Why trust the number.
Your predicted finish is calculated, not guessed. AI writes the words around it, never the number. Here is the working for one example runner aiming at a marathon.
One example, worked through.
Step 1 of 6
Your efforts go in
Races, the benchmark times you give us, and once you sync, the fast reps inside labelled workouts. This runner has two benchmarks and six logged weeks of about 40 km.
- 5K
- 21:30
- Half marathon
- 1:42:00
Step 2 of 6
Your own fade
Everyone slows as the distance grows. The standard rule (Riegel, 1981) says time grows with distance to the power 1.06. Two efforts at different distances let us fit your own number instead.
- Standard
- 1.06
- This runner
- 1.08
Step 3 of 6
Each effort, carried to race distance
Every effort is projected to the marathon with that fade. Here both agree, because the fade was fitted from them. With more efforts they spread, and that spread feeds the range.
- From the 5K
- 3:35:51
- From the Half marathon
- 3:35:51
Step 4 of 6
Blended across your efforts
The projections are weighted by how recent they are, how close to race distance, and whether they were a race. No single good day decides it.
- Blended
- 3:35:51
Step 5 of 6
The marathon check
Short-race speed alone tends to predict a marathon 10 to 20 minutes too fast for most runners (Vickers and Vertosick, 2016). So time is added back when the plan's peak week is under 64 km or no run of 32 km has synced yet, scaled by how far the evidence is from marathon distance.
- Peak week
- 50 km
- Longest synced run
- None yet
- Added
- +3:01
Step 6 of 6
A range and a confidence level
The answer comes with a range and a confidence level. The range is wider on the slow side, and wider still when we are extrapolating far, like a marathon from a 5K.
- Predicted
- 3:38:52
- Range
- 3:21 to 4:10
- Confidence
- Moderate
Worked by RacePacer's own predictor when this page was built, for a made-up runner: 5K 21:30, half 1:42:00, a 50 km peak week, no runs synced yet.
When it says “not yet”.
Before you have logged 3 weeks, you get a rough range from the benchmark closest to your race, marked low confidence. With no benchmark, it shows no number at all and tells you the full prediction arrives after 3 logged weeks.
If you bring your own plan.
Your plan's paces are kept exactly as written. The prediction comes from what you actually run, never from the paces in the plan. For a marathon, the volume check above does read your plan's peak week.
Sources.
- Riegel, P. S. (1981). Athletic records and human endurance. American Scientist, 69(3), 285 to 290.
- Vickers, A. J. and Vertosick, E. A. (2016). An empirical study of race times in recreational endurance runners. BMC Sports Science, Medicine and Rehabilitation, 8, 26.