Does MFP really overestimate exercise calories burned?

2»

Replies

  • robertw486
    robertw486 Posts: 2,398 Member
    Ponkeen wrote: »
    Just wondering, as I have seen this mentioned in the threads. Has anyone reading actually put this to the test against another tracker? I would rather hear from those who have in fact put it to the test, or see a link to a reliable study, than hear people repeating what they have also read from other people in the forums here. Verifiable is preferred- how does one know for sure how reliable, or not, the exercise calorie tracking is?

    A lot depends on what kind of training that you do, for some things it's pretty reliable, for others it's very dependent on self reporting of actual exertion, so it can be a bit hit and miss.

    I've done some comparative work for running, road cycling, mountain biking and walking.

    I've used both Polar and Garmin HRMs, Runkeeper, Strava, Endomondo, Runtastic and MF assessments to get some idea. I've generally found MFP to be in the middle of the distribution curve with Endomondo being least conservative and Strava being most conservative. That's been on both Apple and Android phones. I also calculated manually.

    In principle you could assess that the discrepancy is based on different apps taking different data inputs, but

    I've also got access to a wide range of simulators, so taking a common data feed, from my Garmin, I an take the data through a pretty wide range of different approximation algorithms. Again a spread of results, although slightly narrower than the preceding.

    What I'd observe is that the spread of results was reasonably narrow for running and road cycling, about 300 cals between highest and lowest for a 10 mile run burning about 1000cals, and a similar ride consuming about 1000cals. Spread for mountain biking was pretty broad, aout 600cals, but that's understandable and my Polar was by far the highest by about 400 cals from the paper calculation. For walking it was again a broad spread, again HR being highest, but that's very much because walking isn't an appropriate exercise to approximate using HR.

    There are lots of things where there is no reliable way o estimate calorie expenditure, so it's near impossible to say whether it's overinflating actual effort or reported effort.

    fwiw I'm a control engineer by training, hence having access to the simulators that most people aren't going to be able to get hold of.

    I've found that between Endomondo and Strava, they are both very good at calories expended biking. However, Endomondo reports gross and grossly overestimates my BMR. With that taken out of the equation, both seem really close in net calorie burn. I did a lot of math to figure out if they were even ballpark!

    Endomondo reports I burn 139 calories per hour if I don't move. For a guy less than 200 pounds. Yes really.




    Lot's of great information being shared here, and it seems many of us just want to get the real calorie burn a little close to reality. I'd rather not have an app or program that does the math for me if the math is wrong. I understand nothing is perfect, but the variations can be huge.
  • kcjchang
    kcjchang Posts: 709 Member
    edited October 2015
    Ops, lower calories efficiency increase calories expended. Got a chance to run the calc and was too high on original post.
    kcjchang wrote: »
    MFP is overestimating by a factor of 29% 40% for me. I log the results of my bicycle rides based on power recording from my power meter (closest one can get other than actual laboratory testing although some may frit about left only, left/right, and branding); however, I have not paid for laboratory testing to determine my calories efficiency and arbitrarily set that 25%. I'm probably closer to 22-23% which would lower raise the overestimation closer to 16-21% 50%. I use virtual power (VP) to estimate calories from my walks (2.3 miles route, flat, and my pace is usually around 4-4.3mph) and they are roughly equal to MFP's estimates. I get more spikes and valleys with VP (probably due to GPS error) but it evens out on the long run. I have no ideal how accurate the VP is but the underlying equation used matches very well to the MFP's empirical extract at the intensity that I walk. The VP, the underlying equation, is really meant for running so I can't imagine both are based on the same empirical formula (which makes sense since the app's VP for cycling is based on fundamental physics). The app is call IpBike.

    Here is the stats from my afternoon ride:

    Time Riding: 1:53:45
    Distance (miles): 35.84
    Work (kJ): 1330
    Average Speed (mph): 18.9
    Calories @25%: 1272 (+29% different from MFP)
    Calories @23%: 1382 (+21% different from MFP)
    Calories @22%: 1445 (+16% different from MFP)
    MFP: 1655

    Checked a few other rides with different duration but similar average speed (18-20) and percentages stayed petty consistent for "Bicycling, 16-20 mph, very fast (cycling, biking, bike riding)". Haven't checked my rides where the average speeds are higher but expect something similar. Sorry, don't have anything slower.
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    robertw486 wrote: »
    Lot's of great information being shared here, and it seems many of us just want to get the real calorie burn a little close to reality. I'd rather not have an app or program that does the math for me if the math is wrong. I understand nothing is perfect, but the variations can be huge.

