THE ENDPOINT CLOSE-SPEED STUDY
Every fall. Every return.
Range: P10–P90 · mark / line: median (P50) · range colour: endpoint open-to-close direction. Gaps: unavailable. Bracket and drawdown below: endpoint close-speed episodes, not percentiles.
Tap a route segment to select its window. S = start · F = finish. Route colours show elapsed-average speed, not median speed.
Full window facts
What these measurements mean · privacy
The speed chart shows distance-weighted P10–P90 and median of canonical smoothed speed. Drawdown is 100 × (endpoint close speed / running peak endpoint close speed − 1). Only actual endpoint closes set the peak; percentiles, wicks and mean pace do not. A tied close regains the level. Reference, trough and regained positions are sampled window ends, not exact turning locations. The bracket marks these endpoint-based episode distances, not a percentile decline. This is retrospective speed analysis, not a physiological measure or prediction.
Climb and descent are independent totals from smoothed GPS elevation: fixed 5 m grid, centered triangular smoothing over 20 m. Missing samples carry forward; leading gaps use the first known elevation. Duplicate distances use the last value. Edge samples clamp. Smoothing can soften short hills; GPS is not surveyed terrain.
GPX analysis stays in this browser. The map requests external OpenStreetMap tiles, revealing requested map areas and your IP address to the tile provider. Switching versions navigates to another app; your GPX is not transferred.