Mi Band 10 - Scientific Review (AMAZING!)

00:38:21
https://www.youtube.com/watch?v=qnriFKto38U

概要

TLDRIn this video, Rob, a post-doctoral scientist, tests the Xiaomi Mi Band 10 for heart rate and sleep tracking. He unboxes two bands, connects them to different phones, and conducts initial tests including fake workouts and real exercises like indoor cycling and weightlifting. The heart rate tracking shows promising results with a correlation of 0.99 compared to a Polar H10 chest strap. However, sleep tracking results are inconsistent between the two bands, with one performing better than the other. Overall, the Mi Band 10 performs well for heart rate tracking but lacks reliability in sleep stage tracking.

収穫

  • 📦 Unboxing two Mi Band 10s for testing.
  • 📱 Connecting to different phones and accounts.
  • 💪 Initial tests include fake workouts and real exercises.
  • 📊 Heart rate tracking shows a correlation of 0.99 with Polar H10.
  • 😴 Sleep tracking results are inconsistent between the two bands.
  • 💰 Mi Band 10 is priced around 40-50 dollars.
  • 👍 Good option for heart rate tracking.
  • ❌ Not reliable for sleep stage tracking.
  • 🔄 More tests planned for consistency.
  • 📈 Overall, promising results for a budget device.

タイムライン

  • 00:00:00 - 00:05:00

    Rob introduces the video, explaining that he will test the heart rate and sleep tracking capabilities of the Xiaomi Mi Band 10. He mentions his background as a post-doctoral scientist and outlines the plan to unbox and set up the devices for testing.

  • 00:05:00 - 00:10:00

    After unboxing the Mi Bands, Rob connects them to two different iPhones and Xiaomi accounts. He discusses previous syncing issues with Xiaomi devices and hopes to successfully sync the heart rate data to the Strava app for analysis.

  • 00:10:00 - 00:15:00

    Rob performs a quick indoor cycling workout to test the heart rate tracking of both Mi Bands. He notes the initial heart rate readings and mentions the need to reconnect the devices to Strava for data syncing.

  • 00:15:00 - 00:20:00

    Following the indoor cycling session, Rob analyzes the heart rate data from both Mi Bands compared to a Polar H10 chest strap. He finds a high correlation of 0.99, indicating that the Mi Bands performed well in tracking heart rate during the workout.

  • 00:20:00 - 00:25:00

    Rob transitions to weightlifting, noting that tracking heart rate during this activity is more challenging. He compares the heart rate data from both Mi Bands and finds that while they performed decently, they struggled to accurately capture peak heart rates during intense lifts.

  • 00:25:00 - 00:30:00

    Rob then tests the Mi Bands during outdoor cycling, analyzing the heart rate data from two rides. He finds that the Mi Bands performed well, with correlation values of 0.91 and 0.96, indicating good tracking accuracy compared to reference devices.

  • 00:30:00 - 00:38:21

    Finally, Rob discusses the sleep tracking performance of the Mi Bands, revealing discrepancies between the two devices. He compares their sleep stage data to an EEG device and concludes that the Mi Bands did not perform well in this aspect, particularly in tracking REM and light sleep stages. He summarizes the overall findings, highlighting the Mi Band 10's strong heart rate tracking but poor sleep tracking capabilities.

もっと見る

マインドマップ

ビデオQ&A

  • What is the main focus of the video?

    The video focuses on testing the heart rate and sleep tracking capabilities of the Xiaomi Mi Band 10.

  • Who is the presenter of the video?

    The presenter is Rob, a post-doctoral scientist specializing in biological data analysis.

  • How did the Mi Band 10 perform in heart rate tracking?

    The Mi Band 10 showed promising results in heart rate tracking with a correlation of 0.99 compared to a Polar H10 chest strap.

  • What were the results for sleep tracking?

    The sleep tracking results were inconsistent between the two Mi Band 10s, with one performing better than the other.

  • Is the Mi Band 10 a good option for heart rate tracking?

    Yes, for its price point, the Mi Band 10 performs well in heart rate tracking.

  • What are the limitations of the Mi Band 10?

    The Mi Band 10 lacks reliability in sleep stage tracking.

  • What other devices were mentioned in the video?

    Other devices mentioned include the Polar H10 chest strap, Whoop Strap, and Huawei Watchfit.

  • What is the price range of the Mi Band 10?

    The Mi Band 10 is priced around 40 to 50 dollars.

  • What does Rob plan to do next with the Mi Band 10?

    Rob plans to conduct more tests and possibly have others test the device for consistency.

  • What is the overall conclusion about the Mi Band 10?

    The Mi Band 10 is a good option for heart rate tracking but not reliable for sleep stage tracking.

