Is this another run cheat?

Beware the mileage champions wrote:

No sense in any implied public shaming till I understand what we are dealing with. At this point, I don't, not in totality.

I think you are reading way too much into this data. I suspect it is just poor quality data. When you look at the cadence when it is around 170 there is a healthy degree of scatter which is what you would expect from anything of human origin. We are humans not machines. When the cadence drops to about 120 it is basically rock steady, no change from day to day, no change with duration of the run. You can draw a straight line through it. If he was cycling there would be a similar scatter to that seen when running, possibly worse due to gear changing etc.

So, either he is jumping on a machine that beats out a cadence of 120 hour after hour, day after day, or the cadence measurement craps out. I'd put my money on the latter. Regarding the pace measurements: the increased pace portions also coincide with an increase in the downward spikes to very low paces. I strongly suspect that the average pace does not change.

I disagree that this is a clear case of bad data without analyzing it further. The OP has clearly provided annotated screenshots showing long durations with high paces when the cadence is zero. Cadence is not affected by bad GPS signals. I would need to analyze the raw data to see what is what.

OP, you can delete the name of the person the the links to the Strava activity by using a text editor like Notepadd++ and searching and replacing his name and Strava URLs.

I'm don't think you will be able to anonymize the GPX file by stripping out the location information because that's basically what a GPX file is.

You could always email me the files and I could process them into my Excel reports with all the location information stripped before posting the results here for people to analyze.

https://notepad-plus-plus.org/

[quote]scam_watcheroo wrote:

.... screenshots showing long durations with high paces when the cadence is zero. /quote]

I would not say they were long durations but rather multiple single point durations, the difference is significant. I find this issue interesting, not in terms of is this guy a hero or villain, but rather in terms of data quality and measurement accuracy. My own Garmin often spits out bizarre data but then it is a low end one. It only has GPS, no cadence or heart rate. I have assumed that the top end devices are more accurate. I also wonder if a single device can do all three measurements accurately or whether it would be better to carry three dedicated devices, one for each measurement. (if you are a training fanatic or going for some record)

It would be very interesting if the RoJos would fund a series of field trials where a runner sets off with four or six different devices strapped to his wrists and then the results can be compared for an identical run. Trials could be conducted over long and short runs, slow and fast runs, hot, sunny and cold, wet days etc.

For my own device the distance run can vary by up to 2% though I have never tried to work out what factors influence this. The average pace is accurate (to within the 2% distance error). The kilometer splits are believable but anything shorter than that has to be taken with a grain of salt. Generally sharp direction changes or relatively quick changes in pace (i.e. cresting a hill) seem to throw if off. I've had it freeze several times but that seems to relate to heavy tree cover and loss of satellite signal. I've never had it freeze due to sweat build up around the contacts and that does surprise me.

I have written a report up on this suspect after analyzing the data (which coincidentally, I already had access to from the past). The identity of the suspect and the data have all been anonymized per request:

https://drive.google.com/file/d/1cb_KrgX3VHsfi7YLQvyhD2UUDH0hG2BF/view?usp=sharing

https://drive.google.com/drive/folders/1gHxLMTenHPsYXajiTTKwXFmCVsNKqgre

scam_watcheroo wrote:

I have written a report up on this suspect after analyzing the data (which coincidentally, I already had access to from the past). The identity of the suspect and the data have all been anonymized per request:

Scam W: The report you did on Amy Hughes was excellent primarily because the data was based on physical hardware - treadmill revolutions. This report is not so good because it is based on data from an unknown source and quality. In my work we had two maxims:

Hardware never lies

Data needs to be validated

Let me present to you some data and ask you to draw a conclusion from that data.

Each day I measure my temperature: On Monday it was 98.6 °F, on Tuesday it was 98.6 °F, on Wednesday it was 98.6 °F, on Thursday it was 98.6 °F, on Friday it was 42.0 °F.

Which of the following conclusions is most likely:

1) I died Thursday night - if so how am I typing this

2) I am a space alien and my human cloaking device has stopped working

3) I really do have a temperature of 42.0 °F and on the other days I must have been running a fever

4) Maybe the Friday reading is garbage

I appreciate you playing devil's advocate here, but your example is incomplete and does not draw a direct parallel.

Add in a second data measurement, multiple sequences of data, and more prior knowledge. We could add heart rate measurements of 60, 60, 62, 55, and 1000. Also add 17 other data sequences, half of which show the same anomaly. Add in the prior knowledge that aliens tend to have low temp with high hr. Given what we know about humans, aliens , and the probability of two data measurements simultaneously showing huge outliers, we can reasonably conclude that you are probably an alien.

