Treadmilling on Apple watch

[imagineadam]

What app do you guys use when treadmilling?

I have the series 2 and I have calibrated my watch but it's always a little bit off compared to the treadmill which I suppose is expected since nothing is perfect.

The annoying thing is when I go to change and correct the distance in the nike app then it records yet another work out in the workout app. Super annoying. I don't want to have to turn off its access to writing to the health app every time I treadmill in order to avoid it recording multiple workouts.

So would I be better off just using the Other workout in the workout app then manually putting in the correct distance into the Nike app and turning off its access to writing workouts to the health app each time? Would it still provide pretty accurate calorie counts compared to it knowing that I'm running? What do you other runners do? Do you just live with the slightly off distance the apple watch usually provides? Or do you do something else? Thanks!

 

[Resqu2]

I run on a treadmill everyday of the week, I have found that when I do a 10 min mile or faster pace the distance is almost dead on it for 3.1 miles. If I do a slower run at a 12 min mile pace I an ahead on my Apple Watch by over a quarter of a mile by the time I finish my 3.1 mile run. I only use the indoor run exercise on my watch. I have done many many walks and runs outside to calibrate it before Winter got here and forced me onto a treadmill everyday. As for the difference, I just pretty much live with it knowing I will be back to trail running in a few weeks.

 

[pedrowerner]

My watch is always 600m behind the treadmill

 

[Alvi]

My series 1 is spot on when I run on the treadmill and I think I never calibrated it running outdoors. Maybe try to reset it and calibrate it again.

 

[AlliFlowers]

Treadmilling never closed my circle, so when I get on the treadmill I set the activity to "Other."

 

[BarracksSi]

I run on a treadmill everyday of the week, I have found that when I do a 10 min mile or faster pace the distance is almost dead on it for 3.1 miles. If I do a slower run at a 12 min mile pace I an ahead on my Apple Watch by over a quarter of a mile by the time I finish my 3.1 mile run. I only use the indoor run exercise on my watch. I have done many many walks and runs outside to calibrate it before Winter got here and forced me onto a treadmill everyday. As for the difference, I just pretty much live with it knowing I will be back to trail running in a few weeks.

For you, it sounds like your tempo, or cadence, is the same even if you're running slower. That is, your rhythm is the same, but your stride length is shorter, so you're covering less distance with the same number of steps.

The AW is only measuring the number of footfalls, not distance, for indoor running. It'll then estimate distance based on how many steps you took and an average of how long each step was when you ran outdoors.

[/mansplaining]

 

[Purga]

The AW is only measuring the number of footfalls, not distance, for indoor running. It'll then estimate distance based on how many steps you took and an average of how long each step was when you ran outdoors.

Wrong. The watch also uses your armswing (the magnitude of accelerometer peaks) to approximate the distance for each footfall based on outdoor calibration.

 

[Julien]

Wrong. The watch also uses your armswing (the magnitude of accelerometer peaks) to approximate the distance for each footfall based on outdoor calibration.

Arm swing and foot cadence are directly correlated when running or walking. So 30 arm swings equal about 60 steps. Cadence (steps) in NO way indicates stride length (distance). Stride length must be calibrated by running and walking outside with GPS. Even then it can and will often be different on a treadmill.

For instance if you are running 8 minute mile with a cadence of about 170 your stride length will be about 1.2m. However if you run the same 8 minute mile but with a stride length of about 1m your cadence will be about 205.

 

[Purga]

Arm swing and foot cadence are directly correlated when running or walking. So 30 arm swings equal about 60 steps. Cadence (steps) in NO way indicates stride length (distance). Stride length must be calibrated by running and walking outside with GPS. Even then it can and will often be different on a treadmill.

For instance if you are running 8 minute mile with a cadence of about 170 your stride length will be about 1.2m. However if you run the same 8 minute mile but with a stride length of about 1m your cadence will be about 205.

What you say would be right if it was only possible to detect the occurance of an arm swing. Then of course the number of arm swings just equals half the number of foot steps.