    I think the snag is that for most activities there are so many areas of uncertainty that it's near impossible to have a consistent method.

    Running is pretty straightforward, although elevation and surface make a difference. I'll burn far more trail running than road running for example. Cycling is affected by wind resistance that varies according to speed, posture on the bike, mechanical condition, tyre pressure etc.

    As soon as you get into self reporting on the Kolb scale you're into the judgement that lemurcat identifies. Does the level of effort that means 5 for me mean the same as 5 for you, or 5 for my partner, who happens to be disabled?
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    Jruzer wrote: »
    lemurcat12 wrote: »
    I think that's why the elliptical (which presumes intense) often seems to be one of the worst offenders.

    The MET value for moderate Elliptical is the same as the MET for walking at an average pae on a level, hard surface. Lots of people seem to assume somewhat more effectiveness.

    Interesting. I hadn't heard that before. I typically get ~ 9 kcal/min net from the elliptical and 5 kcal/min net from walking.

    On the 2011 release of the research both of those have a MET factor of 5. That is for walking at 4mph, which is a comfortable pace.
  • ScubaSteve1962
    ScubaSteve1962 Posts: 609 Member
    Ponkeen wrote: »
    Just wondering, as I have seen this mentioned in the threads. Has anyone reading actually put this to the test against another tracker? I would rather hear from those who have in fact put it to the test, or see a link to a reliable study, than hear people repeating what they have also read from other people in the forums here. Verifiable is preferred- how does one know for sure how reliable, or not, the exercise calorie tracking is?

    A lot depends on what kind of training that you do, for some things it's pretty reliable, for others it's very dependent on self reporting of actual exertion, so it can be a bit hit and miss.

    I've done some comparative work for running, road cycling, mountain biking and walking.

    I've used both Polar and Garmin HRMs, Runkeeper, Strava, Endomondo, Runtastic and MF assessments to get some idea. I've generally found MFP to be in the middle of the distribution curve with Endomondo being least conservative and Strava being most conservative. That's been on both Apple and Android phones. I also calculated manually.

    In principle you could assess that the discrepancy is based on different apps taking different data inputs, but

    I've also got access to a wide range of simulators, so taking a common data feed, from my Garmin, I an take the data through a pretty wide range of different approximation algorithms. Again a spread of results, although slightly narrower than the preceding.

    What I'd observe is that the spread of results was reasonably narrow for running and road cycling, about 300 cals between highest and lowest for a 10 mile run burning about 1000cals, and a similar ride consuming about 1000cals. Spread for mountain biking was pretty broad, aout 600cals, but that's understandable and my Polar was by far the highest by about 400 cals from the paper calculation. For walking it was again a broad spread, again HR being highest, but that's very much because walking isn't an appropriate exercise to approximate using HR.

    There are lots of things where there is no reliable way o estimate calorie expenditure, so it's near impossible to say whether it's overinflating actual effort or reported effort.

    fwiw I'm a control engineer by training, hence having access to the simulators that most people aren't going to be able to get hold of.

    an engineer, ahhh so that explains alot (LOL)

  • RobynLB83
    RobynLB83 Posts: 626 Member
    Jruzer wrote: »
    lemurcat12 wrote: »
    I think that's why the elliptical (which presumes intense) often seems to be one of the worst offenders.

    The MET value for moderate Elliptical is the same as the MET for walking at an average pae on a level, hard surface. Lots of people seem to assume somewhat more effectiveness.

    Interesting. I hadn't heard that before. I typically get ~ 9 kcal/min net from the elliptical and 5 kcal/min net from walking.

    On the 2011 release of the research both of those have a MET factor of 5. That is for walking at 4mph, which is a comfortable pace.

    I am a climber, boxer, runner, and avid hiker, and I can attest that walking 4mph is not a comfortable pace, and not even a pace that many people can hold for a mile.
  • RobynLB83
    RobynLB83 Posts: 626 Member
    edited October 2015
    RobynLB83 wrote: »
    Jruzer wrote: »
    lemurcat12 wrote: »
    I think that's why the elliptical (which presumes intense) often seems to be one of the worst offenders.

