Strava Calorie Calculations

TheBigYin

One or two people may have noticed that the Calorie consumption figures for a given bike ride may well be different to the figures that are calculated using your Garmin or other Logging devices... I noticed this a while ago, and being the kind of numerically obsessed geek that I am, I decided to look into it further.

It is my contention that rather than simply taking the calorie calculation that your HRM/GPS logging device provides, Strava has taken upon itself to try and give what could potentially be a more accurate figure - certainly if the data logging device is a non heart-rate-monitor linked smartphone...

As I'm sure you have all noticed, Strava works out a "Virtual Power" figure for your ride, which is based on the speed you're riding at and the amount of up/down/steepness of gradients you have encountered. So, for the entire ride, Strava has a figure for the number of Watts of power you are "laying down". Therefore, it can also add this up and arrive at the number of total watts expended, which, divided by the number of seconds ridden, gives a total amount of energy expended in driving the bike and you along. (Basic Physics - 1 Joule = 1 watt per second).

So - to take a worked example - my Sunday ride from the 13th October 2013 - http://app.strava.com/activities/88753940 - we have a "Total Work" figure off 1,555 kJoules of energy needed to drive me and the bike around that particular route. Okay, so how do we get from that to Calories. Simples - google tells us that 1 joule = 0.239005736 calories. For the purposes of this calculation, lets simplify a little and just work to 4 decimal places shall we...

Therefore, the actual energy used to move me and the bike around was 1,555 *0.2390 = 371.645 kCalories.

Wait a Minute Mark, I hear you say... Strava reckons that you burned 1656 kCalories on that ride, not 372! ... What's going on ??

Again, it's a fairly simple thing. While the Pedal Cycle is a wonderfully efficient machine, the Human body is a lousy inefficient engine (some more than others - but I'll get to that later!) So, while I pushed out 372kCal of energy riding, I also wasted energy in sweating, in digesting my breakfast, in thinking about what I was going to have for dinner, worrying about if I'd get home in time to hand over a Turbo Trainer to Fran, producing Snot, My heart beating approximately 20,000 times, Talking to fellow cyclists, filling (and subsequently emptying) my bladder and 1001 other little things that the human body does while riding a bike that AREN'T directly related to shoving the pedals round. This is where things go from the directly measurable (because, in fairness, MY power figures WERE from a real Power Meter, rather than Strava's Guesstimates) to the "taking an average figure and hoping it's right"

So, let's call the difference between what you shove, and what you burn to shove it the "Metabolic Efficiency Factor" (or MEF for short, because I can't be doing with re-typing that again and again. Again, a quick Google seems to show that studies tend to put this MEF at somewhere between 20 and 25%... So - I did a little digging...

All this data is based on MY ride information over the last week or so...



I've basically used the Calories and Work figure to come out with the MEF that Strava is using... Allowing for Strava only displaying whole digit numbers, it would appear that the MEF for my rides seems to be coming out pretty consistently at 21.43%. Interestingly, the data also shows that its using the same figure for rides on my Roadbike (which has a power meter on), on the Indoor Trainer (which was taking power figures from the Indoor trainer's calculations of resistance) and from the MTB, where the power figures were wholely down to Strava's own power guessing routine...

Just for the sake of thoroughness, I also decided to have a look at a cross section of the members of the GS MyFitnessPal "club" and see if if this MEF changed (say, depending on Age, Gender, Height, Weight, Calculated Body Mass or any other wierd and wonderful way.)



NOPE. Same figure more or less... and that list covers a fair cross-section of ages, genders and body shapes I reckon... It appears that they've decided on a "one size fits all" approach...

Just out of interest, here's a good bit of data...

Calories 6621, Work 5938, 21.43%... want to know who that was... Laurens ten Dam - the Pro from the Belkin Team..

( http://app.strava.com/activities/86637795 for confirmation )

Now, call me a defeatist, but I seriously doubt that MY metabolic efficiency is likely to be exactly on a par with one of the better Pro's :laugh:



But, you may also have noticed another column in the first little chart I posted... The "Garmin Calories"...

Well, it's self explanatory I suppose - thats the calories that My Garmin read at the end of the ride. One thing to notice is - they're all over the shop... you'd expect them to be all either higher or all lower, or all somewhere near, wouldn't you...

So did I at first. Then I thought about it a little more. You see, a few months ago, I went and submitted myself to one of those "NewLeaf" Metabolic Profiling tests. There's a really good writeup on the DC Rainmaker blog if you want to know more about the test ( http://www.dcrainmaker.com/2012/01/look-at-testing-with-new-leaf-fitness.html ) - mine was much the same, but on a Wattbike rather than a treadmill thankfully.

