Review of Run App

Tested Run 0.06 and Recorder 0.09
Unfortunately due to the Recorder settings CRASH issue my test is invalid.

The GPS on Run will have taken fixes every 1 second and stated I had done 4.2m which I know was about right. I did almost 8k steps. The recorder was unfortunately set at 10 second intervals (and I could not have changed it due to the bug reported) so downloading the recorder GPX file and loading it into view ranger says I only did 3.56 miles. But when you look at route it is not following the path as it is skipping points. To be expected. If we want these to come out the same then I think Run and Recorder MUST use the same recording resolution. There is an argument for an interface to the recorder that says start the recorrder client=run, use these parameters for the recording.

I also recorded heart rate data as I wanted to compare against an Amizfit Bip and see exactly where heart rate is failing. But my data is going to be fairly useless. I was attempting to test for my max heart rate and recorded 174 on the Amizfit at one point.

A few question arise: For Heart rate in the recorder do we use the confidence value at all, do we take a rolling avarage say ? On Bangle 1 I came to the conclusion that confidence was a waste of time and it was better ignored and heart rate averaged.

What does source mean in the CSV output ?

I have attached the CSV file if it will help anyone. Really dissappointed that the settings was every 10 seconds. I had set it that was before but really think the client app should be passing the settings across to the recorder. Going to be next weekend before I can do a really thorough test like this again.

2 Attachments

• Bangle.js Track 1.csv
• 

On Bangle 1 I came to the conclusion that confidence was a waste of time and it was better ignored and heart rate averaged

My impression is exactly opposite right now. Should have a real statistic by tomoorow. But it seems HR is really good when confidence is high (>90) and way off with low confidence which would make any averages useless. I got this impression from looking on BT HRM which show all relevant values.

If you look at the first 4 rows in the screenshot. Heart rate 99 - confidence 40, then confidence is 0 for the next 8 readings but most were close to 100 bpm.

Did some statistics. We were both right sort of:
While the standard deviation gets much bigger with lower confidence and also the mean error gets bigger, it is still not as totally off as I would have expected:

See scatter plot of dHR over confidence (dHR is differnce between bangle reading and a polar H10) and the standard deviation over confidence in the two attached pictures.

This recording is from walking on pavement and driving car as a passanger. For an other recording with walking on rough terrain, the result is pretty much the same, just higher amount of data with low confidence.

3 Attachments

I think with your statistical approach it might be possible to estbalish a minimum confidence level where taking an average is going to be good enough for most people.

The question is how much accuracy people are after when measuring heart rate. For example I dont really mind if its +-3bpm out against something more accurate BUT I would like the HRM to work when I am running and up to 7min mile pace would cover most casual amature runners. Obviosuly once you are going over that you are pretty much a known athlete (local club level) for 1/2 marathons etc, and accuracy will matter a lot. For the likes of me (currently tying to get into the couch-5k bracket) I'm not too bothered about HRM accuracy, just want it to record 150 when I am running at 6 miles an hour etc.

For meauring one version against another what should the test methodology be at running speed. EG run at 4mph , 5mph and record heart rate to within X% of some other device. Aplogies for talking miles, all my main running was done before quoting everything in K became the norm.

For meauring one version against another what should the test methodology be at running speed

Right now the statiscs only reflect the algorithm used at recording time and are not suitable to compare differenct parameters/approaches.
Before we can start assembling a test suite for the runners use case we first need to make sure we can record all data necessary, especially ppg, acc and some bt reference in sync and are able to do something useful with that data.
To ensure the last point we have halemmerichs hrmaccevents for now as a great starting point for local use at the computer as said.

But then yes, a set of files from different runners is what I think of. I don't know how relevant the concrete speed is above some point - depends on style I would guess, but we see the speed roughly from number of steps or more accurate from GPS if provided in the file anyway.
[Due to an injury, I can't run for some time unfortunatly, will start with cycling]

And sorry that I posted those statistics here in the Run App Review Thread
I will stop discussing this here and continue in the PPG thread until we have a conrete call for data donations.

