Gross efficiency is a parameter the apps are assuming
Just to recap, the power meter measures the amount of work you did to the crank1. Your body burns energy, and wastes most of it as heat (all machines do this, it basically can't be helped due to physics). The work done field (reported in kJ) is the directly measured input to the cranks. The calories burnt field is estimated calories. Strava assumes how efficient your body is, presumably by at least browsing peer-reviewed literature. Its assumption might be different from another app, so if you see calorie discrepancies between apps, this is why. If you want to search, this parameter is called gross efficiency (GE).
Calories burned = Work (kJ) / (GE * 4.184)
A complicating factor is that every person's GE varies a bit. Moreover, your own GE varies a bit depending on power output - it tends to increase a bit with higher workload (open access article).
One study I found estimated that of its sample of experienced female cyclists, the average gross efficiency at lactate threshold was 23.2%, with a standard deviation of 3.5 percentage points at 60% of VO2max. That is, if the sample is representative, 95% of female cyclists should have gross efficiency within about 2 standard deviations of the mean - implying a range of 16.2% to 30.2%. That's quite a big range. Men may have a slightly lower gross efficiency and this sample of men had less variability. This study is small in absolute terms.3
There may be other studies on gross efficiency, which I didn't bother to search for. If you are searching on Pubmed, "gross efficiency" may be a standard keyword, which would make it easier to search. Note that not all studies may use the same standard keyword, however.
If you did VO2max testing in a lab, some labs should be able to estimate your GE at various intensities.
A worked example incorporating individual variation
On a 78.4 mile ride I did in 2022, Strava reported:
- 1,916 kJ total work (i.e. measured at the drivetrain)
- 1,963 calories burnt (i.e. estimated work done by my body)
One calorie is 4.184 J. So, you divide the work done by the GE assumption to get kJ produced by the body. You divide by 4.184 to get calories. Using arithmetic, it appears that Strava assumed a GE of 23.3% at the time.
To illustrate how calories burnt by an individual vary, we can use the standard deviation (SD) around the GE estimate. If the data are normally distributed (which the study authors said appeared to be approximately true - it's always an approximation), 68% of people are within 1 SD of the mean, and just over 95% are within 2 SDs.
Let's take that ride, and apply the GE estimate at an output of 150W for male cyclists. That's closer to my average power for the ride. The study reported average GE by sex at 150W, 180W, lactate threshold, and at 60% of VO2max. At 150W, the study estimated a GE for men of 19.9%, SD 1.8 percentage points. Interested parties can do the work on their rides with a different GE estimate if they want.
I believe the following interpretation is reasonable. To do 1,916 kJ of work at about 150W:
- The average trained male cyclist probably burned about 2,301 calories.
- About 2/3 of male cyclists should have burned 2,110 to 2,530 calories.
- Almost all male cyclists should have burned 1,949 to 2,809 calories, barring unusual physiology.
You might want to assess the study population. The cited study is not professional athletes, but it is very fit - the men had an average VO2max just over 60, and the women had an average of nearly 49. The average age was in the mid 30s. I don't know how cycling GE is in less trained cyclists, but I assume it should increase with training up to some limit (e.g. can elite athletes increase it substantially? Maybe not?). And perhaps it changes with age. Interested parties might want to consider that in searching.
Basically, if you are on a strict diet, then do remember that even with a power meter, you should refine your intake depending on other measurements, like weighing yourself regularly. I am not currently aware of a practical method to estimate your own gross efficiency.
What if you absolutely must have an accurate calorie count? You could wear a portable VO2 analyzer - I know one sports scientist who compared the one I linked favorably to the usual metabolic cart. However, this means you need to exercise with a mask strapped over your face. The manufacturer reported battery life is about 8 hours, with the mask using single AAA alkaline batteries. On the ride above, which was 3+ hours in the heat, wearing the mask doesn't sound practical. Nevertheless, you should do you.
What determines GE?
As far as I know, one known factor is the proportion of slow-twitch (type I) muscle fibers. That is, if you're anaerobically dominant, your GE may be lower than average and you deserve or at least you can ingest more french fries, cookies, ice cream, etc. Aside from that, the physiological determinants of GE aren't well understood.
In running, it's possible that running form may influence GE, although this isn't that well-understood either. As for cycling, this 2008 open-access article discussed the state of knowledge at the time.2 This article discussed the link between training pedaling technique and cycling performance - many of us were probably told to pedal in circles, but there is no biomechanical evidence that this is helpful. Empirically, interventions to improve pedaling technique aren't connected to performance improvements. If anything, there's some evidence that those interventions increase energy expenditure, i.e. they reduce GE.
Footnotes
- The ride linked in the comments doesn't have actual power meter data. It reports power estimated from Strava's algorithm that makes a lot of simplifying assumptions. The cyclist in question probably wanted her actual numbers not visible for tactical reasons.
- Edward Coyle is one of the authors. He published an article that showed what he thought were longitudinal changes in Lance Armstrong's gross efficiency as he developed as an elite athlete. That study showed Armstrong's GE increasing by about 8% (NB: not percentage points) over 7 years. The findings of that study were called into question and I believe it was later retracted. One calculation error was found from a partial dataset. Coyle wasn't willing to release the full dataset. I don't know how this affects how the sports science community evaluates Coyle's full body of work. His unwillingness to release the full dataset to scrutiny does increase my uncertainty about his reliability. However, his full body of work is larger than this one study.
- The range for women is pretty wide. The study in question only had 13 men and 13 women. It would be better if we had a bigger study. I don't know the typical studies of GE are this small. Researchers could use a technique called meta-analysis to combine multiple studies. This would increase our certainty about the mean GE. Each study is usually weighted by the inverse of its variance - that is, in general, larger studies will be weighted more than small ones. That weight depends in part on the standard deviation of the data. I'm not currently aware of a technique to produce a meta-analytic estimate of the pooled SD of multiple studies. NB: In meta-analysis, you get a standard error (SE). This reflects your uncertainty about the mean. In this study's case, with 26 total people, we aren't that certain about the average GE. If you meta-analyzed many studies, you might get a pooled estimated mean GE with a small SE. You can't take that SE and make the inference above. The SE tells you the range in which the mean is likely to be (i.e. 1.96 * SE gives you the upper and lower bounds of the 95% confidence interval).
1J = 1Ws
, whereJ
,W
, ands
are the units Joule, Watt, and second, respectively. Please correct the formula in the first paragraph. Also note that Americans have the annoying habit of calling the kilo-caloriekcal
, or big calorieCal
, simply calorie. Which, of course, is plain wrong. And which easily, and unnecessarily confuses such calculations by a factor of 1000:1cal = 4.184J
and1kcal = 1Cal = 4.184kJ
.