ビデオをもっと見る

AIを活用したYouTubeの無料動画要約に即アクセス！
字幕
en
オートスクロール:
  • 00:00:00
    I just ordered and received two Mi Band
  • 00:00:02
    10 and today we're going to
  • 00:00:04
    systematically and scientifically test
  • 00:00:06
    them to find out how good the heart rate
  • 00:00:08
    tracking and the sleep state tracking is
  • 00:00:10
    of these cheap Xiaomi watches. By the
  • 00:00:12
    way, for those of you who don't know me,
  • 00:00:14
    my name is Rob and I'm a post-doal
  • 00:00:15
    scientist specializing in biological
  • 00:00:17
    data analysis. Now, this video is going
  • 00:00:19
    to be a bit different from my normal
  • 00:00:20
    ones in two ways. First of all, this is
  • 00:00:23
    going to be an initial test, just a few
  • 00:00:25
    exercises, one or two nights of sleep
  • 00:00:27
    maximum, just to get an initial
  • 00:00:28
    impression of how well these watches can
  • 00:00:30
    track your heart rate and your sleep.
  • 00:00:32
    And second, we're going to together go
  • 00:00:34
    on the journey to install these watches,
  • 00:00:36
    see if we can actually get to the data,
  • 00:00:38
    because that's one problem with Xiaomi
  • 00:00:40
    devices. Often, it's very hard to
  • 00:00:42
    actually get to the raw data and do the
  • 00:00:44
    analysis I like to do. So maybe we'll
  • 00:00:46
    have a lot of trouble and you'll have
  • 00:00:48
    the same frustrations as me, but you'll
  • 00:00:50
    also be informed that it's hard to
  • 00:00:51
    actually sometimes get to this data. Now
  • 00:00:53
    to get started, we need to unbox them
  • 00:00:54
    and connect them to two different phones
  • 00:00:56
    and two different Xiaomi accounts. So
  • 00:00:58
    let's do that. So I have two iPhones
  • 00:01:00
    here, both with a separate Mi Fitness
  • 00:01:03
    account, both syncing to both Apple
  • 00:01:05
    Health, also two separate Straa
  • 00:01:07
    accounts. So there I should be able to
  • 00:01:09
    get the heart rate data. And this has
  • 00:01:11
    been an issue in the past. Sometimes for
  • 00:01:13
    some reason it just doesn't sync. So
  • 00:01:15
    before I actually tested both the Xiaomi
  • 00:01:17
    Watch and the Xiaomi Smart Band 9 Pro
  • 00:01:20
    and only for the Band 9 Pro did I
  • 00:01:22
    actually get the heart rate data on
  • 00:01:23
    Straa and for some reason it just didn't
  • 00:01:26
    sync at all for the Xiaomi Watch. So I'm
  • 00:01:28
    actually reusing the account I had for
  • 00:01:30
    the Band 9 Pro and I created a new
  • 00:01:32
    account on a new phone for one of the
  • 00:01:34
    other bands. So both should hopefully
  • 00:01:36
    now sync to Straa. Let's first quickly
  • 00:01:39
    unbox both devices. So I actually
  • 00:01:41
    ordered two different colors so we can
  • 00:01:43
    tell them apart. a white one and a black
  • 00:01:46
    one. Well, the boxes are simple as
  • 00:01:49
    always. We have the band and the charger
  • 00:01:52
    right here. Still using the old USBA. It
  • 00:01:55
    is what it is. And then we also have our
  • 00:01:59
    black one right here. Same setup again.
  • 00:02:03
    USB typeA
  • 00:02:05
    and the black version.
  • 00:02:11
    So, let's first set up the black one. As
  • 00:02:14
    always, the bezel is pretty good. So,
  • 00:02:16
    let's set up the black one on my gold
  • 00:02:19
    iPhone.
  • 00:02:22
    So, that was easy enough. A quick update
  • 00:02:25
    and then at the same time, we can
  • 00:02:27
    install the other one. Here we have the
  • 00:02:31
    white version. So, here you can see
  • 00:02:33
    previously I had the Xiaomi Band 8 9 Pro
  • 00:02:36
    and another 9 Pro connected, but we're
  • 00:02:39
    actually going to connect a new device.
  • 00:02:44
    All right, so both devices are now
  • 00:02:47
    updating. Let's wait for that to be done
  • 00:02:50
    and then we're going to do a test
  • 00:02:51
    exercise. Just a 5 minute fake exercise
  • 00:02:54
    to see if the data will sync to Straa.
  • 00:02:56
    And if that works, it's time to get to
  • 00:02:57
    the real testing. But now, as always,
  • 00:02:59
    since I have to wait for those updates
  • 00:03:01
    to get done, I can make myself a quick
  • 00:03:04
    coffee and we can enjoy that while we're
  • 00:03:06
    waiting.
  • 00:03:08
    [Music]
  • 00:03:21
    Okay, so the update is done. The coffee
  • 00:03:24
    is ready.
  • 00:03:27
    And now I'll start two fake workouts
  • 00:03:28
    just to see if that data syncs. So let's
  • 00:03:30
    do that now. That also means I cannot
  • 00:03:32
    test the Vivo Active 6 from Garmin for
  • 00:03:34
    at least today and tomorrow. And I'm
  • 00:03:36
    actually also testing another device you
  • 00:03:38
    guys are very interested in, the
  • 00:03:40
    Amazefitit Helios band. So we have the
  • 00:03:43
    strap of it right here. So this is the
  • 00:03:45
    strap part, but I'm actually going to be
  • 00:03:47
    able to wear it at the same time because
  • 00:03:49
    I hacked that together by using the
  • 00:03:51
    Whoop Strap 4.0 biceps. I don't know
  • 00:03:54
    what you call these things. like a bicep
  • 00:03:56
    sleeve or something. So, I'll be wearing
  • 00:03:58
    that as well for the coming weeks in
  • 00:04:00
    addition to a second Helios band I'll be
  • 00:04:02
    wearing on my wrist and then we'll also
  • 00:04:03
    do that review in a week or two. I'm
  • 00:04:05
    very excited about this one. This could
  • 00:04:07
    be a whoop replacement without a
  • 00:04:09
    subscription. So, wait for that as well.
  • 00:04:10
    If you're interested, subscribe to the
  • 00:04:12
    channel. But for now, of course, the
  • 00:04:14
    next two days we're going to focus on
  • 00:04:15
    the Xiaomi Mi Band 10. Let's first try
  • 00:04:17
    an indoor workout because sometimes
  • 00:04:20
    those don't sync to Straa. There's some
  • 00:04:21
    weird things going on. Again, you have
  • 00:04:24
    to wear the device a bit away from the
  • 00:04:26
    wrist. So, that should be okay. So,
  • 00:04:28
    we'll go.
  • 00:04:30
    And also for the second device, same
  • 00:04:33
    information. Let's go. Okay. So, we'll
  • 00:04:37
    wait 5 minutes or so. You can see the
  • 00:04:40
    heart rate is not even being detected
  • 00:04:42
    yet. I don't know if you can see that
  • 00:04:43
    screen right now. So, as you can see,
  • 00:04:46
    it's not detecting my heart rate yet,
  • 00:04:48
    which is weird. Let's see if my heart
  • 00:04:50
    rate will pop up. Yes. 76 73 which is
  • 00:04:53
    reasonable I would say. And for the
  • 00:04:56
    other Mi Band, will the heart rate pop
  • 00:04:58
    up? 96. That's probably a bit too high.
  • 00:05:02
    5 minutes later. Okay, so we're 5
  • 00:05:04
    minutes in the moment of truth. Let's
  • 00:05:08
    stop both the watches. So, the black one
  • 00:05:13
    and the white one. Okay, one
  • 00:05:17
    worked. The other one worked as well.
  • 00:05:20
    Where is strong? Here we go. Activities.
  • 00:05:25
    No, no Sava drink. No, no Straa sync.
  • 00:05:30
    Let's try again. So, we did a freestyle
  • 00:05:32
    workout. It recorded my heart rate with
  • 00:05:35
    the Xiaomi Ben 10, but it's not on
  • 00:05:37
    Straa. Let's reconnect, maybe.
  • 00:05:42
    [Music]
  • 00:05:45
    And we'll be back in another 5 minutes.
  • 00:05:47
    One long angry line later. All right,
  • 00:05:50
    we're back. Five minutes later. Let's
  • 00:05:52
    end the exercise. Whoopsie. Let's check
  • 00:05:56
    Straa. Yes.
  • 00:05:59
    Okay, apparently I needed to reconnect
  • 00:06:01
    or maybe for some reason was connected
  • 00:06:03
    to a different account. All seems to
  • 00:06:06
    work now. Okay, that's good. All seems
  • 00:06:08
    to work now. So, my plan is to first do
  • 00:06:11
    maybe an indoor cycling session, then a
  • 00:06:14
    short weightlifting session, probably
  • 00:06:16
    chest and triceps. Then I actually have
  • 00:06:18
    a party to go to, but I'll cycle there.
  • 00:06:20
    So, we'll get some cycling data back and
  • 00:06:22
    forth. And then, yeah, that's it. Maybe
  • 00:06:24
    tomorrow I'll go for a run and then we
  • 00:06:26
    have a bunch of data to look at. Of
  • 00:06:28
    course, it's a very initial test, but it
  • 00:06:29
