Field notes · July 2026
Every number carries its window
A percentage on its own is a rumor. A bare +22% means nothing until you know: compared to what, over how long, measured how. Most running tools show you the number and leave you to supply the story. Plover ships the story with the number.
Every finding carries its window. Not “your stride is short,” but “stride length, at the same easy pace, down 4% over six weeks against your usual.” The comparison is the evidence; hiding it would be hiding the part that matters.
When Plover can’t make a fair comparison, it says so. A card will read “needs a few more comparable runs” before it will build a baseline out of three data points. A treadmill run that never recorded running dynamics leaves the mechanics analysis marked unavailable — not estimated, not filled with a population average. Missing data stays missing. That is a feature, and we show it with some pride.
The same discipline runs deeper than the copy. Every detector reads only the data up to the day it is asked about, so this week’s feed and the historical replay run the exact same code — no peeking at the future to make a past call look clever. When Plover says a signal would have fired, it would have fired.
And some questions get a flat no. Plover makes no injury prediction. No data reliably forecasts an individual runner’s injury; the best-evidenced signals are population hazard rates, true across a cohort, never a promise about your Saturday. A recovery dip never turns red on its own, because on its own it cannot earn that. Saying less, but meaning all of it, is the whole design.