Your Race Predictor Is Lying to You (A Little, and for Good Reasons)

Put a recent 5K into a race time predictor and it will hand you a marathon time with total confidence. A 24:00 5K becomes roughly a 3:55 marathon, printed to the second, no hedging.

Here’s the thing every experienced coach knows and most calculators won’t tell you: that number is a little bit of a lie. A useful lie, an honest lie, but a lie. And understanding exactly how it lies is the difference between a prediction that guides your training and one that wrecks your race.

What the formula actually assumes

Most predictors, including ours, run on Riegel’s formula: take your time at one distance, scale it by the ratio of distances raised to the power 1.06. That exponent is the whole model. It encodes one idea: as races get longer, your pace decays at a predictable rate.

For a huge range of running, that idea holds up beautifully. 5K to 10K predictions land within seconds for most runners. 10K to half marathon is nearly as good. The formula was fit to real race data and it earns its keep daily.

But hiding inside it is an assumption nobody reads: the formula assumes you are equally trained for both distances. Not equally fit. Equally trained. And that assumption fails hardest exactly where people lean on the prediction most: the marathon.

Why the marathon breaks the math

Your 5K is a test of your engine. Your marathon is a test of your engine, your fuel tank, your durability over three-plus hours, your fueling execution, and your patience. Riegel’s exponent knows about the engine and quietly assumes the rest is handled.

For a runner logging 25 miles a week, the rest is not handled. The aerobic engine that produced the 24:00 5K is real, but the fat-burning efficiency, tendon durability, and glycogen management that a 3:55 marathon requires are built by volume, and volume is exactly what a busy runner is shortest on. This is why the folk wisdom exists that Riegel runs optimistic at the marathon: for most amateurs, doubling your half marathon and adding ten minutes tells the truth better than the elegant formula does.

Our predictor lets you raise the exponent to 1.08 or 1.10 for exactly this reason. Higher exponent, more pessimistic prediction, more honest for lower-volume runners. If your weekly mileage is under 35 and the goal race is a marathon, the 1.06 number is your ceiling on a perfect day, not your plan.

The other direction matters too

Predictions can also sandbag you. A runner who trains almost exclusively slow and long often races a 5K worse than their fitness because they have not touched top-end speed in months. Feed that soft 5K into a predictor and it lowballs the marathon they are actually built for. The formula does not know what kind of runner you have been for the last six months. It only knows one afternoon.

This is why the input race matters as much as the output. Use the most recent race you actually trained for, at a distance you took seriously, ideally within the last few months. A five-year-old PR predicts the runner you were five years ago.

The number the prediction is actually good for

Here is the reframe that makes predictors genuinely valuable: stop treating the output as a promise about race day and start treating it as an input for training.

A predicted marathon time sets your training paces today. It tells you what marathon pace to rehearse in long runs, what threshold pace means this month, what easy should feel like. The training pace calculator does exactly this, computing your paces from the same race result using Daniels’ methodology, and it will happily disagree with Riegel by a minute or two at the marathon. That disagreement is not a bug. The two models answer slightly different questions, and reading them together tells you more than either alone.

Then, as training progresses, re-race something short and feed the new result back in. The prediction moving is the point. It is a fitness gauge you update, not a prophecy you receive.

How I use predictions with athletes

The protocol is simple:

Predict from a recent, honest race. Not a workout, not a watch estimate, not the PR from your twenties.

Adjust the exponent for your volume. Standard 1.06 for well-trained runners at the target distance, 1.08 or higher when the goal race is much longer than anything your training has rehearsed.

Set training paces from it, not identity. The number tells you what to practice. It does not tell you who you are.

Reserve final judgment for the tune-up race. A half marathon four to six weeks out, raced honestly, predicts your marathon better than any formula fed with a 5K. Distance proximity beats mathematical elegance.

On race day, plan slightly behind the prediction. Pacing for a time a hair slower than predicted costs you a minute if you feel great, because you can spend the surplus late. Pacing for a time ahead of the prediction costs you ten minutes when the assumption gap comes due at mile 20.

The predictor is a map drawn by someone who has never seen your training log. Good maps are worth following. They are just not worth trusting more than the terrain.