    will show us for two independent Xiaomi
  • 00:06:32
    Mi Band 10s if they're good at tracking
  • 00:06:33
    my heart rate. So, let's get started.
  • 00:06:35
    So, we'll start with indoor cycling
  • 00:06:38
    right here, which should be relatively
  • 00:06:40
    easy for a device to track. So, let's
  • 00:06:42
    see if the Xiaomi Mi Band 10 can track
  • 00:06:44
    my heart rate during indoor cycling.
  • 00:06:47
    [Music]
  • 00:06:54
    Okay, that was quite an intense workout.
  • 00:06:56
    Just looking at my wrist a few times,
  • 00:06:58
    checking the Mi Band 10's heart rate,
  • 00:07:00
    and I compared that to the chest strap
  • 00:07:02
    by Garmin. Actually, that looked very
  • 00:07:03
    similar. So, that looks good at a first
  • 00:07:05
    glance, but we're actually going to do
  • 00:07:06
    formal analysis and do some coding on my
  • 00:07:08
    computer. Now, just one thing I want to
  • 00:07:10
    mention, for some reason, Garmin always
  • 00:07:12
    gives me a way too high V2 max. So, it
  • 00:07:15
    says I have a V2 max of 68, which is way
  • 00:07:18
    below what I actually have. I have a V2
  • 00:07:19
    max of 58 according to my last
  • 00:07:21
    measurement, but that's a side note.
  • 00:07:23
    Usually, the Garmin can measure my heart
  • 00:07:24
    rate quite accurately, but we're going
  • 00:07:26
    to use the Polar H10 as a reference in
  • 00:07:28
    this case. But from what I've seen,
  • 00:07:30
    basically, whatever chest strap you get,
  • 00:07:31
    you get a good heart rate. So, use the
  • 00:07:33
    chest strap by Polar as a reference and
  • 00:07:35
    see how good the Mi Band 10 actually
  • 00:07:36
    did. Okay, the results are in for indoor
  • 00:07:39
    cycling and the results are looking
  • 00:07:41
    surprisingly good. Now in this plot you
  • 00:07:44
    see the overview for the Mi Band 10 worn
  • 00:07:46
    on my left wrist with along the
  • 00:07:48
    horizontal axis the reference device. So
  • 00:07:50
    the Polar H10 chest strap and along the
  • 00:07:52
    vertical axis the Mi Band 10 where each
  • 00:07:54
    dot is a matching measurement between
  • 00:07:56
    the two of them. And if they would
  • 00:07:58
    perfectly agree all values should be
  • 00:08:00
    along this blue line right here. And as
  • 00:08:02
    you can see almost all points are on or
  • 00:08:05
    at least super close to the blue line.
  • 00:08:07
    And for such a cheap device as the Mi
  • 00:08:09
    Band this is really good. We can
  • 00:08:12
    actually quantify this agreement with a
  • 00:08:14
    correlation, this R value up here, which
  • 00:08:16
    is 0.99 for the one on my left wrist.
  • 00:08:19
    So, super good. I'm really happy with
  • 00:08:21
    this. And we can look at this same thing
  • 00:08:22
    for the device on my right wrist. And
  • 00:08:24
    those results are right here. Now,
  • 00:08:26
    they're not identical. There's this
  • 00:08:28
    little bit of deviation right here, but
  • 00:08:31
    overall, again, a correlation of 0.99,
  • 00:08:34
    which is really good, and almost all
  • 00:08:36
    points on or close to the blue line.
  • 00:08:39
    Now, let's take a look at the individual
  • 00:08:40
    sessions to see why there's still a bit
  • 00:08:42
    of deviations, but this result really
  • 00:08:44
    surprises me so far. Along the
  • 00:08:46
    horizontal axis, we have the clock time
  • 00:08:47
    and my heart rate along the vertical
  • 00:08:49
    axis with in blue green the reference
  • 00:08:51
    device and the Mi Band 10 from my left
  • 00:08:53
    wrist in red and the red line and the
  • 00:08:55
    blue line basically overlap perfectly.
  • 00:08:58
    Like there's a little bit of deviation
  • 00:08:59
    here in the beginning and maybe
  • 00:09:01
    sometimes in the lower heart rate range,
  • 00:09:03
    a tiny bit of deviation, but overall
  • 00:09:06
    definitely good enough performance. And
  • 00:09:08
    if we look at the one on my right wrist,
  • 00:09:10
    this is more or less the same. So a very
  • 00:09:12
    good agreement. Only in this moment
  • 00:09:15
    right here, they detect a slightly too
  • 00:09:17
    high heart rate, but just for a little
  • 00:09:18
    bit. Overall, this is really good and
  • 00:09:21
    good enough for me. And we can actually
  • 00:09:22
    put these results into the context of
  • 00:09:24
    many of the other devices I've tested
  • 00:09:25
    before. And those results are displayed
  • 00:09:27
    right here. So that correlation value
  • 00:09:29
    we're looking at before is along the
  • 00:09:30
    horizontal axis. And on the vertical
  • 00:09:33
    axis, I ordered the watches from worst
  • 00:09:35
    on the bottom to best on top. So the
  • 00:09:37
    further to the right and the higher
  • 00:09:39
    devices, the better is its agreement
  • 00:09:40
    with the reference device. And here I
  • 00:09:42
    marked the Mi Band 10 on my left wrist
  • 00:09:44
    in red. And we'll zoom in a bit in a
  • 00:09:46
    moment so you can also see where the one
  • 00:09:47
    on my right wrist is. And here we have
  • 00:09:49
    that zoomed in view. So these are just
  • 00:09:50
    the devices with a correlation of 0.9 or
  • 00:09:53
    higher. So some of the better devices
  • 00:09:55
    with again the one on my left wrist
  • 00:09:57
    marked in red right here and the one on
  • 00:09:59
    my right wrist right here. So they're
  • 00:10:01
    super close to each other and among some
  • 00:10:03
    of the better devices out there. They're
  • 00:10:05
    really doing well. Now, this is a very
  • 00:10:07
    limited initial test, so just a single
  • 00:10:09
    spinning session, but the fact that both
  • 00:10:11
    of them performed really well is at
  • 00:10:12
    least promising. So, overall, I'm quite
  • 00:10:15
    surprised and really happy with this.
  • 00:10:17
    Let's see if we can see some of the
  • 00:10:18
    other Mi Bands in this overview. So, we
  • 00:10:20
    have the Mi Band 8 right here. So, quite
  • 00:10:22
    a bit worse. Again, I will probably have
  • 00:10:25
    done more testing with that one. And the
  • 00:10:27
    Mi Band 9 is in this overview. And
  • 00:10:30
    indeed, the Mi Band 9 also didn't do
  • 00:10:31
    that poorly. It's right here. is also
  • 00:10:34
    quite close to the Mi Band 10. So, I
  • 00:10:36
    would say the Mi Band 10 might be doing
  • 00:10:37
    a tiny bit better than the Mi Band 9,
  • 00:10:39
    but overall really good performance. But
  • 00:10:42
    let's next take a look at a much harder
  • 00:10:44
    exercise for this device to track
  • 00:10:46
    weightlifting. Well, for any device to
  • 00:10:48
    track. So, with weightlifting, there's a
  • 00:10:49
    lot of tension on your arms. And this
  • 00:10:51
    makes it very hard for any tracker out
  • 00:10:53
    there to accurately track your heart
  • 00:10:55
    rate on the wrist. So, let's do that
  • 00:10:56
    now. Afterwards, we'll take a look at
  • 00:10:58
    cycling outside. I actually need to fix
  • 00:10:59
    one of my bikes back there. We'll fix
  • 00:11:01
    that and then go for cycling. But let's
  • 00:11:03
    first do a weightlifting session. And
  • 00:11:05
    here we have a similar overview to
  • 00:11:06
    before, but now for weightlifting with
  • 00:11:08
    again the reference device along the
  • 00:11:10
    horizontal axis and the Mi Band 10 on my
  • 00:11:13
    left wrist along the vertical axis. And
  • 00:11:15
    we do see a typical pattern I would say.
  • 00:11:18
    So in the lower heart rate range almost
  • 00:11:20
    all points are on or at least super
  • 00:11:22
    close to the blue line. So there the
  • 00:11:24
    measurements are okay. But here in the
  • 00:11:26
    higher heart rate range, there are many
  • 00:11:28
    points below the blue line, indicating
  • 00:11:30
    that in these moments, the Mi Band 10
  • 00:11:31
    detected a too low heart rate. Now, this
  • 00:11:33
    generally means that when I have a peak
  • 00:11:35
    in my heart rate, so there's a lot of
  • 00:11:37
    tension on my arm, the Mi Band 10 cannot
  • 00:11:39
    detect my heart rate accurately. But
  • 00:11:40
    let's now take a look at the one on my
  • 00:11:42
    right wrist. And here we have those
  • 00:11:44
    results. And these are mostly the same,
  • 00:11:46