However, arm swings differ based on running speed. You can detect the "strength" of a stride or better say approximate the strength based on the sensor data (using the magnitude of the accelerometer data rather just using its cadence). This is of course different for individuals and pre-calibrated based on a lot of training data; but is also calibrates individually when you run outdoor.

So it is not only the stride length (which obviously differs based on running speed), but also arm swing characteristics that are getting calibrated and help in calculating the speed and hence the running distance

 

[BarracksSi]

Cadence (steps) in NO way indicates stride length (distance).

Right. Another way to illustrate it --

Say we've got two runners, one 5 feet tall and the other at 6 feet, with legs proportional to their size. Running together at the same tempo, with their feet landing at the same times, the short runner will feel like he's taking big, leaping, giant strides to keep up with the taller runner.

And/or, if they were both running at the same pace (8 min/mile), the shorter runner will have to turn over a higher cadence.

Part of why Usain Bolt is so fast is because he's so big. He's tall for a sprinter. But he's got such a fast turnover in his stride that he can gobble up those big steps in a hurry.

So it is not only the stride length (which obviously differs based on running speed), but also arm swing characteristics that are getting calibrated and help in calculating the speed and hence the running distance

That may be, but it's ignoring the differences between a fast, smooth-running runner with long strides and a violently arm-flapping runner with short strides. Pace is pace, and has nothing to do with Amount Of Flailing. It can probably help discern whether someone is running versus walking, but for recording cadence over distance, it's just simpler to measure the number of arm swings.

Running like Phoebe would not give good "arm swing quality" data at all:

Your post was so adamant, though. Does the AW actually use this data? What's the source of your info?

 

[Julien]

...However, arm swings differ based on running speed....

Not necessarily. Again arm swing is directly correlated to cadence. Using my example and comparing running an 8 minute mile to a 9 minute mile but with the same number of arm swings/cadence per minute.


8 minute mile with a cadence of 170 (85 arm swings) and a stride length of about 1.2m
9 minute mile with a cadence of 170 (85 arm swings) and a stride length of about 1m

Same number of arm swings/cadence per minute but 2 different running times/speeds based only on stride length differences.

Most experienced runners keep a more or less consistent cadence on flat terrain and very their speed based on stride length.

For kicks here are some stats from a 6 mile run I just completed. Would have been a better time but my last mile was up a major hill (875' in 6 miles ). Killer (for me) sub 7 minute 4th mile.

https://connect.garmin.com/modern/activity/1624134482

 

[Purga]

Say we've got two runners, one 5 feet tall and the other at 6 feet, with legs proportional to their size. Running together at the same tempo, with their feet landing at the same times, the short runner will feel like he's taking big, leaping, giant strides to keep up with the taller runner.

yes

And/or, if they were both running at the same pace (8 min/mile), the shorter runner will have to turn over a higher cadence.

yep

Part of why Usain Bolt is so fast is because he's so big. He's tall for a sprinter. But he's got such a fast turnover in his stride that he can gobble up those big steps in a hurry.

agree

That may be, but it's ignoring the differences between a fast, smooth-running runner with long strides and a violently arm-flapping runner with short strides. Pace is pace, and has nothing to do with Amount Of Flailing.

It can probably help discern whether someone is running versus walking, but for recording cadence over distance, it's just simpler to measure the number of arm swings.

You seem to think that part of finding a good algorithm is "keeping it simple" and only including the best and most obviously correlated variables, in this case step cadence (amount of arm swings) and average stride length.
Of course these variables are used and of course they mainly serve to calculate running distance. My point is that they are not exclusively being used.

Iam not taking about flailing, arm-flapping or whatever.
With "sensor data" i am referring to raw inertial data (3 axis accelerometer and gyroscope) which among others serves to calculate the cadence (the amount of arm swings). Running however is such a consistent rhythmic movement that fine differences can be extracted when you run at different speed. The arm swing of a runner running at 9 and 9,5km/h might look and feel the same to you, but there is fine differences in the raw sensor data that can be picked up.

