Race predictions
Runsei predicts your race times from your own recent hard efforts, not a generic formula. It looks at your running from about the last five months, including the fast stretches *inside* runs, like the reps of an interval session, and keeps only the efforts that are your best for their distance. Easy runs don't count, and neither does anything that isn't running.
How the prediction works. Everyone slows down as races get longer, but by how much differs from runner to runner. Runsei fits that curve for you: race time grows as a power of distance, and the power is your own fatigue factor, which says how quickly your pace fades as the distance grows (it stays between 1.04 and 1.10). Recent efforts count most: an effort from six weeks ago counts half as much as one from today, so the predictions follow your current fitness rather than an old best. An effort run in hills, heat or wind counts for less, but its time is never changed. With only one usable effort, Runsei uses the classic factor of 1.06. And if you have actually run a distance faster than the curve says, it shows that time instead, marked you ran this.
Honesty is built in:
- Every prediction comes with a range: about ±2.5% at distances you have covered hard, widening up to ±5% the further it has to project and the less evidence there is.
- The whole set carries one confidence (High, Moderate or Indicative), based on how many recent efforts it rests on. It never hides the numbers; lower confidence only widens the ranges.
- A distance beyond your efforts is marked as a projection. The marathon usually is one: fuelling and late-race fatigue aren't in your shorter efforts, so treat long-race numbers as a ceiling, not a promise.
- With no usable efforts, Runsei says so rather than inventing a time. A few hard efforts at different distances sharpen it a lot.
Compare models. In your race predictions you can also see the same efforts run through other well-known approaches, each with its one-line theory: VDOT (Daniels' VO₂max-based race tables, close to what Garmin and Strava estimate), Cameron (an empirical curve for how pace fades with distance), Riegel (the classic formula with a fixed 1.06 factor) and, when your heart-rate data allows it, Critical Speed (the speed you could in theory hold for a very long time, plus a small reserve of distance above it). They often disagree at the shortest and longest distances, and we are learning which fits runners best.
Critical Speed is not your threshold pace. It sits close to, and a touch faster than, the lactate-threshold pace your threshold sessions are built on. Runsei prescribes that work from your measured threshold pace.
The research this rests on
- Blythe & Király (2016). Prediction and Quantification of Individual Athletic Performance of Runners. PLOS ONE 11(6), e0157257. ReadEach runner follows their own power law of time against distance, with a personal exponent: the idea the prediction is built on.
- Riegel (1981). Athletic Records and Human Endurance. American Scientist 69(3), 285–290. The classic fixed-exponent formula (first published in Runner’s World, 1977): the default factor and one of the compared models.
- Daniels & Gilbert (1979). Oxygen Power: Performance Tables for Distance Runners. Self-published, Tempe, Arizona. The VDOT model among the compared models.
- Jones, Vanhatalo, Burnley, Morton & Poole (2010). Critical power: implications for determination of V̇O₂max and exercise tolerance. Medicine & Science in Sports & Exercise 42(10), 1876–1890. ReadWhat Critical Speed is, and why it marks the boundary just above your threshold.