    though even a little bit better. There
  • 00:11:47
    are more points in the higher heart rate
  • 00:11:49
    range close to the blue line, but we
  • 00:11:51
    still see many points below the blue
  • 00:11:53
    line right here. The correlation value,
  • 00:11:55
    that value you're looking at before is
  • 00:11:56
    now 0.92 for the one on my right wrist
  • 00:12:00
    and 0.89 for the one on my left wrist.
  • 00:12:02
    And this is actually quite decent
  • 00:12:04
    compared to the competition. But let's
  • 00:12:06
    first take a look at the individual
  • 00:12:07
    session and then we'll compare to the
  • 00:12:09
    competition. And here are the results
  • 00:12:10
    for the Mi Band 10 on my left wrist with
  • 00:12:13
    again the Mi Band 10 in red and the
  • 00:12:14
    reference device in blue green. Now,
  • 00:12:16
    each time I did a set of exercise, my
  • 00:12:18
    heart rate showed a peak and you can see
  • 00:12:20
    that clearly by these blue lines right
  • 00:12:22
    here. And it is pretty obvious that the
  • 00:12:24
    Mi Band 10 often couldn't fully detect
  • 00:12:26
    the peaks in my heart rate, at least for
  • 00:12:28
    the one on my left wrist. And if we look
  • 00:12:29
    at the results for the one on my right
  • 00:12:31
    wrist, these are potentially a bit
  • 00:12:32
    better. Some of the peaks were detected.
  • 00:12:35
    I would say these two, so this one right
  • 00:12:36
    here, this one right here, and also this
  • 00:12:38
    first one actually were more or less
  • 00:12:40
    fully detected, but many peaks were also
  • 00:12:43
    missed. So I wouldn't say this is good
  • 00:12:44
    enough. But compared to the competition,
  • 00:12:47
    this actually isn't that bad. So let's
  • 00:12:49
    take a look at that. And here we have
  • 00:12:50
    those results. So a similar overview to
  • 00:12:52
    before where we want the devices to be
  • 00:12:54
    as far to the top right as possible with
  • 00:12:56
    the Mi Band 10 on my right is marked in
  • 00:12:59
    red right here with that correlation of
  • 00:13:01
    0.92. It's actually pretty decent among
  • 00:13:03
    some of the better devices out there.
  • 00:13:05
    And if we zoom in a bit, we can read
  • 00:13:07
    those labels even better. These are all
  • 00:13:09
    devices with a correlation of 0.7 or
  • 00:13:11
    higher. And here we have the Mi Band 10
  • 00:13:13
    on my right wrist. And a little bit
  • 00:13:15
    below that right here is the Mi Band 10
  • 00:13:17
    on my left wrist. And they're super
  • 00:13:19
    close actually. So more or less giving a
  • 00:13:21
    similar result. I do see some difference
  • 00:13:23
    sometimes in performance for devices on
  • 00:13:25
    my left and right wrist where actually
  • 00:13:27
    in general my left wrist seems to be a
  • 00:13:29
    bit easier. For the Mi Band I cannot see
  • 00:13:31
    any difference so far. And with this
  • 00:13:33
    correlation it is among some of the
  • 00:13:34
    better devices out there. Maybe doing
  • 00:13:37
    about as well as many Huawei devices.
  • 00:13:39
    Not quite as good as Apple watches.
  • 00:13:41
    These are some of the best devices out
  • 00:13:43
    there for heart rate tracking. But for
  • 00:13:45
    the price, this is honestly really good.
  • 00:13:47
    I'm not disappointed with this. Now, for
  • 00:13:49
    weightlifting, I still wouldn't use it.
  • 00:13:51
    I generally wouldn't use any device on
  • 00:13:53
    the wrist except for maybe the Apple
  • 00:13:55
    Watch. But if you want to track your
  • 00:13:56
    heart rate during weightlifting, I
  • 00:13:57
    really recommend using a chest strap.
  • 00:13:59
    But let's next take a look at cycling.
  • 00:14:01
    That will be the most important test to
  • 00:14:03
    do because that's quite a bit harder
  • 00:14:05
    than cycling indoors and also quite
  • 00:14:07
    important for many people, but not quite
  • 00:14:09
    as hard as weightlifting. But before
  • 00:14:11
    doing that, I actually need to fix one
  • 00:14:12
    of my bikes. So, I have a bunch of bikes
  • 00:14:14
    right here. Actually, have a few more
  • 00:14:16
    bikes. I actually need to fix the
  • 00:14:18
    tension on the spokes of the back wheel
  • 00:14:20
    right here. I already partially fixed it
  • 00:14:22
    before, but there's still a small issue.
  • 00:14:24
    I need to replace the pedals right
  • 00:14:26
    there. And then hopefully it'll be ready
  • 00:14:28
    to ride. I actually used that bike for
  • 00:14:29
    many years for my road biking. My
  • 00:14:32
    friends had all these fancy bikes and I
  • 00:14:33
    was still on my old steel bike with
  • 00:14:35
    frame shifters. It still worked. It was
  • 00:14:38
    a bit of a challenge keeping up with
  • 00:14:39
    them, but good for my fitness. I now
  • 00:14:41
    have a somewhat newer bike, still from
  • 00:14:43
    the 2000s, I think. Let me quickly show
  • 00:14:45
    you that one. This is actually my
  • 00:14:46
    current road bike right here. It's from
  • 00:14:48
    the early 2000s. I think back then it
  • 00:14:51
    was an amazing bike, but it's still
  • 00:14:53
    mostly aluminium. So, if there are any
  • 00:14:55
    bike brands out there who want to
  • 00:14:56
    sponsor me, I'm a really keen cyclist. I
  • 00:14:59
    cycle a lot, but on quite old bikes and
  • 00:15:01
    I'm happy to try out some carbon bikes
  • 00:15:03
    sometimes. Decathlon for instance, I
  • 00:15:05
    send you a message. I love your VR rial
  • 00:15:07
    stuff. I have a lot of your kit. So, if
  • 00:15:09
    you ever want to have me try one of your
  • 00:15:10
    bikes, I'd love to. But I'd be happy
  • 00:15:12
    with any brand that produces quality
  • 00:15:14
    carbon frames.
  • 00:15:16
    [Music]
  • 00:15:20
    All right, I'm back from my bike rides.
  • 00:15:22
    I already analyzed the data and it's
  • 00:15:24
    super interesting. There were just two
  • 00:15:26
    short bike rides, both about 30 35
  • 00:15:28
    minutes or so, but the data is pretty
  • 00:15:30
    clear, I would say. And with two
  • 00:15:32
    watches, we basically have four rides to
  • 00:15:34
    look at. So, let's take a look. And
  • 00:15:36
    sorry I'm not wearing a shirt in some of
  • 00:15:37
    this video, but it's about 27° C now in
  • 00:15:41
    my room. Way too hot for me. I don't
  • 00:15:43
    know what it is in Fahrenheit. Alex can
  • 00:15:44
    put it here. But let's take a look at
  • 00:15:46
    the data now. Instead of jumping
  • 00:15:47
    straight to the correlations, I first
  • 00:15:49
    want to take a look at the bike rides
  • 00:15:51
    themselves. This is the first ride right
  • 00:15:53
    here with the reference in blue green
  • 00:15:55
    and the Mi Band 10 worn on my right
  • 00:15:57
    wrist in red. And this looks really
  • 00:15:59
    good. Honestly, the Mi Band 10 mostly
  • 00:16:02
    was able to follow along with the
  • 00:16:03
    changes in my heart rate. Yes, some of
  • 00:16:05
    the details are missing, but overall
  • 00:16:08
    looking pretty good. This would be good
  • 00:16:10
    enough for me. And let's now take a look
  • 00:16:11
    at this same ride, but now for the one
  • 00:16:13
    on my left wrist. And those results are
  • 00:16:16
    right here. Also looking pretty good. It
  • 00:16:18
    makes some different mistakes, but
  • 00:16:20
    overall more or less the same
  • 00:16:22
    performance, I would say. And the second
  • 00:16:24
    bike ride actually looks a lot worse for
  • 00:16:26
    the one on my right wrist. Still not as
  • 00:16:28
    bad as we've seen for some devices. We
  • 00:16:31
    see quite some deviation especially here
  • 00:16:33
    in the first half of the ride. We can
  • 00:16:35
    see that the red line is sometimes too
  • 00:16:37
    high like right here for instance or
  • 00:16:39
    right here and sometimes too low like
  • 00:16:41
    right here. But the one I wore on my
  • 00:16:44
    left wrist actually did a bit better for
  • 00:16:46
    this ride. This is much closer to
  • 00:16:48
    reality. So this again would be good
  • 00:16:50
    enough for me. Let's now take a look at
  • 00:16:52
    those correlations. And let's start with
  • 00:16:54
    the Mi Band 10 on my right wrist because
  • 00:16:56
    that showed worst performance for the
