any observable miscalibration of the bike sensor vs GPS distance measures.

I've updated the data on gicentre.org/oytt to reflect his full distances including the privacy zone riding.

[Exit Stage Left]

Bruce would seem to have been called Dave back in the early 2000s.

He seemed to peak in the Surrey League 5 day in 2003. With a good showing on day 3.
http://autobus.cyclingnews.com/road.php?id=road/UK/2003/aug03/aug15-19evanssurrey5day033

The afternoon's stage of 84 miles around the South Downs saw a breakaway group of 13 riders go clear in the opening miles and build a lead that at one time exceeded 4 minutes. Most of the main teams were represented in the break with VC St Raphael, Energy Cycles, Evans RT and Parrot Print RT each having two riders but the greatest beneficiary looked to be Dave Berkeley (Sigma Sport RT) who looked like becoming the new race leader. However, in an exciting finale the main bunch reduced the group's lead to less than two minutes and at the finish Justin Hoy (Evans RT) produced an outstanding sprint to win the stage ahead of Timmy Barry (Ras Mumhan) and Berkeley.

He's a contemporary of Gethin Butler, who won that event in 2000. Bradley Wiggins rode that level of event at the time.

2003 was his big year, there are a few results from 2004, but nothing before. He's not listed as having a category, so may have held an NZ licence.
https://www.britishcycling.org.uk/points?person_id=28236&d=4&year=2003

[TheRedEyeJedi]

Seems like he did a bit of riding with Team Sky yesterday as part of 212 miles at 18.7mph.  Some decent pics of them leading him out on twitter and facebook.   He is also using a go pro on board which could open the possibility for some nice footage being released after his attempt

[Freya]

What sort of support is he using? He says he has 'shredded' a couple of tyres in  three days but is still pumping out the miles in a fantastic fashion.

[marcusjb]

Hoppo saying on Facebook that Bruce has been disqualified from HAMR.

Would not be a surprise as he has not had the same data transparency as the other contenders (particularly HR).

[Ham]

Could it be "Solo riders only; no teams"?

[Jack_P]

Hoppo saying on Facebook that Bruce has been disqualified from HAMR.

Would not be a surprise as he has not had the same data transparency as the other contenders (particularly HR).

Can't be the reason, the power data is proof of his efforts, unless they think he is on an e-bike ? And why does this leak out via Hoppo, the HAMR web pages don't serve the competitors in the challenge very well

[clarion]

Oof!  Seems harsh.  I wait to hear the details.