Wing foiling · data project
Six months of wing foiling, measured
Raw Garmin GPS tracks + real offshore wind from the Chimet and Bramblemet beacons, turned into an honest picture of whether I'm actually getting better.
Built by directing an AI coding agent (Claude Code) — the analysis was described in plain English, not hand-coded.
Longest flight
78s → 23m
Jan → Jun
Time on foil
3 → 35%
monthly average
Turns carried
67 → 90%
through the turn
Sessions analysed
21
point-by-point
How it works
Garmin logs a GPS point every couple of seconds — about 5,000 per session. The analysis derives speed from raw latitude/longitude, detects the exact moments on foil vs. in the water (speed hysteresis), counts every run and turn, and matches each session to 5-minute beacon wind so conditions can be separated from skill.
What the data shows
- Longest unbroken flight grew from 78 seconds to 23 minutes — the hardest thing to fake in foiling.
- Turn carry-rate (coming out of a gybe still flying) rose from ~67% to ~90% — the cleanest skill signal, because it barely depends on the wind.
- Time on foil scatters with conditions; the monthly trend is the real story.
Privacy by design. This project ships code, method and aggregate results only. Raw GPS tracks, heart-rate data and session notes are personal and stay on my machine — they are never published. You point the scripts at your own local Garmin exports.