  • 00:16:58
    second ride. And here we have an
  • 00:17:00
    overview of those results which still
  • 00:17:02
    have a pretty decent correlation at 0.91
  • 00:17:05
    which is much better than many devices
  • 00:17:07
    out there. And generally most points are
  • 00:17:10
    close to the blue line. Though as we can
  • 00:17:11
    see here in the lower heart rate range
  • 00:17:14
    sometimes detected too high heart rate
  • 00:17:16
    and in the higher heart rate range it's
  • 00:17:18
    sometimes detected at too low heart
  • 00:17:20
    rate. And this is related to what we saw
  • 00:17:22
    for that first half of the second
  • 00:17:24
    session where sometimes that are too
  • 00:17:25
    high and sometimes too low heart rate,
  • 00:17:27
    but overall still not that bad. And if
  • 00:17:30
    we take a look at the one I wore on my
  • 00:17:32
    left wrist, we have even better results.
  • 00:17:34
    The correlation is now really good at
  • 00:17:36
    0.96
  • 00:17:38
    and most points are on or at least close
  • 00:17:40
    to blue line. And given the price point
  • 00:17:42
    of this device, which is probably around
  • 00:17:45
    50 bucks or so, but depending on when
  • 00:17:47
    you buy it, it might be even cheaper.
  • 00:17:49
    That's just really good. Now, there are
  • 00:17:51
    some reasons why it might be performing
  • 00:17:52
    well. One of them is, for instance, the
  • 00:17:54
    temperature outside. As you saw, it's
  • 00:17:56
    pretty hot inside now, 27.4 degrees, and
  • 00:17:59
    that's because it's hot outside. And
  • 00:18:01
    potentially, that increased temperature
  • 00:18:03
    leads to better blood flow through my
  • 00:18:05
    arms and therefore an easier signal for
  • 00:18:07
    the device to pick up on. I don't know
  • 00:18:09
    if this is really the case. This is
  • 00:18:10
    something I really want to investigate
  • 00:18:12
    at some point, see if there's some kind
  • 00:18:13
    of association between the performance
  • 00:18:15
    of a device and the temperature outside.
  • 00:18:17
    But overall, this looks pretty good. But
  • 00:18:19
    let's put this in the context of many
  • 00:18:20
    other devices out there. And you can see
  • 00:18:22
    that in this overview right here. And
  • 00:18:24
    you're familiar with this kind of
  • 00:18:25
    overview by now. So, we want the device
  • 00:18:27
    to be as far to the top right as
  • 00:18:28
    possible. And here I marked the Mi Band
  • 00:18:31
    10 on my left wrist in red. And it's not
  • 00:18:34
    that far from the Mi Band 10 on my right
  • 00:18:36
    wrist, which is right here. This is a
  • 00:18:38
    bit difficult to read, but I just want
  • 00:18:40
    to emphasize how far to the top right
  • 00:18:42
    these devices are. At least based on
  • 00:18:44
    this very initial test, they're doing
  • 00:18:46
    quite well compared to the competition.
  • 00:18:48
    But let's zoom in a bit so we can read
  • 00:18:49
    those labels better. And that zoomed in
  • 00:18:51
    views is played right here. So these are
  • 00:18:52
    just the devices with a correlation of
  • 00:18:54
    0.8 or higher with the Mi Band 10 on my
  • 00:18:57
    left wrist marked in red right here. And
  • 00:19:00
    a little bit lower, we have the Mi Band
  • 00:19:02
    10 on my right wrist. But both are
  • 00:19:04
    honestly doing quite well. They're doing
  • 00:19:07
    about as well as, for instance, the
  • 00:19:09
    Fibbit Charge 6, which is another
  • 00:19:11
    smaller device, the Huawei Watchfit 3
  • 00:19:14
    and the Huawei Watchfit 4, both of which
  • 00:19:15
    are quite good heart rate trackers as
  • 00:19:17
    well. I have more data for those two
  • 00:19:19
    than I have for the Mi Band 10. So, I
  • 00:19:20
    need to do more testing in the future,
  • 00:19:22
    but this initial test looks really good.
  • 00:19:25
    It's not quite as good as the Pixel
  • 00:19:27
    Watch 3 and also different Apple
  • 00:19:28
    Watches. These are really still the best
  • 00:19:30
    heart rate trackers out there, but for
  • 00:19:32
    the price, I'm honestly quite impressed.
  • 00:19:34
    So, so far that performance in terms of
  • 00:19:36
    heart rate tracking is pretty good for
  • 00:19:38
    the price. Now, this is an initial test,
  • 00:19:40
    but it's at least a good indicator of
  • 00:19:41
    what the device might be able to do.
  • 00:19:43
    There's still one exercise we need to
  • 00:19:45
    check, though, and that's running, and I
  • 00:19:46
    always use the Runa app for that, but
  • 00:19:48
    more on that later. I'll do the running
  • 00:19:50
    test tomorrow because I really need some
  • 00:19:52
    sleep. It's already quite late and I
  • 00:19:54
    already did quite a few exercises today.
  • 00:19:56
    So, we'll do that tomorrow. and also
  • 00:19:57
    take a look at the consistency of the
  • 00:19:59
    sleep stage tracking of the two Mi Band
  • 00:20:01
    10s because they should be tracking the
  • 00:20:03
    same sleep stages. If not, then they're
  • 00:20:05
    likely not very reliable. And we'll also
  • 00:20:07
    see if we can compare to an EEG device.
  • 00:20:09
    It actually measures my brain waves. But
  • 00:20:11
    with one night, it can sometimes go
  • 00:20:12
    wrong. And I'll give the word now to
  • 00:20:14
    Tomorrow Rob. So, I slept my first night
  • 00:20:16
    with the two Mi Band 10s and the results
  • 00:20:18
    are not great at first glance at least.
  • 00:20:22
    So, I can put the two sleep stages side
  • 00:20:24
    by side. So, this is one of the two
  • 00:20:26
    devices, and if I put the others side by
  • 00:20:28
    side, there's actually a big discrepancy
  • 00:20:30
    between the two. The moment I woke up is
  • 00:20:33
    more or less similar, but otherwise,
  • 00:20:35
    they don't appear to be much the same.
  • 00:20:37
    So, I'm actually going to try also to do
  • 00:20:39
    some kind of formal analysis, but it's
  • 00:20:41
    actually hard to get this data out. I'm
  • 00:20:42
    going to have to manually put it in some
  • 00:20:44
    Excel sheet, and then hopefully the data
  • 00:20:46
    quality of my EEG is good enough as
  • 00:20:48
    well. So, I'm going to have a look at
  • 00:20:50
    that as well. But first, want to go for
  • 00:20:51
    a run because it's already heating up.
  • 00:20:53
    It's already about 27° C again in my
  • 00:20:56
    room here and actually my working room
  • 00:20:58
    it's even hotter and we don't have air
  • 00:21:00
    con here in Europe. So I'm going to go
  • 00:21:01
    for a quick run before it gets too hot.
  • 00:21:03
    Then we're going to take a look at that
  • 00:21:04
    data and after that we're going to do a
  • 00:21:06
    more formal analysis on the sleep stage
  • 00:21:08
    data to see if they're consistent
  • 00:21:09
    between the two watches and if they
  • 00:21:11
    agree well with the EG if that data
  • 00:21:13
    quality is good enough. But first we're
  • 00:21:15
    going to go on that run. Well, first a
  • 00:21:17
    coffee and then we're going to go on
  • 00:21:18
    that run. But I'm actually not a runner
  • 00:21:20
    by training. I'm a cyclist by training.
  • 00:21:22
    I've been cycling for about 15 years or
  • 00:21:24
    so for sports and my whole life just for
  • 00:21:26
    commuting. And for running, I've always
  • 00:21:28
    struggled to make my running plans. And
  • 00:21:30
    that's where Runner comes in. I
  • 00:21:31
    affiliate with Runner. Runner is the
  • 00:21:33
    best running coach app out there. So, it
  • 00:21:36
    will make your whole running plan for
  • 00:21:37
    you if you have a certain goal or if in
  • 00:21:39
    general you're just looking to get into
  • 00:21:40
    running. And during the run, for
  • 00:21:42
    instance, on your Garmin device or your
  • 00:21:44
    phone, it will show you your pace and if
  • 00:21:46
    that pace is in the range or if you
  • 00:21:47
    should speed up or slow down. I really
  • 00:21:49
    love it. It's changed the way I run in a
  • 00:21:51
    very positive sense. It gives me sort of
  • 00:21:53
    a sense that I actually know what I'm
  • 00:21:54
    doing because before I would start way
  • 00:21:56
    too fast, way too much because I'm
  • 00:21:58
    generally fit because of cycling and
