How I calculated the accuracy of my caloric burn measurements.

MachiavelliNZ

I use Fitbit to track my activity and MFP to track my food intake and I couldn't live without either of them.

While it is quite easy to try to improve how accurate my food tracking is, it always bugged me that I didn't really know how accurate my Fitbit's measurements were. Since "Calories In - Calories Out = Weight Loss", getting an accurate reading of Calories Out is really important.

So I sat down and figured out the difference between what my Fitbit says I burned and what was probably true in reality. I'm pleased with myself so I thought I'd share it here, too, in case it gives any of your the kick you need to do the same for your readings.

Before I talk about the steps I did to calculate my fitbit inaccuracy, keep in mind that I am assuming that (1) my scales and the system I use to weigh myself are both perfect, and, (2) my food intake tracking using MFP is perfect. In reality, the inaccuracies are due to everything in this process and not only my Fitbit but, since I'm already doing my very best with food intake tracking, it's easier to modify my exercise to change the overall result.

I used a spreadsheet and wrote down "Cals burned" (I'm assuming that this is TDEE) and "Cals eaten" weekly for every report I have received. I used the spreadsheet to calculate my caloric deficit each week. I then did a quick google search and used the popular rule of thumb that 1 pound = 3500 kcal to calculate my theoretical ideal weight loss each week based on the deficits measured by Fitbit.
I then wrote out what my actual weight loss was. Using those numbers, I figured out what my actual caloric deficit really must have been each week. Finally, that allowed me to figure out my actual total "Cals burned" figure. Comparing my fitbit results versus these numbers, I discovered that Fitbit is giving me a result for Calories burned by activity and metabolism that is, on average, 15.79% overestimated.

Obviously that 15% inaccuracy might not actually be due to my fitbit flex. It could be that I'm just recording intake wrong sometimes. But it does mean that I can use this knowledge to perfectly stay on track to my weight loss goal by only eating back 85% of my exercise calories. Knowledge really is power. Feels good.

TL;DR - I geeked out and used MFP, my fitbit weekly reports, plus maths to discover my estimated calories burned is 15% too high.

Amanda4change

Kudos for taking the time to figure it out. This is why I don't eat more than 20-25% of my exercise calories back (it's rare that I eat any of them), and why I set my deficit off of a TDEE based on sedentary activity levels. I just don't have the patience to do this type of research into it.

segacs

I don't have a fitbit, but I'm essentially trying to do the same thing with TDEE. I have mine set to "lightly active" and I'm not logging any of my exercise separately. I'm going to track through 60 days of data and see how accurate that is for me, and adjust accordingly.

Trying to estimate the per-workout calories was not only an exercise in futility, in my mind, but also led to me having vastly different calorie goals depending on how much I worked out each day. That made it really hard to meal plan. I find this method easier for my lifestyle so far.

Lourdesong

That's really cool. I've been meaning to do something like this, but I keep putting it off. I have enough data now (6 months total, 3 with fitbit) that all I really need to do is just sit down and take the time to figure it out.

shadow2soul

It feels great to have a decent grasp of how many calories you burn on average, doesn't it?

I did something similar and found that fitbit for me is around 10% underestimated. For that reason, I can actually eat the full adjustment back without worry. Now this could be because I only log 80% of the calories from my HRM or any other numerous logging inaccuracies, but whatever the reason is, it is still nice to know.

xapyk

Great explanation! Thank you!

I use almost the same method to figure out how accurate my exercise calories.
The only difference is I measure my weight every day and take moving average for 14 days. That means for today's weight I use average for last 14 days. This method allows to have weight graph smoother. Otherwise I have oscillations in my weight graphs up to 3 kg up and down. I believe that a cause of those oscillations is water balance in my body. That's why I prefer to have moving average for 14 days.

Grey line is actual weight and yellow line is averaged weight:



After checking accuracy of my cyclo Garmin I figured out that it underestimate my cyclo calories for 10-15% approx.
So I eat all my exercise calories back and have about 10-15% extra deficit.

MachiavelliNZ

Lourdesong wrote: »

I've been meaning to do something like this, but I keep putting it off. I have enough data now (6 months total, 3 with fitbit)

Do it! That's about the same total fitbit use as me, it's definitely worth a little time and it doesn't take that long. All this exercise, calories, and food can get complex so I think it's a good idea to do it just to cut out as many unknowns as possible.

shadow2soul wrote: »

It feels great to have a decent grasp of how many calories you burn on average, doesn't it?

I did something similar and found that fitbit for me is around 10% underestimated.

Wow, that's genuinely surprising. I figured fitbit must run a little high for everyone. Shows what I know. It's really fantastic to hear your experience, actually, because it tells me that everyone needs to find out how their own personal exercise works for them and not just go by what others think. If you took my calculations for yourself it would have been totally wrong.

And yes, actually knowing how many calories that I burn with relative certainty has changed my entire weightloss experience for the better. I feel totally in control now and know that I can reach my goal.

xapyk wrote: »

I measure my weight every day and take moving average for 14 days. That means for today's weight I use average for last 14 days. This method allows to have weight graph smoother. Otherwise I have oscillations in my weight graphs up to 3 kg up and down.

That's a really good idea. I measure my weight at the first thing each morning before any food or drink. I try to keep variables as consistent as possible but despite that (and despite consistent weight-loss trend) my weight oscillates up and down in a wave. I have just been accepting that as the normal way of it but maybe I should try something like your method.

shadow2soul

MachiavelliNZ wrote: »

xapyk wrote: »

I measure my weight every day and take moving average for 14 days. That means for today's weight I use average for last 14 days. This method allows to have weight graph smoother. Otherwise I have oscillations in my weight graphs up to 3 kg up and down.

That's a really good idea. I measure my weight at the first thing each morning before any food or drink. I try to keep variables as consistent as possible but despite that (and despite consistent weight-loss trend) my weight oscillates up and down in a wave. I have just been accepting that as the normal way of it but maybe I should try something like your method.

Have you linked your fitbit account to trendweight.com?
It takes your day to day weigh in's and gives you an average.

Example:

xapyk

MachiavelliNZ wrote: »

I have just been accepting that as the normal way of it but maybe I should try something like your method.

I accepting that as well. I consider that as my personal characteristic. If I eat too much simple carbs (sweeties, chocolate, brioches, cookies etc) on next day I can see 3-4kg plus. That's understandable. 1g of carbs takes 3g of water and they are stored all together in glycogen depots.
Also after some enduro exercises like cycling 150km or running 15-25 km I can see 3-5 kg minus. And that understandable as well: my body is just dehydrated.

So it doesn't make sense to take into account changes in water balance. But 14 day average gives much more realistic picture

MachiavelliNZ

shadow2soul wrote: »

Have you linked your fitbit account to trendweight.com?
It takes your day to day weigh in's and gives you an average.

I had never even heard of that. I'll check it out.

That image you showed has a surprising trend line. It appears to be hesitant to more closely track those drops in weight. If I were doing my own trend line I would try to make it sit more closely to the data. Can you vary the sensitivity?

SnuggleSmacks

Miss my BodyMedia. This was part of its reporting. It would spit out excel spreadsheet filled with this plus quite a bit of other information. Extremely helpful to know how accurate your gadgets are

MachiavelliNZ

I went and got linked up to Trendweight. It's great that you can retroactively set a start date for the measurements.

"You will reach your goal around June 6, 2015."

Bring it on!