Step accuracy

[jfdavis17]

I would like to know has anyone tested the step accuracy, and what they found. I have been comparing mine to the garmin vivofit and the steps are way below what the vivofit gives me. I am not saying the Band is wrong, but either it is counting low or my vivofit is counting high.

What I am seeing is when I get about 10,000 steps with the vivofit, I only have about 8000 with the Band give or take 500 steps.

Anyway, others observations would be greatly appreciated.

 

[DroidUser42]

Seems like the thing to do is figure out a way to count your steps and see which one is wrong. Either a cheap waist-mounted pedometer or a measured course.

However, I tend to suspect the difference is in the areas of various arm movements during the day that may or may not be a "step".

 

[kittengirl]

I have a Vivofit, but have not compared the two. However, I know from using my Vivofit at the same time as my Garmin GPS, that for distance it's pretty inaccurate, usually overstating by quite a bit when walking.

 

[nmercy]

My FitBit One shows about a 1,000 steps more daily than my Band. It seems that the algorithm doesn't pick up small trips, only longer trips... so like going to the bathroom/kitchen it doesn't pick up, but when I go for a walk it keeps pace with the One.

 

[Richard Servello]

The band is smarter than those devices. It doesn't count bounces, or bouncy car rides. It counts actual steps. Other pedometers count a step each time it bumps.

 

[invisik]

I wonder about this as well. I've noticed a few times while I'm driving the Band will vibrate and say I've met my steps goal for the day. Other times I'll look at it mid-day at the office where I know I've been sitting the whole time, and it will say 1000+ steps.

On the flip side, when I do a "Run" on the treadmill at the club, it is accurate within 0.01 of a mile for distance to what the treadmill screen says. This is a different mode I realize, but the concept should be similar and is hyper-accurate somehow.

One day I'll try to check the Bands step count against manually counting my steps around the office and see how it compares. I do like the thing overall...

-m

 

[nmercy]

It is true that is supposed to be smarter, but the day after I got mine, I woke up went to bathroom, went down stairs, got my breakfast, sat down to eat and it had registered 0 steps. The algorithm for determining false steps definitely needs work. As I said, it's fine once it's determined you're walking, but after it has it's very accurate.

 

[DroidUser42]

I wonder about this as well. I've noticed a few times while I'm driving the Band will vibrate and say I've met my steps goal for the day.

What position was your arm in? I haven't really looked into it, but it seems to stop counting if I hold my arm horizontally while checking the step count.

 

[gadgetrants]

On the flip side, when I do a "Run" on the treadmill at the club, it is accurate within 0.01 of a mile for distance to what the treadmill screen says. This is a different mode I realize, but the concept should be similar and is hyper-accurate somehow.

-m

This is pretty interesting -- I wonder if the accelerometer and gyrometer are doing the work here? I've gotten pretty good distance estimates on a few indoor runs, but only when I'm going at a slow pace.

​-Matt

 

[sjjones]

I have the Fitbit force and band yesterdays numbers
Fitbit: 22,855 steps
Band:28,800

I have a treadmill desk and the display on that was closer to the Fitbit ......................................................

 

[eyecrispy]

I tested it on a 2.5 mile treadmill run this morning.
Both my Up24 and my Band were around 5600 steps at the end of it (including a walking cool down). It was close enough for my liking. That said, I have not calibrated my Band yet (have had it less than 24 hours). It has been too dang cold for me to run outside.

 

[Richard Servello]

I tested it on a 2.5 mile treadmill run this morning.
Both my Up24 and my Band were around 5600 steps at the end of it (including a walking cool down). It was close enough for my liking. That said, I have not calibrated my Band yet (have had it less than 24 hours). It has been too dang cold for me to run outside.

I've had no issues...but I run 2 times a week and also use GPS when we go out walking on the weekend.

 

[onlysublime]

You need to calibrate your Microsoft Band: Here's how - Neowin

Apparently, we have to calibrate the Band's step count by activating the GPS. It uses the GPS data and our profile to determine what is a step. Part of the "machine learning."

To activate the GPS, go to the Run tile and start a "run" and walk around for awhile with the GPS on.

Try it out!!! rly:

 

[mijacs]

I use a fit bit zip and for a while the Band would be about 1000 steps lower. Now they are pretty close. I know things like using shopping carts will not count with the Band. The Band seems to be fairly aggressively filtering out non steps. For example, someone with the flex can rock in a chair and it will count but the Band won't count it.

 

[eyecrispy]

I use a fit bit zip and for a while the Band would be about 1000 steps lower. Now they are pretty close. I know things like using shopping carts will not count with the Band. The Band seems to be fairly aggressively filtering out non steps. For example, someone with the flex can rock in a chair and it will count but the Band won't count it.

Yes. The band is much more aggressive filtering out nonsteps. Almost too aggressive. My Up24 registers steps when I'm walking and not swinging my arms while the band does not. Kind of annoying. They need to update it with the next firmware.

 

[RiftPoint]

I calibrated mine on Sunday with a 6.7 mile run, and it seemed very accurate then. But when I did a 15min ~2mi run on the treadmill this morning it registered me as running almost 2min slower... It seems to be registering steps no problem, but how it's calculating it's step distance seems to need a little adjusting. It would be nice (and probably help them calibrate step accuracy) if you could manually correct the band with the treadmill's results...

 

[eyecrispy]

I calibrated mine on Sunday with a 6.7 mile run, and it seemed very accurate then. But when I did a 15min ~2mi run on the treadmill this morning it registered me as running almost 2min slower... It seems to be registering steps no problem, but how it's calculating it's step distance seems to need a little adjusting. It would be nice (and probably help them calibrate step accuracy) if you could manually correct the band with the treadmill's results...

Agree. They really need to allow manual correction.

 

[DroidUser42]

Where you running the same way? If I was designing it, I'd probably have calibrations for walk, jog and run as I'd think each one would have a different stride.

 

[RiftPoint]

Where you running the same way? If I was designing it, I'd probably have calibrations for walk, jog and run as I'd think each one would have a different stride.

As far as I know I was running the same way. I'd say the only differences would be no elevation variation and I did some interval work as well. I had seen someone post somewhere in another post that it seemed accurate at a slower pace of around 11min/mi, but when they got up to faster than an 8min/mi pace the accuracy was off.

 

[DroidUser42]

I'd think the only way the device could tell which "stride" table to use would be the step frequency. But that raises the question about hills. On a hill I'd think you'd decrease stride along with step frequency going up but that may give a different result than simply slowing down on flat ground. The more I think about it, the messier it gets.