  • 00:22:00
    then for running I would always get
  • 00:22:01
    injured and it's now a lot better. So
  • 00:22:03
    after my coffee we're going to go on
  • 00:22:04
    that run together with runna and analyze
  • 00:22:06
    that data.
  • 00:22:10
    I'll try a bit of multitasking and
  • 00:22:12
    stretch at the same time. So during the
  • 00:22:14
    run I was looking at the watch and also
  • 00:22:16
    at my reference and they seemed to be
  • 00:22:18
    kind of okay. There was one moment where
  • 00:22:20
    I saw some deviation but maybe that was
  • 00:22:22
    just a very momentary difference. All
  • 00:22:24
    the analysis are running now so I can
  • 00:22:26
    show you. So all the analysis for the
  • 00:22:28
    heart rate are running right there. I'm
  • 00:22:30
    doing some analysis for the sleep
  • 00:22:32
    staging right there. Once I'm done
  • 00:22:34
    stretching, we'll look at the results
  • 00:22:36
    together. And the running results
  • 00:22:38
    actually look really good. So this is
  • 00:22:40
    the one on my left wrist right here. And
  • 00:22:43
    you can see that the red line of the Mi
  • 00:22:45
    Band 10 overlaps almost perfectly
  • 00:22:47
    honestly with the blue green line of the
  • 00:22:49
    reference device. So here I have nothing
  • 00:22:51
    to complain about. Let's take a look at
  • 00:22:53
    the other one. So the one on my right
  • 00:22:55
    wrist is also basically perfect. There
  • 00:22:58
    are some small deviations, but this is
  • 00:23:00
    so minor. I would have nothing to
  • 00:23:02
    complain about here. Here in the
  • 00:23:03
    beginning, you can see that when I
  • 00:23:04
    started it, it needed a moment to catch
  • 00:23:06
    my heart rate, but this is only a few
  • 00:23:08
    seconds. And after that, it's basically
  • 00:23:11
    perfect. But let's put a number on this.
  • 00:23:13
    Let's calculate that correlation and
  • 00:23:14
    look at that correlation plot. And here
  • 00:23:16
    we have that correlation plot for the Mi
  • 00:23:17
    Band 10 on my left wrist. And the
  • 00:23:20
    correlation is more or less perfect. So
  • 00:23:22
    a rounded correlation of 1.00. And given
  • 00:23:25
    that the correlation cannot be higher
  • 00:23:27
    than 1, 1.00 is close to perfect. There
  • 00:23:30
    are still tiny deviations, but this is
  • 00:23:32
    really really as good as it gets
  • 00:23:34
    usually. But let's take a look at the
  • 00:23:36
    one on my right wrist. This is also
  • 00:23:38
    basically perfect. We see that in the
  • 00:23:40
    beginning it needed a second to catch up
  • 00:23:41
    on my heart rate. That's why for a few
  • 00:23:44
    of the measurements the heart rate was
  • 00:23:45
    too low, but after that again super
  • 00:23:47
    close to the blue line. Not many watches
  • 00:23:50
    that I've seen perform better than this
  • 00:23:51
    device. But let's take a look at that in
  • 00:23:53
    a more detailed way by looking at the
  • 00:23:55
    performance of many devices I've tested
  • 00:23:57
    in the past. And here we have that
  • 00:23:58
    overview where again we want the devices
  • 00:24:00
    to be as far to the top right as
  • 00:24:02
    possible. And we have the Mi Band 10 on
  • 00:24:04
    my left wrist marked in red right here.
  • 00:24:06
    And the one on my right wrist is
  • 00:24:08
    actually very close. It's right here. So
  • 00:24:10
    about five devices down or so, but
  • 00:24:13
    they're both among some of the best
  • 00:24:15
    devices out there. And I honestly didn't
  • 00:24:17
    expect this beforehand. I don't know if
  • 00:24:19
    the current weather conditions and if
  • 00:24:21
    the way I ran, as in not too many
  • 00:24:24
    intervals, somehow benefited this
  • 00:24:26
    performance, but overall this looks
  • 00:24:28
    really good. And we also saw for the
  • 00:24:29
    other exercises that I did quite well.
  • 00:24:31
    But let's zoom in a bit so we can read
  • 00:24:32
    those labels better. And here we have
  • 00:24:34
    that zoomed in view. in this case, just
  • 00:24:36
    the devices with a correlation of 0.9 or
  • 00:24:38
    higher. And the Mi Band 10 on my left
  • 00:24:41
    wrist is the best performer out of all
  • 00:24:43
    devices I've tested. Now, again, this is
  • 00:24:45
    for a single run under sunny conditions,
  • 00:24:48
    but still very good. And the one on my
  • 00:24:50
    right wrist basically performed the
  • 00:24:52
    same. So, we at least have one replicate
  • 00:24:54
    measurement on both of my wrists. I'm a
  • 00:24:56
    bit afraid, but I don't think this is
  • 00:24:57
    the case, that maybe the Mi Band 10 is
  • 00:25:00
    searching for Bluetooth signals of heart
  • 00:25:02
    rate and picking up on them and using
  • 00:25:04
    them without letting me know. I don't
  • 00:25:05
    think this is really the case. That
  • 00:25:07
    could be the one deviant explanation for
  • 00:25:09
    why it's doing so well. But assuming
  • 00:25:11
    that's not the case, it's doing really
  • 00:25:13
    well. And the heart rate monitor within
  • 00:25:15
    the Mi Band is able to measure my heart
  • 00:25:17
    rate quite well. And for the price, this
  • 00:25:19
    is basically the best device out there.
  • 00:25:21
    The only close contender is the Huawei
  • 00:25:23
    Band. So, for instance, the Huawei Band
  • 00:25:24
    9 did quite well on me for running about
  • 00:25:27
    the same as the Mi Band 10. So, both of
  • 00:25:29
    these are quite cheap devices. With the
  • 00:25:31
    Huawei Band, you get a bit of a bigger
  • 00:25:33
    screen, but otherwise, I'm really
  • 00:25:35
    content with the Mi Band 10. It's doing
  • 00:25:37
    about the same as, for instance, the
  • 00:25:39
    Whoop MG that I wear on my biceps or
  • 00:25:42
    different Apple watches. Now, I'll do
  • 00:25:44
    more testing. I hope to also give it to
  • 00:25:46
    Raphael for some testing or maybe
  • 00:25:47
    somebody else to independently test if
  • 00:25:50
    the Mi Band 10 is good at heart rate
  • 00:25:51
    tracking. But this is a good first
  • 00:25:53
    indication. I don't know what you think.
  • 00:25:55
    I didn't expect this. Maybe you have
  • 00:25:57
    different experience with it. Actually
  • 00:25:58
    want to look at the sensor of the Mi
  • 00:26:00
    Band 10 at some point and compare to the
  • 00:26:02
    previous Mi Bands. I didn't do much
  • 00:26:04
    running before, so I didn't do a lot of
  • 00:26:06
    testing with older Mi Bands for running.
  • 00:26:08
    But as a first indication, this is very
  • 00:26:10
    good. So those are really surprisingly
  • 00:26:12
    good results. I now want to take a look
  • 00:26:15
    at this sleep stage tracking, but that's
  • 00:26:16
    going to take me some time to both
  • 00:26:18
    process the data, actually get the data
  • 00:26:20
    into my computer. There's going to be a
  • 00:26:21
    lot of manual work. We're going to do
  • 00:26:23
    that analysis. First, I have to eat
  • 00:26:24
    something, but then I'll get back to
  • 00:26:25
    you. I analyzed all the sleep data, and
  • 00:26:27
    we're going to start by looking at how
  • 00:26:29
    the Mi Band 10 performed compared to the
  • 00:26:31
    reference device I have, which is the
  • 00:26:33
    Zmax EG headband, which can actually
  • 00:26:35
    measure my brain waves, and is
  • 00:26:37
    specifically designed for sleep
  • 00:26:38
    tracking. And here you can see how the
  • 00:26:40
    Mi Band 10, the black version compares
  • 00:26:43
    to that reference device with the sleep
  • 00:26:45
    stages is measured by the reference
  • 00:26:47
    device on top and those according to the
  • 00:26:49
    Mi Band 10 on my left wrist here on the
  • 00:26:51
    left. Now each column here sums to 100%.
  • 00:26:55
    Meaning that we can see what percentage
  • 00:26:56
    of each of the sleep stages according to
  • 00:26:58
    the reference was detected as a sleep
  • 00:27:00
    stage by the Mi Band 10. And if they
  • 00:27:02
    would perfectly agree, all values along
  • 00:27:04
    the diagonal here should be 100%. Now I
  • 00:27:06
    should mention two things here. There
  • 00:27:08
    were some small data issues with the
  • 00:27:09
    reference device, mostly leading to more
  • 00:27:12