The end result is a sort of "profile" of my metabolic efficiency factor, broken down into 10bpm "bands" so the Garmin knows that if my HR is showing between 80-89 bpm, i'm burning 6.76561kCals /minute of fat and a total of 9.02081 kCals, whereas between 140 and 149 i'm burning through 0.76040 kCals of fat and 13.80201 kcals in total. It has data for basically all points between 30bpm and something like 240bpm, though if I hit either of those figures, i'm sure my calorie expenditure would be the last thing on my mind.

I guess that all this waffle shows is that estimating Calories from exercise is a) difficult b) an inexact science and c) if you give 5 people a chance to do the calculations, you'll get 5 different ways of performing the calculations and AT LEAST 5 different answers.

For me, I Have a sort of hierachy of how much I believe the data... From least belief to most it runs...

MFP's Guesstimates based on time of a "exercise" (often referred to as the random number generator)

Endomondo based on smartphone tracking but no HR data

GPS and HR tracked data

GPS / HR and POWER tracked data on a "Vanilla" Garmin Device

GPS / HR and POWER tracked data on a "NewLeaf Calibrated" Garmin Device

...



But there's a REALLY good rule of thumb to work by... if you've got 2 or more different figures - LOG THE SMALLEST and err on the side of caution.

TheBigYin

(and just to throw a spanner in the works - the Garmin figure on todays ride was probably complete and utter garbage, because the HR belt battery died on my 2 hours in, so I rode the last 45 minutes whilst clinically dead, and therefore not accruing many Calories :laugh:)

cloggsy71

Whoa... What a post!

Chapeau Sir!

I always use the lowest figure (which is normally Strava's).

narak_lol

This is very thorough, TBY!

Only have 2 figures - Garmin & Strava. I use the Garmin figure, which is consistently a lot less than Strava (around 40-50%). Having said that I don't normally log the rides :laugh:

RiotMTB

TBY this is an awesome post!! Thank you. I have always posted my results using my HRM calories because I could never make sense of Strava's calculations...

TheBigYin

Another "interesting" thing is the differences you can get between HRM units... the Garmin's are a case in point... The 605/705 were notorious for over-counting calories... they got quite a bit of negative feedback regarding this, and the 500/800's went, if anything, slightly over cautious - at least "out of the box" - but they had the option of the NewLeaf tweak...

Next time I get a chance to go out for a ride I'll take my 705 and the 800 out and log the same ride, from the same HR belt, Cadence/Speed sensor and Power Meter, and post up the 2 links to compare them

verdemujer

I will throw another variable into the mix. I don't have a Garmin or a HeartRate monitor. I have a BodyMedia Link Armband. http://www.bodymedia.com/ I'm also not the math whiz or geek that Yin is so I'm not doing the math/science for myself or this post.
I have always taken the view that its a best guess estimate anyways since none of us run around with one of those sceintific heart rate monitors hooked up to us 24/7. I was never happy withe 'burns' recorded from websites like MFP though as it always seemed like they either went overboard or under and were based soley on time spent doing the exercise and not really taking into account my true body figures. So, I bought an Armband. I figured if it works for the biggest losers, it would work for me. I chose it at the time because it actually records sleep patterns which can also play into weight lost (or not). I have learned two things - I don't sleep as well as I should, especially during the school year when I tend to go to bed later than I should and yet have to be up way earlier than I would like if given my choice. And my burns are much lower than what sites like MFP record for the activity. Two hours of karate off this site is 1667 calories but according to my armband its more like something between 300 to 600. The variance for my figures really depends on if we are doing a lot of actiivty kicking/punching verus the more quiet kata practice. I always use my armband figures for the time frame over MFP.

All that being said, Strava is often very close to my Armbands data. The difference can be about 20 calories either way. I've noted that but I tend to stick with my Armbands data for the ride/run over Strava just to be consistent here. I am now noting the average of percentages for everyone from Strava irregardless of body type, age, gender, bike, ride type, and location. Its very interesting. Big Yin, I thank you for the data. Basically though, I'll stick with the thought process that its a balancing act of exercise vs calories and so long as the averages are working out, I will continue to lose. I'm not one to hang up on the perfection of the numbers.

