Run a 22-minute 5K, feed it into a race-time predictor, and ask for a marathon. The number that comes back is about 3 hours and 31 minutes. It looks exact, because the arithmetic behind it is — Pete Riegel’s formula multiplies your time by a fixed exponent and hands you a result to the minute. What it is not is a promise, and the gap between those two things is where a lot of over-ambitious marathon goals get set on a laptop screen and never questioned again.
What the formula actually computes
The race time predictor and the marathon time predictor both run on the same equation: T₂ = T₁ × (D₂ ÷ D₁)^1.06. Feed it a time at one distance and a target distance, and it scales the time up by the ratio of the distances, raised to the power of 1.06. That exponent is the whole model — it exists because pace fades as distance grows, and 1.06 is Riegel’s estimate of how much. It is not a measurement of you specifically. It is a curve fitted to how distance running behaves in general, applied to your one data point as if you were the average case it was built from.
The stretch the caveat warns about
The formula is most trustworthy when the two distances sit close together — a 10K predicting a half marathon, say. Stretch it from a 5K to a marathon and you are asking one data point, run in under half an hour, to stand in for a race that will take you three to five hours. The further apart the distances, the more the answer depends on your endurance base rather than your speed — and endurance base is exactly what a 5K cannot show. A fast 5K tells you how efficiently you move for a few minutes. It says nothing about whether you can keep moving for four hours.
Put a number on the stretch itself. Ask that same 22-minute 5K to predict a 10K instead of a marathon, and the model scales the time by a factor of roughly 2.08 — a predicted 10K of about 45:52. Ask it to predict a marathon and the scaling factor jumps to roughly 9.59, over a race lasting nine times as long as the one the model was actually given. Both are the same formula and the same exponent; the difference is entirely how far the target sits from the evidence, and the 5K-to-marathon jump is close to the largest one a runner can rationally ask it to make.
Why a half marathon is the better question to ask it
Ask the same model to predict a marathon from a recent half marathon instead, and the answer earns more trust — not because the arithmetic changes, but because the input already contains more of what a marathon demands. A half marathon takes over an hour for most runners; holding pace that long already tests fuelling and pacing discipline, not just top-end speed. A 5K can hint at potential. A half marathon has already rehearsed part of the problem it is being asked to predict.
The variable no formula can see
Two runners can post identical 5K times and receive identical marathon predictions, then run twenty minutes apart on race day. The model cannot see weekly mileage, long-run history, or whether either of them has practised eating on the move. The marathon is unusually good at punishing exactly what shorter distances let you get away with — a thin endurance base, a fuelling plan improvised on the day, going out ten seconds a kilometre too fast because the prediction said you could hold it. None of that shows up in a finish time from a 22-minute 5K. All of it shows up around 32 kilometres.
Treat the number as a ceiling, not a plan
The honest way to use a race predictor is as the best case your current speed makes possible, conditional on training you have not necessarily done yet. If the number in front of you reflects ambition more than evidence, it is worth cross-checking against a method that does not depend on one good race day. The VO₂ max calculator estimates the same underlying fitness from a 12-minute time trial, or from resting and maximum heart rate, with no race performance involved at all. Three different methods landing in the same range is a stronger signal than one race extrapolated a long way past where it was actually run.
Train at the fitness you have, not the finish line you want
A predicted finish time is not a pace to train at — it is a distant outcome, and treating it as a daily target is how runners end up running what is really marathon effort on what was supposed to be a threshold session. The number that actually belongs in a training plan is the VDOT the VO₂ max calculator returns, because that same value sets the easy, marathon, threshold and interval paces behind a Daniels-style plan. Train at the paces your current, honestly measured fitness supports, and a faster finish time tends to follow from the months of work. Chase the predictor’s number directly, without the build that would justify it, and the first place its optimism shows up is around 32 kilometres, not on the results page.
A predictor is not lying when it hands you 3:31 off a 22-minute 5K — it is answering exactly the question it was built to answer, which is not quite the one most runners think they asked. Use it to set a ceiling, then build the long runs and the fuelling rehearsal that would let you actually reach it. Skoma turns the block that follows into a book once the number on the watch has stopped being a guess.