Interesting indeed.

Excellent work again Scam!

But please reveal their identities. We don't need to protect the feelings of cheaters. Because there is no institution like WADA involved in these stunt runs, public humiliation is the only punishment they will ever get.

I second the excellent work but unmasking the person probably not the right thing to do which would be the OP airing his issues in person to the individual.

1 2 3 greeen wrote:

Add in a second data measurement, multiple sequences of data, and more prior knowledge.

What is the second data measurement? We don't know what device or devices this person used. If only one device was used then all the data collected is a function of the accuracy, precision, repeatability and limitations of that device. All those are unknowns. A true second data measurement would require an independent measurement system which we don't have.

Multiple sequences of data. All this shows is the device consistently craps out and the failure mode is fairly consistent. Though the time to failure is pretty random and the failure trigger is unknown.

More prior knowledge. There's the rub. I have no idea who this individual is, nor do I want to know. It seems you do and would probably relish this person being exposed as a cheat. In so doing you are making a fundamental error in reviewing data, you have a preferred conclusion that you want to achieve. Don't be offended, it is a common trap to fall into. Many times in my career I've clashed with corporate vice presidents over cherry picking data. There have also been times when I have really wished I could come to a different conclusion but in the end you have to let logic prevail. We used to refer to it as going into Mr. Spock mode.

There are two claims being made from this data; one, that a world record of some sort was set, and two, that the person is a cheat. The data supports neither.

This analysis is useless without peer review. The only way to give this any credibility is to reveal the identity of the individual concerned, and the stunt run claimed, so that others can also review the evidence.

Sam, you are a pathetic POS with no life. How bout dah?

It would be nice if there was video evidence capturing all the runs but I don't think that exists.

Both the analysis from the raw unsmoothed data and the smoothed data using a 10 point rolling average with the trapezoid rule for variable time intervals are included in the Google Drive location. Noise exists in all data but I don't believe this is simple noise. The suspiciously high speeds exist even after smoothing the data.

From the 17 runs, there is one of them that I think was run entirely legitimately. See the graphs for Run #14, which don't indicate a significant amount of extreme stride lengths and the speed histogram don't indicate unusually high speeds:

I can add that most of these runs were done by running multiple loops through communities with the exact loop used changing each day. I think we can rule out poor satellite signal as an alternative theory because of both the different loops used per day and because we would then expect to see a regular pattern of jumps in speed on each loop.

We know the suspect is using Garmin watch(es) but I don't have information on which particular model or models.

From Pete Kostelnick's and Sandra Vi's run across America data that used a variety of Garmin watches, we have a large amount of reference data to compare to for what someone expects to see.

scam_watcheroo wrote:

We know the suspect is using Garmin watch(es) but I don't have information on which particular model or models.

Strava tells you what device is used underneath the elapsed time etc.

I'm all for verifying the data source, routes, etc. prior to outing the guy. I understand your hard stance on making claims based on reliable data. I am inclined to repeat that sentiment back at you when you assume I know this person and have an agenda just because of the side I am taking. Based on that logic I might say you are the cheater himself trying to claim faulty equipment. (I don't know him or the op.) Anyway, you seems to want laboratory style data collection when we can just apply some reasonable assumptions. This isn't that big of a deal.

The second measurement is cadence (but you already knew that). We have pace and cadence. It could be a foot pod. But it is reasonable to assume that it came from the same watch, since that is a standard Garmin feature. Just because it came from the same device doesn't mean both measurements becoming suspicious at the same time should be dismissed.

Yes, the "failure" mode is consistent. It's also consistent with what one would expect a bicycle ride would look like.

Prior knowledge did not reference prior knowledge of this person. It means prior knowledge of reliability of Garmin gps watches in pace, cadence, and how failures typically look. Anyone who has ever used a Garmin can look at that graph and conclude that the cheating is far more likely than device failure.

ExpertKipWatcher wrote:

I second the excellent work but unmasking the person probably not the right thing to do which would be the OP airing his issues in person to the individual.

I just did a bit more research. Our suspect has another extreme distance running challenge starting very soon.

Gathering some more details, our suspect will be attempting to run multiple marathon distances per day for a much longer period of time than even his first extreme challenge. This is planned to happen on public roads connecting many communities so there should be ample time for someone to track him to provide video evidence. This is happening very soon.

scam_watcheroo wrote:

Gathering some more details, our suspect will be attempting to run multiple marathon distances per day for a much longer period of time than even his first extreme challenge. This is planned to happen on public roads connecting many communities so there should be ample time for someone to track him to provide video evidence. This is happening very soon.

It won't be possible to track him unless you tell us who it is.