TheBigYin

Part of my reason for throwing these figures out there is this... If you're going out riding for 100+ miles, it's useful to know that the figures on your Garmin are somewhere near - purely as a reminder to eat something... I actually set my Garmin to throw a "Calorie Reminder" every 500 Calories, which is a pretty handy reminder that it's time to eat/drink something. Especially in poor weather, it's often easy to forget to eat/drink and before you know it, you've met the man with the hammer and you're desperately stuffing gels down your neck and grovelling along for the next 15 miles until your blood sugar gets back up...

Short rides (under a couple of hours / 35 miles or so) it's not exactly critical - but if you've a couple of back to back metrics or "proper" centuries in a weekend, proper feeding is going to be an essential part of getting through 'em - necer mind getting a decent time

verdemujer

Aha! I can set reminders with the armband. I'll have to remember to do that I as I gear up the training for my first century ride. I'm not crazy about gels. I would much rather have real food on me and in me but that takes pre-planning.

TheBigYin

thread bump

Archon2

Just read this due to your bump, TheBigYin. Really enjoyed your analysis and the discussion. So far I have been trying to play it safe and eat back 2/3 of what Strava tells me. Although I'm now just started uploading Garmin 500 data w/HRM & Cadence data to Strava. Not sure if Strava uses its own formula anyway or also uses additional info from the Garmin fit files -- such as using HRM info for improved calculations.

Anyway, based on this info you posted, I think I'll just keep eating 2/3 back approximately. Seems to be working out.

TheBigYin

That's the thing, strava without a Powermeter is just working out its own virtual power figure, based on speed, rider weight and gradient. This makes a number of assumptions... Notably rider build, position on bike, road surface condition, and the biggie... Wind speed.

If you had a 50 mile flat straight, that you rode out on with a 15mph tailwind and managed 20mph average, then turned around and rode back at 20mph, it'd say you'd worked equally as hard in both directions...and award you the same calories in each direction. We all know that the 20mph with the tailwind, you'd be barely ticking over, with an effective wind resistance of 5mph - conversely on the way back you'd be suffering like a dog with an effective wind resistance of 35mph.

As power to overcome wind resistance increases with the square of the wind speed the "virtual" power from strava would be materially flawed taken in isolation, though in that particular non real world situation it evens out.

One thing I haven't experimented with is recording the same ride with a pair of Garmins, one paired to a heart rate belt, the other not, but without power in both cases, then upload em both and see if strava does actually take into account physical stress (measured via hr ) into it's power calculation.

That'll have to wait until after LeTour though...

Archon2

One thing I haven't experimented with is recording the same ride with a pair of Garmins, one paired to a heart rate belt, the other not, but without power in both cases, then upload em both and see if strava does actually take into account physical stress (measured via hr ) into it's power calculation.

That'll have to wait until after LeTour though...

Good empirical test idea. If you do that, I, for one, would want to see the outcome. I'd do it myself, but I only have the one Garmin unit (which I just got)

TheBigYin

yeah, ideally, I'd try and scrounge another Garmin 800, but I suppose I could just use the 705 and 800 in tandem...

if nothing else, using the 2 mis-matched head units would give an idea of how Garmin's algorithms changed as well... probably need to do a series of tests thinking about it...

1. Garmin 800 c/w hr against 705 with same hr
2. Garmin 800 c/w hr against 705 without hr
3. & 4 same as above but with transposed head units

veloman21

Another point of interest is that if you have a power meter, Garmin and Strava report different calorie burns though each has the same value for total work. Garmin's approach is to use a 1:1 ratio therefore Total Work = calories burned. Strava adds a multiplier such that Total Work x 1.1147 = calories burned.

Is the Strava constant based on body efficiency or are they adding a value for RMR to the work rate to calculate the total calorie expenditure in that time period?

TheBigYin

Another point of interest is that if you have a power meter, Garmin and Strava report different calorie burns though each has the same value for total work. Garmin's approach is to use a 1:1 ratio therefore Total Work = calories burned. Strava adds a multiplier such that Total Work x 1.1147 = calories burned.

It's actually not QUITE that simple... this is based on a Garmin 800, but the 810 works in much the same way I believe

If the Garmin Has a NewLeaf Metaboloc Profile installed, then the calculation is based on heart rate, with a series of "bands" of 10beats between 30 and 240 beats... each of these bands has a "calorie per second" value and a "fat calorie per second" value. So for every second in a given "band" your calorie counter is incremented by YOUR personal burn values. It does this even if you feed the head unit with power meter data - quite rightly, because the NewLeaf data actually takes your personal figures from an accurately measured test, and while theres a degree of innacuracy from the "10 beat band" granularity, its generally less than any assumptions of "standard metabolic efficiency"

If there's no NewLeaf profile, then I'll take your word for it that the head unit falls back to a 1:1 Kilojoules : kCals reading - i've had the NewLeaf profile in the head unit since way before I had a power meter on a regular basis, and the only way to remove it is to do a factory reset on the head unit, which frankly I can't be arsed to do - at least not until after LeTour.... I kind of rely on the 800 to be there and stable and I don't want to mess with it.

