Now! (drum roll of wooden pencils on a desktop) the uber geeky results!

Previously:
Pedometer Testing: Round II – Revenge of the Pedometers
Pedometer Testing: Data Tracking
Issues with the Withings Pulse & Elevation: Part I: Dissed by the Pulse, Part II: Withings Pulse Doesn’t Track Elevation for Crap and Part III: THIS IS NOT HOW MATH WORKS

First, the pedometers:
Omron HJ-720
Withings Pulse
FitBit Ultra
Omron HJ-323
FitBit One
Accupedo
Noom Walk

Essentially, if a pedometer gave me a measurement, I tracked it, and compared it to other pedometers that measured the same thing.

All pedometers measured step count (of course). I knew from previous testing that the FitBit Ultra over-counted steps, and that held true this time. The FitBit One, however, seemed far more accurate, or at least closer to the average.

To see how different activities affected step count, I noted the step count at various points in the day. I then averaged those measurements, and plotted how far each pedometer was from the average for that time period. (I did this for every day, but I chose Thursday to show here, because it was easy to read, while being a good representative of what I saw over the course of the testing.

Here is a look at how each pedometer differed from the average over the entire 11 day course.

Even those these measurements are against the average for each time period, you can see that the FitBit Ultra gets more and more inaccurate as the day progresses. My observation is that it over-counts when I am at rest, as I tend to doing most of my running around in the morning, and stick closer to my desk in the afternoon (and in the evening, plop my butt on the couch).

The Noom Walk step count was quite off from the average–it was always higher than another other pedometer, so I didn’t count those steps when calculating the average, although I did add it to the Thursday chart, so you can see how far off the average it was.

This chart is showing how each pedometer performed over the entire time period as compared to the average.

The first section shows whether the step count for each pedometer for 11 days was over or under the average. The next section is the minimum–the most each pedometer under-counted. The third section is the maximum–the most that each pedometer over-counted. You can see that the Ultra and the Noom had the highest over-counts.

That chart is essentially showing you both consistency and accuracy. (The Ultra consistently over-counts. The Omrons consistently under-count.

Most converted your step count into miles (the Noom Walk didn’t).

I compared the pedometers against each other (I took the average mileage (throwing out the Noom, which was an obvious outlier) and saw how each pedometer did compared to the average.

Don’t hold that 3/15 data against the pedometers. We took a bike ride that day, and pedometers don’t do well with bike rides (unless you tie the pedometer to your shoe, which I was not willing to do).

I also compared mileage over a period of hours with the mileage generated by my GPS.

In this case, the pedometers which gave a slightly higher mileage are probably closer to true, because as anyone who has every walked with me can attest, I wander all over the path/sidewalk.

Only the Omron pedometers counted aerobic steps. For the Omron, aerobic steps are when you have been walking for ten minutes without stopping.

The Omron HJ-720, Withings Pulse, FitBit Ultra, FitBit One, and Accupedo all gave measures of Active / Aerobic time. These were all slightly different, and FitBit and Withings both gave you measure of different levels of activity.

I used only Very Active and Intense Activity for my comparison data. Because there are different activity levels and measurements, the different pedometers had very different measures, although they all seemed to be consistent within themselves.

To be honest, I don’t pay a lot of attention to Active Time. I’m measuring activities I’d be doing anyway. I’m not looking to break any records or train harder.

Calorie count was very interesting.

First and foremost, FitBit gives you your total calorie output, including the calories you expend just because you’re breathing and your heart as beating. To get a comparable measurement, I arbitrarily decided FitBit was giving me a BMR (basal metabolic rate) of 1600, and subtracted that from the daily totals.

The results were… odd.

The FitBit One and Accupedo were closest to average, but that doesn’t necessarily mean they’re the most accurate. Just that they were similar. Why were the Pulse and the Ultra so high? I would guess part is because the Ultra consistently has a higher step count than any other pedometer.

If you’ve been reading along, you’ve seen that I have some serious issues with the Withings elevation count.

As in, it doesn’t seem to be reality based, or even based on any math system with which I’m familiar.

I figured the best comparison would be to take what the pedometers gave me over a set period of time (when I was walking outside) and compare that to what the GPS gave me.

Because the Withings data was so cracked out, I chose the elevation counts generated by the website, because they seemed to be the closest to reality. But as you can see, that didn’t make them very accurate.

The 3/21 data was a walk into town and back. The 3/22 data was from a 4 3/4 mile hike at Coopers Rock.

I fully expected the GPS to have a higher elevation gain than the pedometers, because it counts all elevation, even slight rises that aren’t really noticeable when you’re walking (such as walking up High Street). I was surprised at how close the FitBits were to the GPS readings for our hike at Coopers Rock. Perhaps the steep terrain made them more accurate.

So what does all this mean in the long run?

From a practical point, I’m going to use the FitBit One, so I can monitor whether I’m getting enough exercise to justify eating dessert. I’m also going to keep the Accupedo app, because it’s surprisingly accurate. And for now, I’m going to keep wearing the Omron HJ-323, but I’m not sure if I’ll keep wearing it.

I hope you found that useful, and if you have any questions, I’ll be delighted to answer them. I can also share my raw data if you’d like to see if you trust my manipulations.

But I fully admit that I did this solely because I am a tremendous dork and love doing weirdo things like this.