Self-Hosted Workout Intelligence

For years, I meticulously tracked every run. Metrics are, of course, paramount to a former athlete and current data scientist. But a puppy landing on my doorstep transformed my exercise world overnight. What started as pristine running data became (an irritainingly unlabeled) mix of actual runs but more frequent dog walks. It’s spurned my quest to make my exercise data meaningful again. From developing classification models to distinguish runs from walks and dashboards to explore how adding a four-legged training partner redefined what “consistency” actually means, this exploration of the Choco effect is an open field of study 🐕

Problem

I had scattered CSVs and subjective training notes. No single weekly view, no trend break detection, and no way to keep promises to “future me.”

Approach

Normalize inputs into a tidy model (sessions, metrics, deltas).
Compute consistency and trend-breaks with transparent rules.
Human-centered visuals that emphasize decisions rather than dashboards.

Impact

Weekly ritual, not a one-off report.
Early warning on form/volume changes.
Reduced decision debt and easier experimentation.