How do you predict your marathon time?
Convert a recent half marathon with a power-law model, sanity-check it against your weekly volume and long runs, then subtract some optimism.
The single most informative input is a recent half marathon, because it is long enough to test the same aerobic and durability qualities the marathon does. A power-law conversion — Riegel with an exponent around 1.08 for marathon predictions — turns that into a marathon estimate in one step.
A 10 km will do at a push, and a 5 km is nearly useless on its own. The shorter the input race, the more the prediction extrapolates and the more it flatters you: the marathon's limiters are fuelling, glycogen and the ability to hold form on tired legs, and a 5 km measures none of them.
The second check is training, not arithmetic. A prediction that implies a marathon pace you have never sustained for more than 10 km in training, on weekly volume that does not support the distance, is a number rather than a plan. If your long runs have not comfortably reached 2 to 2.5 hours, the prediction is describing a runner you are not yet.
The third adjustment is the course and the day. Elevation, heat and field congestion all cost time, and this site's course pages compute the grade-adjusted split table for your specific race so the prediction lands on the course you are actually running rather than an idealised flat one.
Key numbers
| Best input | recent half marathon | |
|---|---|---|
| Exponent used | ~1.08 | for marathon predictions |
| Sanity check | long runs of 2–2.5 h |
Work it out for yourself
Race time predictor — Predict your marathon and see how each physiological factor moves the estimate.
Frequently asked questions
Can I predict a marathon from a 5 km?
You can compute a number, but it will usually be optimistic by a wide margin. The 5 km tests aerobic power; the marathon tests fuelling, durability and pacing. Use a half marathon where you possibly can.
Why did I run so much slower than my predicted marathon time?
The usual suspects, in order: insufficient weekly volume to support the predicted pace, going out too fast, under-fuelling, and racing a hillier or hotter course than the prediction assumed. All four are addressable, and all four are decided before the start line.
Sources
- Riegel (1981). Athletic records and human endurance. American Scientist.
- Vickers & Vertosick (2016). An empirical study of race times in recreational endurance runners. BMC Sports Science, Medicine and Rehabilitation.
Related explainers
Glossary: Riegel Model, Joyner Model