Unfortunately due to the Recorder settings CRASH issue my test is invalid.

I just looked into this, it seems like it never worked? Fixed now, but you could just run 'Recorder' from the Launcher to get the menu

On Bangle 1 I came to the conclusion that confidence was a waste of time and it was better ignored and heart rate averaged.

Don't do that now! Confidence is much better and really useful, even on Bangle.js 1.

Also as @Mi said, we shouldn't try and play around with the HRM figures in the app. The internal HRM calculations do what's basically a median filter with an average of the HRM figures in the middle - so there is already averaging done.

Maybe what's needed is just a bigger median filter and more tolerance of different values - but that's something we need to work out based on data.

Don't do that now! Confidence is much better and really useful, even on Bangle.js 1.

No problem. I wont do this, if things have moved on.

I think I only have maybe one Bangle 1 apps where I attempted to use HRM.
Should probably revist at some point.

Fixed now, but you could just run 'Recorder' from the Launcher to get the menu

Have updated and will give it a test tomorrow. Its been a while since I used the recorder but feel I will probably switch over to it.

Re-tested latest Run and Recorder apps today over a 1.03mile loop.
Had recorder set to record every second.

Distances recorded:

• Run showed 1mi (needs 2 decimal places when in mile or km mode)
  
• When the track is viewed in recorder it shows as 1m (no decimal fraction).
  
• Amizfit Bip showed 1.03mile (I have verfied through a Garmin eTrex this is correct)
  
• Downloaded the GPX file and loaded into Viewranger and it said 0.97 miles.
  
  

HOWEVER when the file is imported into viewranger there is a tolerance of 30m and it throws points away if it thinks it can make a simpler polygon.

CONCLUSION: I am satisfied that with the recorder set to 1s intervals that the distance calculations by Run and Recorder come out the same. Any discrepancies will be down to not setting the recorder to match the same GPS fix frequency as Run OR down to import issues when the downloaded GPX file is imported into some mapping software.

2 Attachments

• Bangle.js Track 0.gpx
• Bangle.js Track 0.csv

It's a week or so I'm having issues with the run app. The app doesn't show live data, it displays data only when I click the button to stop it. I have 4 boxes: distance, time, bpm and speed. The only one I have is bpm. Ah, it doesn't belong here but I also have problems with the bthrm.

Am I the only one to have this issue with this app ??
btw, bthrm works better with the new version (but still takes time before the first connexion occurs (10 seconds)).

@Fteacher what version of Run are you using, what version of Firmware? If you add a screenshot of how you have configured RUN I will see if I can reproduce later tonight.

I use run v0.06 and the firmware was 2v12.27 I think. I didn't try it again since then.
The boxes are
box1:time box2:distance
box3:speed box4: bpm.
The gps turns green, as well as the run item when I start it with the side button. The bpm starts and then freezes, nothing else moves until I hit the stop button and then I can see time and distance.
I use French to have kms also.
But all this used to work in a previous setup...

Just a thought, but it could be something to do with BTHRM changes? I haven't changed the Run app in a few weeks

I'm on Run 0.06. FW 2.12.32. I have set run up like yours. I'm using the standard heart rate monitor on the watch. Everything looks like its working for me. The screenshot below I dont have a GPS fix but this was just to check I could see the time ticking over and nothing was freezing.

1 Attachment

I updated, tried different things and ...
it works again ! There was a problem with the recorder app. I kept my settings and asked the run app not to record and it worked. So I switched the recording back on and deleted all previous tracks. It was supposed to record on track 0, and the run app didn't work (still no live data). So I moved to record on track 1 and it works again... Something strange with track 0.

I'm running on a regular basis with the B.js2.
I saw that the time is now readable in dark mode ! Great !
I still think a few improvements would be welcome, like

1. Change the font for the numbers, make it bold (or somehow super readable).
   
2. Have access to a small accessible text file at the end of the run with the stats.
   This is more an open question, but the infos from my external hr device seem to be averaged over a long period of time. I haven't mesured it, but my HR seems to be 30 seconds late. Technically, for my training I try hard not to exceed specific heart rates, but it's difficult because the heart itself has some inertia, and a personnal factor, I'm running in a place with varied declivities (it's called "7 hills") leading to lots of changes in the effort and in the HR. The smoothing is clearly a bit too strong for me (and probably other people with external hr devices), but maybe it makes more sense when using the wrist HRM.
   