    The MET value for moderate Elliptical is the same as the MET for walking at an average pae on a level, hard surface. Lots of people seem to assume somewhat more
  • robertw486
    robertw486 Posts: 2,398 Member
    Ponkeen wrote: »
    Thanks oilphins! That is what I was looking for- an actual test of some sort. The other feedback is helpful, too. I intentionally underestimate the amount of time I spend exercising in order to control for little breather moments, waiting at intersections, etc... I just wasn't sure beyond that if I could trust the count.

    I have two apps that report to MFP, Endomondo and Strava. So when I do my bike rides, they total calories and report to MFP. No matter how MFP classifies them, the calorie count remains whatever the app recorded. Through some personal testing and comparisons with known to be accepted biking calculators, I have found that both apps are fairly accurate in measuring calorie burn while riding. However, I have found that Endomondo reports gross calories burned, including BMR, and also grossly overestimates my BMR. Strava reports only calories actually burned above what I would burn at rest, so it will now be the app I used for all biking.

    For my elliptical miles, I use the calculated calories shown on the elliptical and enter them into MFP. Since the elliptical we have uses inputs such as weight, age, heart rate, ramp, and resistance in the calculation it should be much more accurate than any assigned value per minute an app would assign. I haven't yet done as much work on calculating vs known good formulas, but the elliptical is simply not the same as walking so it's a harder comparison.



    Since several of you are actually using power meters when biking, I personally think the main issue would be software integration. Assuming the power meter is accurate, it really doesn't matter what the terrain, incline, aero profile, rolling resistance, etc are. Power to the pedals should be king regardless in terms of human output. The HRM integration might be useful to somehow compare power output to heart rate to help an app calculate some of the variables mentioned above, but since we know HRMs only work well under certain conditions it would be of little other use IMO.


    Jruzer wrote: »
    lemurcat12 wrote: »
    I think that's why the elliptical (which presumes intense) often seems to be one of the worst offenders.

    The MET value for moderate Elliptical is the same as the MET for walking at an average pae on a level, hard surface. Lots of people seem to assume somewhat more effectiveness.

    Interesting. I hadn't heard that before. I typically get ~ 9 kcal/min net from the elliptical and 5 kcal/min net from walking.

    On the 2011 release of the research both of those have a MET factor of 5. That is for walking at 4mph, which is a comfortable pace.

    But an elliptical isn't the same as walking. On our elliptical (an older Precor) even with the ramp at the lowest setting, the stride movement is somewhat "uphill" if you will, and the amount you have to lift your foot in the forward step is dictated my the mechanics of the machine. Combine that with the variable resistance, and the calorie burn can be changed quite a lot with those two adjustments that the machine allows. As an example, on our machine with the ramp steep and resistance up, it's more similar to walking up stairs. Thus you can go at a slower pace and burn equal calories, or go at the same pace and burn more calories.

    Today I got on our elliptical and did some testing of my HRM on the bike computer vs the HRM on the machine. In steady state mode without working very hard I was showing 7-7.2 METS/10-10.5 calories per minute. This was with minimal ramp and resistance at around 10 IIRC. Bumping resistance and ramp up more I got to 10 METS/15 calories per minute range. Both of those readings were at a cadence of right around 4.8 MPH as measured on the machine. But at any rate, if you allow variance for all the factors that can change, I don't see how any app without those inputs could be anything other than a complete SWAG. And more than likely they use some kind of average.

    I also did some testing on the machine HRM vs a chest strap HRM, which was my original goal. The machine and the chest strap report virtually the same heart rate, and the machine does in fact use heart rate in applying weight to calories burned per minute. During warm up my calories per minute reading was lower until my heart rate came up and stabilized, while remaining at a steady pace/ramp/resistance.
  • robertw486
    robertw486 Posts: 2,398 Member
    RobynLB83 wrote: »

    I am a climber, boxer, runner, and avid hiker, and I can attest that walking 4mph is not a comfortable pace, and not even a pace that many people can hold for a mile.

    I think this is more a subjective thing. If someone rarely walks and has a short stride it's a tougher task. For more long legged people that are thin, more fit, or just walk more and condition those muscles it's much easier.

    What's "comfortable" is the big question that is so subjective to me. That all depends on fitness levels, body shape and weight, as well as the when and where of the walking. But personally I think other than some shorter or heavier people, many if not most people can walk a 4 mph pace for a mile.