This is a similar to electrocardiography (EKG). Think about the sinus rhythm of your heart beat. Sure you can just pick up and easily identify each heart beat, but you can also closer look at the fine details that may be relevant relating to a medical question.

Apple is most likely including all the data they can get to calculate running speed. After all there is a lot of machine-learning involved in finding the right variables. Before the release of the watch, Apple employed hundreds of people running on treadmills and doing activity just to fine-tune these algorithms. And each watch continues and fine tunes this algorithm whenever you run outdoor to better fit it to your individual characteristics.

Rather of a simple algorithm that just includes the number of strides and an average of stride length think about a more complicate algorithm that includes all the variables it can get or better say that machine-learning algorithms showed to improve the accuracy. These variables most likely span, yes, your average stride length and the number of strides but probably also your height, heart rate and very fine characteristics in your arm swing behaviour (after all you are pushing your body faster forward, which is somehow reflected in sensor data that is trying to measure exactly that: acceleration).

Your post was so adamant, though. Does the AW actually use this data? What's the source of your info?

Most likely Apple does not and will never expose their algorithm and iam not working for Apple. However, iam working in the field of wearable devices, building protoypes and working with such sensor data. So my info comes from personal experience with similar technology.

 

[AlliFlowers]

Damn that arm swing! It really screws things up for people who use crutches.

 

[imagineadam]

Thanks for all the replies!

So what I'm finding is that my watch is pretty accurate now actually after I've ran a bit more outside with it. Sometimes it's spot on with the distance of the treadmill depending on the speed I'm running at. Usually if I have the speed slower say 6mph on the TM my watch will count ahead a bit. But once I warm up and crank the speed up a bit more to 7 or 8mph the treadmill catches back up to the watch and sometimes passes it or it stays the same with the watch. It's usually right on around 7mph or 7.5.

So if my watch is a little behind near the end of a run I will just make my arms go a little crazy and I can usually get it to match up with the treadmill.

 

[BarracksSi]

yes


yep


agree


You seem to think that part of finding a good algorithm is "keeping it simple" and only including the best and most obviously correlated variables, in this case step cadence (amount of arm swings) and average stride length.
Of course these variables are used and of course they mainly serve to calculate running distance. My point is that they are not exclusively being used.

Iam not taking about flailing, arm-flapping or whatever.
With "sensor data" i am referring to raw inertial data (3 axis accelerometer and gyroscope) which among others serves to calculate the cadence (the amount of arm swings). Running however is such a consistent rhythmic movement that fine differences can be extracted when you run at different speed. The arm swing of a runner running at 9 and 9,5km/h might look and feel the same to you, but there is fine differences in the raw sensor data that can be picked up.

This is a similar to electrocardiography (EKG). Think about the sinus rhythm of your heart beat. Sure you can just pick up and easily identify each heart beat, but you can also closer look at the fine details that may be relevant relating to a medical question.

Apple is most likely including all the data they can get to calculate running speed. After all there is a lot of machine-learning involved in finding the right variables. Before the release of the watch, Apple employed hundreds of people running on treadmills and doing activity just to fine-tune these algorithms. And each watch continues and fine tunes this algorithm whenever you run outdoor to better fit it to your individual characteristics.

Rather of a simple algorithm that just includes the number of strides and an average of stride length think about a more complicate algorithm that includes all the variables it can get or better say that machine-learning algorithms showed to improve the accuracy. These variables most likely span, yes, your average stride length and the number of strides but probably also your height, heart rate and very fine characteristics in your arm swing behaviour (after all you are pushing your body faster forward, which is somehow reflected in sensor data that is trying to measure exactly that: acceleration).


Most likely Apple does not and will never expose their algorithm and iam not working for Apple. However, iam working in the field of wearable devices, building protoypes and working with such sensor data. So my info comes from personal experience with similar technology.

K.

 

[BarracksSi]

Apple is most likely ...
[snip]


Most likely Apple does not and will never expose their algorithm and iam not working for Apple. [snip again]

So you don't know what Apple is doing to measure total distance.