I just checked and the BPM value is updated with the value that comes straight from the HRM event - so if your BPM measurement is 30 seconds late that's probably because of the Bluetooth HRM device rather than the run app.

Looks like the font is currently a pixel-doubled 6x15 font, so it's 2 pixels wide. Problem with using something like "6x8:3" is it's too wide. If you found a specific font that worked, changing it would be pretty easy though

I think HRM device works poorly when low on battery.
I've seen changes in the latest version of the run app, the notifications ! That's a nice idea. As often, it brings me ideas, to which I will add a few older things.

1. Could the notifications use the "HR alarm" app to setup a max high and max low value ? I had a discussion about that recently with the coder of this app. Runners should do 60 to 85% of their training in the hr zone "endurance"(slow). For me it means that I have quite a few runs in the month that are entirely in this zone. Therefore, including the app directly would make a lot of sense. The HR Alarm app allows for incremental changes of 5 by 5 bpm, it should be changed to 1 (longer to setup, but it doesn't need to be changed often so...)
   
2. I still hope for a small readable text file that gives me my average HR on the watch when I stop the recording, as well as my average speed
   
3. There is A-pace(now) and C-pace(so far), but for kms, there's only speed (A-speed), no C-speed. Besides being a fair addon, this could actually display an average speed at the end of the run (as in point 2). But to have less infos the watch, I anyway limited the number of boxes to 4, which leads me to the next point:
   
4. I still think that having 2 screens that could be configured in the settings would be the perfect trade-off between readability and the number of informations. During the run it could be rotated by a swipe.
   Btw, I criticize a lot as, constructively as possible, but hey, big up to the bangle.js 2 ! Even with a few drawbacks, I now prefer running with it rather than with my Garmin forerunner 220, which is only dedicated to running but full of purposefully designed limitations. I don't tell how much proprietary software sucks, it's all clear to everyone. @Gordon, you defeated the opponent here and surpassed my expectations. Well done. And thanks to the community also !
   

Thanks! That's interesting about the Bangle.js HRM - and that's on the Bangle.js 2?

For the 2 screens - that could be a good idea. I think for that kind of thing it might make sense to fork the current app and add a new one - but it should be a pretty easy change to make.

Other things - yes, those seem like good ideas. However I'm super busy at the moment and I don't really have the time to add those. If someone else wanted to, that'd be great though :)

Sorry, I made plenty of typos and wasn't clear. It's my external bthrm device that works poorly when low on battery.
For everything else, I am patient and already quite satisfied. I'll speak again about HR zones later :) !

@Fteacher, I am the one who added the notifications and it's really cool to hear that someone else likes them! I'm curious if you find them currently strong enough; after a few runs I'm finding that I miss some of the notifications.

I totally agree with the HR alarm idea, but have had issues with my watch reliably showing my HR at all, even when I am motionless. After I get that sorted I plan on adding that feature.

@GrandVizierOlaf The HRM is pretty unreliable for me too, but some of us on this forum use an external Bluetooth HRM especially when running, to get accurate data. That works pretty reliably. It would benefit some of us if the option is already here...
I agree on the vibrations being a bit weak, but I have to try it more.

The B2 heart rate monitor is only accurate when standing still. Its pretty useless when doing exercise in my opinion. In my Daisy watch I display the heart rate value in white if the confidence value is over 50% and red if less than 50%. When exercising I have to stop running to read the heart rate with a confidence value greater than 50%. Its quite a while since I have played with my Bangle 1 - but seem to remember the HRM was a lot more reliable on a B1 and did not have this problem.

Until this gets fixed, needs some real focus and some form of test harness, B2 is not going to be able to claim its a fitness watch.