  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    edited October 2015
    RobynLB83 wrote: »
    Jruzer wrote: »
    lemurcat12 wrote: »
    I think that's why the elliptical (which presumes intense) often seems to be one of the worst offenders.

    The MET value for moderate Elliptical is the same as the MET for walking at an average pae on a level, hard surface. Lots of people seem to assume somewhat more effectiveness.

    Interesting. I hadn't heard that before. I typically get ~ 9 kcal/min net from the elliptical and 5 kcal/min net from walking.

    On the 2011 release of the research both of those have a MET factor of 5. That is for walking at 4mph, which is a comfortable pace.

    I am a climber, boxer, runner, and avid hiker, and I can attest that walking 4mph is not a comfortable pace, and not even a pace that many people can hold for a mile.

    4mph across rough terrain, carrying a daysac and fighting order is a standard planning assumption. That's comfortable, and sustainable. It's the rate I'd expect an able bodied individual to be able to walk on flatter terrain and without a load.

    It's only a 15 minute mile.

    Personally I walk a bit faster than that, so find slowing down to walk with others a bit uncomfortable.
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    edited October 2015
    an engineer, ahhh so that explains alot (LOL)

    You mean the understanding what I'm on about thing?

    Or the not suffering fools thing? That's less about being an engineer and more about being military (combat arm)
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    robertw486 wrote: »
    But an elliptical isn't the same as walking.

    Indeed, but the MET factoring allows for the fact different exercises have different effects. It addresses the difference.

    The tables aren't easy to cut and paste but that MET of 5 is for Moderate intensity, hence being very subjective. Moderate for me is going to be very different to moderate for my partner, who is disabled, for example.

    Essentially there is a vast scope for uncertainty in anything outside a fairly narrow range of activities for any calorie estimation approach. Any testing done is really only valid for that machine on the day it was done, as they all require calibration to be reliable.



  • 47Jacqueline
    47Jacqueline Posts: 6,993 Member
    Any generic assignment of calories burned is based on a general model of circumstances. Usually a 150lb guy. So, that would definitely skew the outcome.

    The best thing is to use an HRM that has been calibrated to your height, weight, age. And even then, realize the results will vary because of metabolism or fitness status.
  • ScubaSteve1962
    ScubaSteve1962 Posts: 609 Member
    an engineer, ahhh so that explains alot (LOL)

    You mean the understanding what I'm on about thing?

    Or the not suffering fools thing? That's less about being an engineer and more about being military (combat arm)

    The I know I'm right about everything and everyone else is wrong!
  • robertw486
    robertw486 Posts: 2,398 Member
    robertw486 wrote: »
    But an elliptical isn't the same as walking.

    Indeed, but the MET factoring allows for the fact different exercises have different effects. It addresses the difference.

    The tables aren't easy to cut and paste but that MET of 5 is for Moderate intensity, hence being very subjective. Moderate for me is going to be very different to moderate for my partner, who is disabled, for example.

    Essentially there is a vast scope for uncertainty in anything outside a fairly narrow range of activities for any calorie estimation approach. Any testing done is really only valid for that machine on the day it was done, as they all require calibration to be reliable.



    I guess for me I just put more faith in hard numbers, rather than selecting exercises based on broad categories of speed and intensity.

    As for the calibration thing, I've seen this stated quite often here on the forums. But on my Precor machine and many like it, there is really next to nothing in calibration once the machine is properly assembled. I have the same manual the field techs use, and the sensors and inputs are more or less designed to be foolproof.

    Now how accurate the software uses those inputs is hard to say.
  • libby328
    libby328 Posts: 287 Member
    I've seen a lot of people say it's high... I wear a chest strapped HRM and the totals on MFP is always super low in relation to my HRM. I have to adjust accordingly. For example, a 30 minute fast job for me says I burn something like 95 cals, while my HRM says around 212.
  • robertw486
    robertw486 Posts: 2,398 Member
    It's a tricky game at times.

    I'm just trying to find apps, machines, etc that log what I think is closest to correct. I would think for walking and jogging that comparisons to the accepted forumlas and calculators is a good test, but even then variances in stride length, heart rates, etc, still add up.



    On the subject, does anyone know if Strava applies weight to heart rates if you get the premium version? I've started using it for my biking, and I'm hoping it's closer with walking and such as well.
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    libby328 wrote: »
    For example, a 30 minute fast job for me says I burn something like 95 cals, while my HRM says around 212.