If there's no power or New Leaf, then the Garmin falls back on it's own internal and slightly under-reading algorithm that works based on heart rate and some "buggeration factor" which involves rider height and weight... you can see this in action by changing either reading in the settings of the head unit and watching the calories move around.

Is the Strava constant based on body efficiency or are they adding a value for RMR to the work rate to calculate the total calorie expenditure in that time period?

As I pointed out in the first post, EVERYBODY gets the same MEF fiddle factor, be it Me,Frannybobs, Laurents ten Dam or Marianne Vos... 21.43% allowing for rounding errors. It's all there in the waffle between the two tables in my initial post... but it's simply the Kilojoules times 0.239005736 (1 joule = 0.239005736 calories says Mr Google) multiplied by their arbitary MEF factor of 21.43%

or put another way

kilojoules * (23.9005736/21.43)

veloman21

Mark,

I don't use the HRM, can't stand wearing one. The calorie data for me is driven purely by the power meter and for sure there is a constant being applied. Not sure if the constant will change when my weight changes but I'll check it out.

I'm not sure of the science but it would seem that if you are accurately measuring work done then knowing the HR does not really improve your guess? But then again, what do I know :-)

The total work figure outputted by the Edge 500 appears to already been adjusted for human efficiency so the extra value that Strava adds must for for something else? That's why I wondered if there were adding some form of RMR for the period of the ride to show total burn? All very odd, I wish each entity would declare what goes into their guess.

I guess you could easily test my situation by not wearing your HRM and see if the programming of the Edge 800 is different?

Frannybobs

Again, it's a fairly simple thing. While the Pedal Cycle is a wonderfully efficient machine, the Human body is a lousy inefficient engine (some more than others - but I'll get to that later!) So, while I pushed out 372kCal of energy riding, I also wasted energy in sweating, in digesting my breakfast, in thinking about what I was going to have for dinner, worrying about if I'd get home in time to hand over a Turbo Trainer to Fran...

Wow - was that really 9 months ago when I finally met an MFP friend in person?! So to burn more cals we need to blank out our minds? Sounds like a nice prospect!!

All really interesting stuff, particularly to a fellow data geek like me. Interesting and slightly odd that Strava use the same MEF for everyone.

I tend to go off the Garmin figure because I always have and I didn't notice until I linked Strava with MFP that it calculated its own calories per ride. It's now unlinked and I have linked to Garmin Connect - but would be interesting to know if Strava does take HR into account.

TheBigYin

Mark,

I don't use the HRM, can't stand wearing one. The calorie data for me is driven purely by the power meter and for sure there is a constant being applied. Not sure if the constant will change when my weight changes but I'll check it out.

If it's just the "chest band" irritation, there are other options - I've recently been trying the "Mio Link" (on recommendation of Dean on here IIRC) which uses an optical sensor and a wristband - being the data geek you've probably already noticed, YES, my first couple of rides were using a HR band paired to the old Garmin 705 and the Mio to the Garmin 800... After 4 rides of pretty much identical (or when not identical it was the chestband that had slipped!) data, I've adopted the Mio as my go-to option.

It's not perfect - it's rechargeable, and has only a 10h battery life, so you NEED to get into the habit of recharging after every ride - and it does result in yet another irritating "tan line" to go with the rest of the set on the arm, but by-and-large, it's MUCH better than a sweaty horrible elastic band around your chest.

veloman21

Mark, thanks for the info on the Mio link, looks interesting.

I don't really understand why heart rate would make the calorie burn values more accurate? It seems to me that it would not necessarily be a good indication of burn? For instance, if I go hard up a mountain, my heart rate spikes but after I summit and am on the way down, my heart rate is still very high for several minutes but I'm not doing any work as I'm coasting. Also even your resting heart rate can be very different day to day depending on your mood, stress, have a cold etc. but again you are not doing any significant additional work.