    awake moments. So, we're going to ignore
  • 00:27:13
    the awake column right here. And second,
  • 00:27:16
    this is just for one night, so we cannot
  • 00:27:18
    draw any definitive conclusions. But I
  • 00:27:20
    do think the patterns are pretty clear.
  • 00:27:22
    Now, first of all, for this Mi Band, the
  • 00:27:24
    deep sleep agreement was pretty good at
  • 00:27:26
    about 90%. However, the light sleep
  • 00:27:29
    agreement and the RAM sleep agreement
  • 00:27:31
    both aren't very good at about 50%
  • 00:27:33
    agreement. and instead a lot of it was
  • 00:27:36
    either predicted as light sleep in the
  • 00:27:37
    case of REM or as deep sleep or REM
  • 00:27:40
    sleep in the case of light sleep. So not
  • 00:27:42
    looking that good honestly. But what is
  • 00:27:44
    even more surprising and worrying in a
  • 00:27:46
    way is how much worse the other Mi Band
  • 00:27:48
    performed. So that's this one right
  • 00:27:50
    here. This is the Mi Band on my right
  • 00:27:52
    wrist. We have pretty poor deep sleep
  • 00:27:54
    agreement where much of it was detected
  • 00:27:56
    as light sleep instead. Again, otherwise
  • 00:27:58
    being predicted as deep sleep or RAM
  • 00:27:59
    sleep and also really bad RAM sleep
  • 00:28:02
    agreement. And the fact that one agrees
  • 00:28:04
    a bit better with the reference than the
  • 00:28:06
    other also means that the two amongst
  • 00:28:08
    themselves likely don't agree very well.
  • 00:28:10
    And I want to show you that by just
  • 00:28:11
    looking at the knights themselves. And
  • 00:28:13
    here we have the actual night of data
  • 00:28:15
    with the reference device on top. So if
  • 00:28:17
    the clock time along the horizontal axis
  • 00:28:19
    and the sleep stage is on the vertical
  • 00:28:20
    axis. And on the bottom a similar plot
  • 00:28:23
    but now for the Mi Band 10. And as you
  • 00:28:25
    can see the agreement isn't that great
  • 00:28:28
    honestly. So in percentages we saw that
  • 00:28:30
    a lot of the deep sleep detected by the
  • 00:28:32
    EEG device was also detected by the Mi
  • 00:28:34
    Band 10. There are three main deep sleep
  • 00:28:37
    segments and all of these were more or
  • 00:28:39
    less detected by the Mi Band 10, but a
  • 00:28:41
    lot of extra deep sleep is detected all
  • 00:28:43
    throughout the night and this isn't very
  • 00:28:45
    realistic. And also the RAM sleep is
  • 00:28:48
    just randomly spread out throughout the
  • 00:28:50
    night. Normally you go through sleep
  • 00:28:52
    cycles, each one ending in RAM sleep. So
  • 00:28:53
    I likely had 1 2 3 4 five sleep cycles.
  • 00:28:59
    And you can clearly see this based on
  • 00:29:00
    the EG device. So there's some deep
  • 00:29:02
    sleep and light sleep followed by RAM.
  • 00:29:04
    Deep sleep and light sleep followed by
  • 00:29:06
    RAM. Deep sleep light sleep followed RAM
  • 00:29:08
    etc etc. And you just cannot see this
  • 00:29:11
    based on the Mi Band 10 data. And if you
  • 00:29:13
    look at the other Mi Band 10, so the one
  • 00:29:14
    on my right wrist, we see a very
  • 00:29:17
    different pattern. So again, it's not
  • 00:29:20
    good and very different. And so there's
  • 00:29:22
    much less deep sleep here in the
  • 00:29:23
    beginning of the night. Also, the time I
  • 00:29:25
    fell asleep is a bit shifted. Now, this
  • 00:29:27
    shift I actually allow for. So, I allow
  • 00:29:29
    in my script for it to find the best
  • 00:29:31
    match between the reference data and the
  • 00:29:34
    Mi Band 10 by shifting plus or minus 15
  • 00:29:37
    minutes because there will always be
  • 00:29:38
    some discrepancies. And with the best
  • 00:29:40
    fit that it found, it really doesn't
  • 00:29:42
    agree very well with the EG device. So,
  • 00:29:45
    it detects a lot of deep sleep near the
  • 00:29:46
    end of the night where I normally have
  • 00:29:48
    very little deep sleep. And also the RAM
  • 00:29:50
    sleep just agrees very poorly with the
  • 00:29:53
    EG device. So overall I'm not very happy
  • 00:29:56
    with this performance and I can actually
  • 00:29:58
    show you that there are many devices
  • 00:30:00
    that perform better and you can actually
  • 00:30:01
    see that in this overview right here. So
  • 00:30:04
    this is the performance in terms of
  • 00:30:05
    sleep stage tracking for all the devices
  • 00:30:07
    I've tested in the past or many of them
  • 00:30:09
    at least where the further to the right
  • 00:30:11
    and the higher devices, the better is
  • 00:30:13
    its agreement with the reference device.
  • 00:30:15
    So, we have the average agreement over
  • 00:30:16
    the individual sleep stages along the
  • 00:30:18
    horizontal axis and the worst sleep
  • 00:30:20
    stage agreement along the vertical axis.
  • 00:30:22
    Now, I actually didn't plot the Mi Band
  • 00:30:24
    10 in this overview yet, but we can do
  • 00:30:26
    that. And if we're being very generous,
  • 00:30:29
    we can take the best agreement right
  • 00:30:31
    here. So, we can calculate the average
  • 00:30:33
    agreement by summing these three numbers
  • 00:30:37
    right here and dividing it by three. So
  • 00:30:40
    we get 65%
  • 00:30:42
    and the worst sleep stage is about 52%.
  • 00:30:45
    So if we take 65 and the other number
  • 00:30:49
    was 52 we would end up somewhere around
  • 00:30:53
    here. So actually doing pretty okay but
  • 00:30:56
    as we saw ending up around here would be
  • 00:30:58
    a bit over optimistic. If we take the
  • 00:31:01
    other percentages, so those were these
  • 00:31:03
    ones right here, we would end up, let's
  • 00:31:07
    see, with an average agreement of 41%
  • 00:31:10
    and a worst of 36. So 41
  • 00:31:15
    and 36 would be somewhere around here,
  • 00:31:18
    somewhere around some Huawei devices, a
  • 00:31:21
    little bit worse than what I had for the
  • 00:31:22
    Mi Band 9 potentially. And where's the
  • 00:31:25
    Mi Band 9 Pro? Right here. So these
  • 00:31:27
    ended up right here. But we can also
  • 00:31:28
    take the average of those two results
  • 00:31:30
    and that will probably be more fair and
  • 00:31:32
    that's where I'll place it for now. Let
  • 00:31:33
    me quickly calculate that. So if we
  • 00:31:35
    assume the average is true, we would
  • 00:31:38
    over the three sleep stages which is now
  • 00:31:40
    an average on average. So this is not
  • 00:31:42
    perfect. I can recalculate for my
  • 00:31:45
    long-term testing. We would end up with
  • 00:31:47
    an average of an average of about 53
  • 00:31:49
    over the three sleep stages. And the
  • 00:31:51
    worst sleep stage right here would be
  • 00:31:53
    about 44. So we would have 53 on average
  • 00:31:56
    and about 44 as the worst sleep stage.
  • 00:31:59
    If we put these two numbers in that
  • 00:32:00
    overview right here, we would have 55.
  • 00:32:04
    So roughly 53 right here, let's say, and
  • 00:32:07
    then 45 would end up right here
  • 00:32:12
    somewhere. Maybe a little bit lower. So
  • 00:32:15
    right here, this is where the Mi Band 10
  • 00:32:16
    would end up. We can actually make it
  • 00:32:20
    readable. Okay, here we are. This is
  • 00:32:22
    sort of the temporary overview we would
  • 00:32:24
    end up with. So we have the Mi Band 10
  • 00:32:26
    right here which had a 54% average of
  • 00:32:30
    average. So not perfect in terms of
  • 00:32:32
    testing, but if I were to recalculate
  • 00:32:33
    it, we would mostly end up with the same
  • 00:32:36
    thing. And then 44% along the other
  • 00:32:39
    axis. And you can at least see that for
  • 00:32:40
    this very initial test, it's not that
  • 00:32:42
    far away from the Mi Band 9, the Mi Band
  • 00:32:45
    9 Pro, and the Mi Band 8 actually did
  • 00:32:48
    quite a lot worse in my testing. This
  • 00:32:50
    might be a fluke, but all of them are
  • 00:32:52
    amongst the lower performing devices
  • 00:32:54
    right here. So, not doing very good. The
  • 00:32:56
    best sleep stage trackers out there are
  • 00:32:58
    the Apple Watch. If you get any more or
  • 00:33:00
    less recent version, you'll get pretty
  • 00:33:02
    good sleep stage tracking. Even with
  • 00:33:04
