IDEO_QS_Meetup

Discussion of self tracking for athletic performance enhancement.

Transcript

1. A quantitative approach to athletic performance Sebastian Ortiz
2. Overview • What I’ve been tracking and why – Weight (DIY scale, wii balance board, Withings scale) – Calorie/food intake (fatsecret) – Experiments (meal timing, IF, calorie/macro cycling, supplements) • Some tools I have used along the way – Research: mendeley – Visualization: Pandas,matplotlib,d3.js – APIs: fatsecret, withings, fitbit, runkeeper, zeo, bodymedia,etc. • Extras – Video tracking of skills – Devices for instantaneous feedback
3. Weight tracking (1st iteration) • DIY wireless scale – 24 bit AD7799 ADC – RN41 bluetooth serial module – At90usb162 microcontroller – Power switching logic (battery life 9-10 months)
4. DIY wireless scale
6. High-tech assembly techniques
7. Wii balance board Also available for android, linux
8. Why track weight/calories? • Consciously/subconsciously change behavioral patterns • Discover long term trends • Quantify effectiveness of actions, supplements
9. Subconscious changes • Tracking weight even without consciously trying to alter it results in measurable change. • Start tracking, weight goes down without conscious effort. • Stop tracking/ track sporadically and weight goes up.
11. Long term trends
12. Experiment (meal timing) Weight(blue) basically stays unchanged after shifting most calorie intake to late at night
13. Experiment(calorie cycling for performance) • Dramatically increase total calories on days with gymnastics workouts • Number of repetitions of skill-based exercises greatly increased • Currently trying refined version of the experiment where only carbs are cycled but calories kept as close to constant as possible
14. Tools • Research • Visualization • APIs
15. How to research experiments
16. pandas • Perfect for time series data • Aligns/re-samples unequally sampled time series for comparison • Rolling window functions for visualizing trends • Nice similarities to R, dataframe,etc.
17. matplotlib • Easy to use, mature state of development • Integrates seamlessly with pandas • Integrates with ipython notebooks for documenting your experiments
18. d3.js • Can produce pretty results • Easy to add interactivity • Quick to share with others using bl.ocks.org • Simple compared to raw SVG, but still fairly involved compared to matplotlib
19. API hacking • Withings • Fatsecret • Fitbit • Instagram • Flickr • Python code available here: – https://github.com/trtg – Writeups (in progress) at my octopress blog here: • http://trtg.github.com/ • Older projects here: http://www.keyboardmods.com
20. Realtime feedback • http://www.keyboardmods.com/2012/01/gy mnastics-tap-swing-trainer.html
21. Questions?