I can see the benefit of having heart rate info to determine what you are capable of aerobically but how does HR improve the work value derived from a power meter? I'm really interested in this as when I got a power meter, my calorie burn rates went down significantly, probably more realistically unfortunately. Would definitely appreciate your insight into this.

It seems Garmin uses a hierarchy approach to calculate calorie burn depending on available resources. It seems that they consider HR trumps Power and it does not look like they use both together on the Edge 500?
https://support.garmin.com/support/searchSupport/case.faces?caseId={471ecbc0-e4f3-11de-d785-000000000000}

TheBigYin

it's partly the "elevated HR when you stop pedalling for a few seconds" that makes a pure power approach slightly low in it's overall reading... the human body isn't like a VW blue-motion car - the engine doesnt stop running when the wheels don't need driving - if your HR is elevated, all your systems (circulatory, energy exchange and so forth) are burning more calories than they would at complete rest... It's fair to say that if you were to take the straight KiloJoule figure from the power meter, swap it to kCals and call that your burn, you'd be erring on the safe side, which, if you're doing manageable exercise levels (under 2 hours, no more than 3 days in a row, no more than 5 in a week) is a pretty saf approach - especially if you are looking to drop a few kg's.

There's fairly widely documented information that after exercise your metabolism remains in an elevated state, and you burn more calories for hours after a hard ride - so - stopping recording ANY calorie burn just because you freewheel 100m doesn't really make that much sense (other than it being an ULTRA conservative way of calculating your expenditure.

Really, the big thing about having HR data as well as the power though isn't necessarily about it being "better" or "worse" with regard to the calories burned, its about being able to chart your progress - if you are doing proper power based training, then if you notice a regular, continual drop in the average HR developed on the same power zone runs, it's an indication of a) getting fitter aerobically, b) time to do another FTP test and re-assess your threshold and zones. Equally, anomalies on the HR front can quickly highlight problems with overtraining, or the onset of other problems.

The Thing is though, that while the power figure is pretty much indisputable (+/- 2% generally), how efficient YOUR engine is in converting the fuel (food) into that power is a variable figure.

Factors that can affect the efficiency of the "engine" include

Overall Weight of rider

- affects the amount of body mass the heart has to fuel - part of which has nothing to do with moving the back wheel - i'd guess that the higher the body fat % of the rider, the less efficient the engine is (there's still blood got to circulate around that lard)

- presumably the higher body mass also results in more blood vessels to serve the extra body mass, therefore more blood needed to get through all the blood vessels, The more blood that needs to be pumped, and the more flesh needing energy to keep alive, the less %wise gets to the legs to turn the cranks. This wouldn't necessarily be restricted to higher BF% people, but again higher the BF% the more inefficient the engine.

The State of Fitness of a Rider

Pretty much self explanatory - someone lean, fit and strong, who's a seasoned athlete is going to be more efficient at converting food into energy, and in turning that energy into muscle movement - in fact thats pretty much the text-box metabolic definition of a "Fit" rider.

The "condition" of the rider

By which I mean things like...

"Is the rider rested and ready to go, or tired from prior efforts in the week" - we sort of all know about that after the MfP LeTour Challenge :laugh:

"Is the rider properly hydrated" - which can alter during the course of a ride

"Is the rider carrying an injury" - even simple things like a bit of rod rash carries an "overhead" in terms of energy that the body is diverting from riding towards healing, and is not something you can "switch off" for the 2 hours of riding your bike.

so there's lots of factors that can affect the "efficiency of the engine"...

The Training Peaks / Coggan/Allen method of equating 1kj=1cal (basically saying it's a 1:1 ratio) is understandable, especially as it's a) dead easy to work with and b) probably pretty accurate when dealing with the highly trained and motivated athletes that are likely to hire a trainer and spend a couple of thousand on a power meter...

as 1 joule = 0.2390 calories 9(near enough), assuming that the above fit rider is going to be 23.9% efficient in turning food into power is a fair assumption.

Strava play it a little more conservative and say 21.4% efficiency - for everyone - which makes it easy, but doesn't make a great deal of sense - personally I think they should have some form of sliding scale between 21% and 25% depending on riders BMI (or ideally BF% if they know it)

when I wrote the initial post, I did an awful lot of googling about and reading various studies which put efficiencies at between 20% and 25%... The lower the efficiency, the more calories you'll burn to be able to develop the power at the back wheel. It doesn't matter that you waste most of that power, you've still burned it.

I'm pretty damned sure that my "Engine" isn't going to be running as efficiently as Laurents ten Dam - but I'm happy that it's running at least twice as well as it was 3 years ago...