Race time predictor
One recent race in, a prediction at any other distance out. Three established models are shown separately, because where they disagree is the most useful thing on the page.
Predicted 10K
52:05
Median of the three models, which agree to within 14 seconds here. VDOT 38.3.
What each model says
Power law with exponent 1.06. Simple and transparent; optimistic when stepping a long way up in distance.
Fitted curve rather than a power law. Usually the most realistic of the three from 5K up to the marathon.
Assumes you are equally well trained for both distances — which is exactly what a 5K runner is not, for a marathon.
Predictions at every distance
The consensus figure at each distance, with the average pace it implies. Rows far from your entered race are the least reliable; see the note above the table where it applies.
| Distance | Predicted | Pace /km | Pace /mi |
|---|---|---|---|
| 1500 m | 6:51 | 4:34 | 7:21 |
| Mile | 7:24 | 4:36 | 7:24 |
| 3K | 14:33 | 4:51 | 7:48 |
| 2 miles | 15:40 | 4:52 | 7:50 |
| 5K | 25:00 | 5:00 | 8:03 |
| 8K | 41:09 | 5:09 | 8:17 |
| 5 miles | 41:24 | 5:09 | 8:17 |
| 10K | 52:05 | 5:12 | 8:23 |
| 12K | 1:03:07 | 5:16 | 8:28 |
| 15K | 1:19:56 | 5:20 | 8:35 |
| 10 miles | 1:26:11 | 5:21 | 8:37 |
| 20K | 1:48:40 | 5:26 | 8:45 |
| Half marathon | 1:55:00 | 5:27 | 8:46 |
| 25K | 2:17:43 | 5:31 | 8:52 |
| 30K | 2:47:01 | 5:34 | 8:58 |
| Marathon | 3:59:47 | 5:41 | 9:09 |
| 50K | 4:47:02 | 5:44 | 9:14 |
The three models
Riegel (1977)
Peter Riegel fitted T₂ = T₁ × (D₂/D₁)^1.06 across a wide set of race results. Its
appeal is that you can do it on a phone calculator and see exactly what it assumes: that the
relationship between distance and time is the same shape everywhere on the curve. That assumption
is good over a doubling and progressively worse beyond it.
Cameron
Dave Cameron's formula uses a fitted function of distance rather than a fixed exponent, so the penalty for going longer grows in a way a power law cannot capture. In practice it is usually the most realistic of the three between 5K and the marathon, which is why it is worth looking at whenever it disagrees with Riegel.
Daniels VDOT
Rather than fitting finish times to each other, Jack Daniels and Jimmy Gilbert modelled the oxygen cost of running at a given speed and the fraction of maximum a runner can hold for a given duration. Your race gives a VDOT; the VDOT gives every other race. It is the only one of the three that also tells you what to run in training, which is what the VDOT calculator does with it.
What the models cannot know
All three assume equivalent training for both distances. None of them knows your weekly mileage, your longest recent run, the course profile, the weather, or whether you have ever run further than the race you entered. For a step up in distance, those factors only ever make you slower than the prediction, never faster, so treat a predicted time as the fastest you could run if everything else were in place.
The practical use is not the number itself but the pace it implies. Take the consensus prediction to the splits calculator to see what it means kilometre by kilometre, and be honest about whether that pace felt sustainable in training.
Common questions
How accurate is a race time predictor?
Over a doubling of distance (5K to 10K, or 10K to half marathon) a well-trained runner typically finishes within about 2% of the prediction, which is under a minute on a 45-minute 10K. Beyond that the error grows quickly, and it is almost always in one direction: the real time is slower than predicted, because the models assume you are equally well trained for both distances.
Why do the three models disagree?
They were fitted to different data with different shapes. Riegel is a pure power law, which is simple but keeps its exponent constant however far you extrapolate. Cameron bends away from a power law and usually lands closest from 5K to marathon. Daniels VDOT comes from a physiological model of oxygen cost rather than a curve fit to finish times. Where they agree, the prediction is solid; where they spread out, that spread is the honest uncertainty.
Can I predict a marathon from a 5K?
You can compute it, but you should not plan a race around it. A marathon is more than eight times a 5K, and the limiting factor changes completely: a 5K is limited by oxygen delivery, a marathon by fuel, muscular durability and how much long-run volume you have done. The predictions on this page for that jump are a ceiling on what your current speed permits, not a forecast of what you will run.
What is the Riegel exponent and should I change it?
It is the power that distance is raised to: time2 = time1 x (distance2/distance1)^k, with k = 1.06 as published. A higher exponent penalises the step up in distance more. If your weekly mileage is modest relative to the longer race, 1.07 to 1.10 tends to reflect reality better. If you are a high-mileage runner stepping up, 1.05 can be closer.
Should I use my best ever race, or a recent one?
A recent one, always. The model predicts from current fitness, and a personal best from three years ago is not current fitness. If your most recent race was run in bad conditions or as part of a hard training week, it will under-read, but a tired recent result is still a better input than a fresh old one.