For those unfamiliar, Atlon is a workout tracker that automatically tracks your workouts and provides you with a detailed breakdown of your performance. Our goal is to bring seamless performance tracking to all athletes as we currently see for running or cycling. If you want to know more, download Atlon on the App Store or check out @atlon.app on Instagram.
Enough introductions, let’s get to it. Atlon tracks your movement while you work out, and uses machine learning to “guess” the exercises you did. While we’re getting better and better, our “guesses” are not perfect and there are a bunch of stuff you can do to get better results.
1. Wear the watch correctly
The most important thing to do is to wear the watch as you’ve configured it in settings. Head to Settings then Watch Orientation, and make sure that it matches how you wear the watch when you work out. Avoid rotating the watch to the inside of your wrist unless you have to (I’m looking at you KB snatch).
Also make sure the watch is tight enough so it doesn’t wiggle around on your arm. The wiggle can confuse the algorithm and lead to worse predictions. This will also improve the heart rate readings.
Another tip is to configure the watch such that the crown is turned away from your hand. Wearing the crown close to the hand can cause accidental pushes when the wrist is flexed, e.g. in an overhead squat or handstand. And if the crown is pressed for a long time, it can contact emergency services. Not great. I actually personally prefer to wear the watch like that as the crown can more easily be adjusted using the thumb.
2. Use multiple recordings
The breakdown is currently optimized to show the result of a single workout at a time. So if you track the whole hour of a class, the warmup, strength, skill, and WOD will all be mixed together. We are aware that this is a problem, and have plans to fix it, but in the meantime, you can fix it yourself by starting and stopping Atlon for each “section”. Just start Atlon when the wall clock counts down and stop it when you’re done.
3. Don’t stress about starting and stopping
You can easily start Atlon a minute before you actually start your workout. We will automatically cut away the time before your first rep so you’ll only see your actual workout time. The same holds true in the end. You don’t have to rush to stop the recording before the barbell hits the ground after your last Clean & Jerk. We detect the last rep, and update the workout time to stop after the last rep.
4. Do consistent reps
An obvious one, but worth mentioning. Try to have good form on all your reps, e.g. also lift your arms overhead on the last burpee of the set. Good form will also help you in the long run to avoid injury and become a more consistent athlete.
Potentially a bit less obvious is to keep your arm movements consistent when doing air squats or lunges. This makes the rep count a lot more accurate. Similarly, try to be consistent in how you lift when doing single handed exercises, e.g. DB snatch, and strive to do an equal amount of reps with each arm.
5. Not all exercises are equal
While we strive to be able to recognize as many exercises as possible, we’re still not quite there yet. Our current model recognizes 24 different exercise types.
The exercise categories we detect today are:
Bike ERG, Box jumps, Burpees, Burpee box passes, Burpees over the bar, Deadlifts, Double-unders, Cleans, Snatches, HSPU, KB Swings, Lunges, Pull-ups, Push-ups, Row ERG, Running, S2OH, Thrusters, Sit-ups, Ski ERG, Squats, T2B, and Wall balls.
This means, if you do e.g. rope climbs, we might misclassify it or not detect it at all. However, if you add a description, we will update the classification manually and add the missing/wrong segments.
We will continually add more exercises to our detection algorithm as we get enough data. So keep doing what you do, and one day, we’ll detect whatever you throw at us.
6. Strict and static movements are hard to identify
Firstly, we absolutely encourage anyone to include strict movements and static exercises into their routine. However, we are not currently very good at detecting and counting these types of exercises. Our algorithms depend on movement patters, so static holds are hard to distinguish from each other, and strict movements also tend to have very little movement. Use the workout description, and we will update the prediction to correctly reflect what you did.
7. Provide descriptions
We can’t overstate how important descriptions are to us. We use them to verify that our algorithms are correct and they help us expand our exercise detector to detect new exercise types. When you provide a description, we will update your workout breakdown to reflect it which means better data for you.
We love activity tracking, and we believe that measuring your performance is key to improvement in any sport. Runners and cyclists have had great tracking options for years, but CrossFit and strength training in general are still lacking a solid solution. At Atlon, we aspire to fill this gap and take the sport to the next level.