    Assuming you mean jog, rather than job, it does depend on what you mean by fast. Personally 30 minutes is about 6km, at about 150-160bpm on level ground. I'll burn about 100cals/mile, as I'm 165lbs, so that'll give me about 350-360cals.

    I'd agree about MFP being slightly on the low side cf my Garmin, with HRM, but only by about 10-20 cals.

    I'd check your MFP settings given the scale of that error.
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    robertw486 wrote: »
    On the subject, does anyone know if Strava applies weight to heart rates if you get the premium version? I've started using it for my biking, and I'm hoping it's closer with walking and such as well.

    It does.

    I did a 60km road session yesterday with Garmin giving 1000cals, feeding the GPX to Strava gave me 1200, Runkeeper gave me 1400, Endomondo gave me 1000 and Training Peaks gave me 1800. MFP gave me 1500.
  • robertw486
    robertw486 Posts: 2,398 Member
    robertw486 wrote: »
    On the subject, does anyone know if Strava applies weight to heart rates if you get the premium version? I've started using it for my biking, and I'm hoping it's closer with walking and such as well.

    It does.

    I did a 60km road session yesterday with Garmin giving 1000cals, feeding the GPX to Strava gave me 1200, Runkeeper gave me 1400, Endomondo gave me 1000 and Training Peaks gave me 1800. MFP gave me 1500.

    Holy crap. I'm trying to nail down an accurate method, but having so many inputs would make me insane! Which do you think are the most accurate of the above?

    I'm also interested in how your Endomondo is recording lower than the Strava. I had the exact opposite, and Endo seemed to be throwing a lot of BMR calories at any exercise.
  • patrikc333
    patrikc333 Posts: 436 Member
    I use the estimated cals for waling only, and it looks like what I use is accurate

    not sure for actual running, as I use only garmin estimate (and strava overestimates)
  • MeanderingMammal
    MeanderingMammal Posts: 7,866 Member
    edited October 2015
    robertw486 wrote: »
    robertw486 wrote: »
    On the subject, does anyone know if Strava applies weight to heart rates if you get the premium version? I've started using it for my biking, and I'm hoping it's closer with walking and such as well.

    It does.

    I did a 60km road session yesterday with Garmin giving 1000cals, feeding the GPX to Strava gave me 1200, Runkeeper gave me 1400, Endomondo gave me 1000 and Training Peaks gave me 1800. MFP gave me 1500.

    Holy crap. I'm trying to nail down an accurate method, but having so many inputs would make me insane! Which do you think are the most accurate of the above?

    I'm also interested in how your Endomondo is recording lower than the Strava. I had the exact opposite, and Endo seemed to be throwing a lot of BMR calories at any exercise.

    And all derived from exactly the same data set. A Garmin 310XT head end with HRM, Cadence and Speed sensors. It uploads to Garmin which syncs automatically to Endomondo, MFP and Strava, then I use another intermediary to send it to Runkeeper and Training Peaks. Looking into it I think Endomondo allows Garmin to drive, so that may not be Endomondos assessment.

    I think 1300-1400 is about right, I had a fair amount of climbing early in the session so the 1000 doesn't seem to reflect that. In terms of effort, about the same as a 13-14 mile training run.

    That said it's a carbon frame and well maintained so minimal additional weight and rolling resistance. My average speed was only 18mph so bottom end of where wind resistance would be a significant factor.
  • kcjchang
    kcjchang Posts: 709 Member
    For me, Endomodo gives the lowest estimate. Strava is always on the high side and MFP in the middle. If you are uploading to Endomondo, it will reflect what in the file's header instead of recalculating based on the data set.

    Strava recalculation based on what's in the data set and makes its own determination of the path you took, active time, speed, etc. Not sure what it uses for calories but I know the free version will use the HR data if it is uploaded (I have never use the app to record an activity). It will display average power over the segments but does not use it to estimate calories in the free version. I think Strava is based more on Skiba than Coggan or TISS.
  • fiddletime
    fiddletime Posts: 1,868 Member
    For a 2 mile flat run my fit bit says about 420, my chest strap polar 220, MFP 205 and the Precor treadmill about 300. I always take 1/3rd off the treadmill. That seems to correlate with my calories and weight. If I do a HIIT session I usually just eat back half the calories of the HRM as the values are even more hit or miss. The Fitbit calories are always way too high.