Which way do I rotate?

Do you ever wonder if you rotate more right or more left on average every day? Me too! My vibe coded watch app says I'm a clockwise guy, but recent scientific research suggest most people have a counterclockwise tendency. 

I spent the last few weeks vibe coding a Connect IQ app for my Garmin Instinct 3 Solar that counts how many net rotations I make in a day. It counts up when I rotate clockwise, and down when I rotate counterclockwise as long as its logging. Annoyingly, the magnetometer can only be run in an "activity" so I was posting dozens of activities on Strava during debugging and logging. 

UI during debugging 

Implementation 

This is my first fully vibe coded project, I did not read a single line of the code that Claude generated in this watch app. I was simultaneously extremely impressed by how fast and well implemented each app was, but also not impressed by the overall strategy the AI took to solve this problem. 

Claude could easily and rapidly deploy new fully functional programs to the watch. That being said, it took several iterations to get it to correctly combine the sensors into something that made sense and wasn't logging different rotations differently or drifting constantly. I also found that it did not have a good overall strategy, I had to suggest ways to approach problems and keywords to research, as well as suggest collecting reference data to simulate against. I had success in collecting large datasets of raw sensor readings while walking around the block counting how many times I rotated and having Claude simulate different sensor fusion algorithms to see what algorithms and tuning match the rotations. The v20 program was the final one that I used on the watch after implementing a sleep mode to decrease battery life and going through several iterations of implementation and UI changes. 

The watch has a 3 axis gyroscope, a 3 axis accelerometer and a 3 axis magnetometer, all three are required to get an accurate count throughout the day. The gyro is fast responding but drifts, the magnetometer is an absolute measurement of what we care about but its sorta crappy especially indoors around steel objects, and the accelerometer provides a gravity axis to reference the spins to. 

The app runs gyro and accelerometer at 25 Hz and magnetometer at 4 Hz. The gyro carries the rotation second to second, and the compass slowly pulls the heading back to truth whenever the field looks believable, and the accelerometer provides the rotation vector. There is a gate on the compass that compares field strength and dip angle against what it has been seeing and throws out anything too far off. I strongly suspect there is some lower level processing happening on the Garmin watch and the API has better filtered data than the raw sensor values; my embedded engineer friends told me that 25 Hz is nowhere close to fast enough but the rotation count was accurate in every reference test I did. 

Results after ~2 weeks of daily life

Logged data overlaid on each day

Same data sequentially

My typical day lands around +15 rotations clockwise by bedtime. There are some notable rotations, including 2 rotations clockwise every morning going down the stairs in my apartment, and 2 rotations counterclockwise in the evening when I'm back. Also very notably on September 27th I went to a farm and rode the Scrambler and Swing Ride resulting in ~75 rotations counterclockwise. 

Mental health impacts

Constantly logging my rotations impacted me more than expected. I found myself constantly wondering which way I was rotating, which way I turned when grabbing coffee, how I put my coat away, which way I got out of my chair, whether a morning walk induced a net rotation. Worse than the mental health impacts, this was also impacting the data since I was not doing my natural rotations. 

The best solution I found is putting the watch on a screen that does not show the rotations. I much more often forgot that it was logging when it didn't show my rotations and didn't have an urge to "unspin" at points during the day. 

Real science around human spinning

For some reason actual scientists have studied this, there is a paper in Nature from June that is weirdly relevant: Individual locomotor bias drives counterclockwise motion in pedestrian crowds. Across five experiments in Spain and Japan, people set loose to roam inside a 5 meter arena drift counterclockwise. It survives when everyone in the group is left handed, in open spaces with no walls to turn at, and in Japan where crowds form lanes the opposite way from Spain. They also rule out handedness, footedness and eye dominance, so where the bias actually comes from is still open. They even studied a bunch of children running around to see if a counterclockwise bias is taught and didn't find that it is. 

My watch says the opposite, 15 clockwise rotations a day on average that I didn't go to a farm and get on a machine that spins me counterclockwise 75 times. Am I built diff? probably