    something like the Apple Watch SE from
  • 00:33:06
    2022, which you can probably get
  • 00:33:07
    refurbished for a pretty good deal, and
  • 00:33:09
    you'll get good heart rate tracking. The
  • 00:33:11
    eight sleep pot, which is my favorite
  • 00:33:13
    sleep improvement device on these hot
  • 00:33:15
    days. This really helps me sleep a lot
  • 00:33:17
    better because it can actively heat and
  • 00:33:19
    cool each side of the bed independently.
  • 00:33:21
    It's been a godsend for my sleep. It is
  • 00:33:24
    very expensive though, so only get it if
  • 00:33:26
    you can really afford it. If you want
  • 00:33:28
    the best discount possible, my affiliate
  • 00:33:30
    link is down here or up here and you'll
  • 00:33:32
    be supporting the channel at the same
  • 00:33:33
    time. Then we also have the Aura Ring
  • 00:33:36
    right here, which is doing pretty well.
  • 00:33:38
    And finally, there's the NUA, now called
  • 00:33:40
    Sleep 2 app, which is specifically
  • 00:33:42
    designed for sleep tracking. Also pretty
  • 00:33:44
    good. You'll need an extra device to go
  • 00:33:46
    with it though because it's just an app.
  • 00:33:48
    And then in the second tier of devices,
  • 00:33:50
    we have different Google and Fitbit
  • 00:33:51
    devices which are also pretty good and
  • 00:33:53
    you can get some Fitbits for a pretty
  • 00:33:55
    decent price. You get the same sleep
  • 00:33:57
    stage tracking no matter if you have
  • 00:33:58
    like a very old Fitbit Charge 2 or a
  • 00:34:01
    newer Fitbit Charge 6 or a Pixel Watch.
  • 00:34:04
    And finally, there's the Whoop Strap,
  • 00:34:05
    which is one of my favorite allround
  • 00:34:07
    devices. I also have an affiliate link
  • 00:34:09
    for that up here or down here and you
  • 00:34:11
    can check out my full review as well.
  • 00:34:12
    But overall for sleep stage tracking,
  • 00:34:15
    the Mi Band 10 isn't amazing. But let's
  • 00:34:17
    now get to my overall conclusions. Okay,
  • 00:34:20
    so that was a lot of data on the Mi Band
  • 00:34:22
    10. Actually, two Mi Band 10s and the
  • 00:34:24
    results are super interesting in my
  • 00:34:26
    opinion and actually better than I
  • 00:34:27
    expected. Wait, let me put you down for
  • 00:34:29
    a second. In terms of heart rate
  • 00:34:31
    tracking, the Mi Band 10 actually did
  • 00:34:33
    pretty good or actually very good, at
  • 00:34:35
    least for the price point. If you get it
  • 00:34:38
    for, let's say, 40 or 50 bucks, it's
  • 00:34:40
    really a good performer. At least based
  • 00:34:42
    on this initial testing, it's doing a
  • 00:34:44
    lot better than many quite a bit more
  • 00:34:46
    expensive devices. Now, you get a small
  • 00:34:48
    screen and it's maybe not the best sort
  • 00:34:51
    of smart watch out there, but in terms
  • 00:34:53
    of heart rate tracking, at least on me,
  • 00:34:55
    it's doing surprisingly well. I'll also
  • 00:34:57
    give it to Rafael, my colleague, for
  • 00:34:59
    testing, or maybe another person because
  • 00:35:01
    it would be very interesting to see if
  • 00:35:03
    these results are replicable on a second
  • 00:35:05
    person. And of course, this was a very
  • 00:35:06
    initial test, just doing one or two of
  • 00:35:09
    the same exercises. So, we need to do
  • 00:35:11
    many, many more to see if these results
  • 00:35:14
    are consistent over a longer time
  • 00:35:16
    period. But I'm happily surprised given
  • 00:35:18
    the price point and even if the
  • 00:35:20
    performance is actually a bit worse in
  • 00:35:22
    reality, you can't really go wrong if
  • 00:35:24
    you want sort of basic heart rate
  • 00:35:26
    tracking or maybe even very good heart
  • 00:35:28
    rate tracking. The sleep stage tracking
  • 00:35:30
    on the other hand, I'm not very excited
  • 00:35:32
    about, but this is sort of what I
  • 00:35:33
    expected. It would take a lot of money
  • 00:35:36
    and effort for Xiaomi to create a better
  • 00:35:38
    sleep stage tracking algorithm. Usually
  • 00:35:40
    across their entire lineup, a single
  • 00:35:43
    brand has the same sleep stage tracking
  • 00:35:44
    algorithm because you need to create a
  • 00:35:46
    big data set of reference data and data
  • 00:35:49
    with the device to actually train that
  • 00:35:50
    model. It's actually sort of an AI or
  • 00:35:53
    machine learning task. You've heard
  • 00:35:54
    these terms thrown around a lot and for
  • 00:35:56
    that you need a lot of reference data or
  • 00:35:58
    training data similar to how for
  • 00:36:00
    instance Google developed their large
  • 00:36:01
    language models or chat GPT was
  • 00:36:03
    developed. You need a lot of training
  • 00:36:05
    data for that to become a good model.
  • 00:36:07
    And the same is true for sleep stage
  • 00:36:09
    tracking. You need a lot of data with
  • 00:36:11
    both the reference device measurements
  • 00:36:13
    and the measurements with the actual
  • 00:36:15
    device you're trying to train for. So
  • 00:36:17
    the device itself can actually learn how
  • 00:36:19
    to predict your sleep stages. And this
  • 00:36:21
    will always be a relatively difficult
  • 00:36:23
    task with a decent amount of data
  • 00:36:25
    required. And I suspect that Xiaomi
  • 00:36:27
    hasn't collected that type of data. So
  • 00:36:29
    who's this device for? Well, based on my
  • 00:36:31
    initial testing, somebody who wants to
  • 00:36:33
    track their heart rate during their
  • 00:36:35
    exercises doesn't care too much about
  • 00:36:38
    the app because I don't really like the
  • 00:36:40
    Mi Fitness app that much compared to
  • 00:36:42
    some of the competition. I still think
  • 00:36:44
    for instance, the Aura Ring or the Whoop
  • 00:36:46
    Strap definitely provide a better app
  • 00:36:48
    experience with more interpretable data
  • 00:36:50
    also on your recovery. Both also have
  • 00:36:52
    much better sleep stage tracking. So, if
  • 00:36:55
    you're interested in your health and
  • 00:36:56
    sleep tracking, then I wouldn't get the
  • 00:36:58
    Mi Band. But if you just want heart rate
  • 00:37:00
    tracking, you want to look at your heart
  • 00:37:02
    rate during your exercise and see if
  • 00:37:04
    you're going too hard or not, then it
  • 00:37:06
    might be a valid choice. And given the
  • 00:37:08
    price, it's actually not bad. Of course,
  • 00:37:10
    think about e-ways. Do you actually need
  • 00:37:12
    this device or is it better to buy a
  • 00:37:14
    better device now instead of the cheaper
  • 00:37:16
    one now and then in half a year
  • 00:37:18
    upgrading again? But if you don't have a
  • 00:37:20
    big budget and you are going to use this
  • 00:37:22
    device for quite a while, it might
  • 00:37:24
    actually be a good option. But I'll keep
  • 00:37:25
    you updated once I do my retesting and
  • 00:37:28
    my long-term review. Now, if you do
  • 00:37:30
    decide to get the Mi Band 10, a Whoop
  • 00:37:33
    strap, an Aura Ring, an HLE pod, another
  • 00:37:36
    device, or anything at all on Amazon for
  • 00:37:38
    that matter, even something as small as
  • 00:37:40
    toilet paper. You want to support the
  • 00:37:42
    channel and save some money, there are
  • 00:37:44
    different affiliate links in description
  • 00:37:45
    below that give you the best discount
  • 00:37:47
    possible. If you're into running, I
  • 00:37:49
    really recommend the Runna app, which
  • 00:37:51
    creates running plans for you and live
  • 00:37:53
    coaching during your run. Or if you want
  • 00:37:55
    to track your glucose levels and want to
  • 00:37:57
    find out which foods cause big glucose
  • 00:37:59
    spikes for you personally, I really like
  • 00:38:02
    the Levels app. Both have an affiliate
  • 00:38:03
    link down below, giving you the best
  • 00:38:05
    deal possible. Now, given that you
  • 00:38:06
    watched this whole video on the Mi Band
  • 00:38:08
    10, I think you might like this video on
  • 00:38:11
    the Whoop Strap or this video on the
  • 00:38:13
    Huawei Watchfit 4 Pro.
タグ
  • Mi Band 10
  • heart rate tracking
  • sleep tracking
  • Xiaomi
  • fitness tracker
  • Rob
  • biological data analysis
  • exercise
  • indoor cycling
  • weightlifting