Free Race Time Predictor: 50K to 100 Miles¶
Nobody goes looking for a race time predictor out of curiosity. You go looking because there is a race on the calendar and four people are waiting on a number. Your crew wants to know whether to be at mile 62 at 9pm or 1am. Your drop bags depend on whether you hit the far side of the course in daylight. The cutoff chart has a column you have been staring at for a week.
So here is the free one I built: the race time predictor on NavRun. Type in a race you already ran, distance and finish time, and it returns predicted finish times from 5K out to 100 miles. No account, no email, no install. It takes about twenty seconds.
This post walks through what each part of the output actually means and where the number stops being worth anything. If you want the derivation of the formula itself, that lives in what one race actually tells you.
What the free race time predictor gives you¶
Four things, and only the last one asks for an email address.
- Predicted finish times across eight distances. 5K, 10K, half, marathon, 50K, 50 mile, 100K, 100 mile, each with a pace per mile and a confidence label.
- Training paces. Easy, marathon, threshold, interval, and repetition paces, derived from the same result through the Daniels VDOT curve.
- A read on your fatigue curve. This one appears once you enter a second race.
- A printable race-day pace band, emailed as a PDF for whichever distance you pick.
The prediction table, the paces, and the fatigue read are all instant and ungated. The pace band goes to an email address because it arrives as a file.
One race in, eight distances out¶
Say you ran a 50K in 6:30. Here is what comes back:
| Distance | Predicted time | Pace | Confidence |
|---|---|---|---|
| Marathon | 5:25:31 | 12:25 /mi | High |
| 50K | 6:30:00 | 12:33 /mi | High |
| 50 Mile | 10:51:58 | 13:02 /mi | High |
| 100K | 13:44:28 | 13:16 /mi | High |
| 100 Mile | 22:58:17 | 13:47 /mi | Medium |
Look at the pace column rather than the times. It drifts from 12:33 at 50K to 13:47 at 100 miles. That drift is the whole model: as distance goes up, pace comes down, and the rate it comes down at is a single exponent.
Up to the marathon the tool uses the exponent Riegel published in 1981, 1.06, fitted on road racing. Past the marathon it uses 1.08, because 1.06 applied to a 100 miler produces a time nobody has ever run on a course with dirt on it. The higher number is a correction, and it is a blunt one.
The confidence column is the part to read first¶
Every row carries a label, and it comes from one thing: how far the tool had to stretch from the race you entered to the race it is predicting.
- High when the target is between half and double the distance you typed in.
- Medium out to a quarter or four times that distance.
- Low past that.
From a 50K, a 100K is exactly double, so it is High. A 100 miler is a bit over three times, so it is Medium. From a half marathon, a 100 miler is nearly eight times the distance, and the label says Low rather than printing a clean number and letting you believe it.
That label is doing more work than the finish time next to it. A Medium on your goal race is the tool telling you the honest version: this is a window, and the window is wide.
Enter a second race and it uses your curve instead of the default one¶
The 1.06 and 1.08 exponents are averages taken across a lot of other people. You do not have to accept them.
Add a second race with a different distance and the tool fits the exponent to your own two results, by least squares through your points in log space. Then it predicts each distance from whichever of your races is closest to it, so it extrapolates as little as possible. The fitted value gets a name, endurance-biased, balanced, or speed-biased, along with the training that tends to move it.
Neither end of that is better. A curve that leans toward endurance and a curve that leans toward speed are two shapes, and plenty of people have finished hard hundreds with either one. The number is a description of two results you have already run.
Two honest limits on this. The fit is clamped between 1.01 and 1.15, so it cannot run off to a value no human race curve produces. And if your two results come from very different courses, a road half and a mountain 50K, the fit pins near that ceiling and what you are looking at is mostly the difference between the two courses, not the difference between your speed and your endurance.
What the number does not know¶
This is a formula. It sees two numbers, a distance and a time. Everything that actually decides an ultra is invisible to it:
- Vertical gain and how it is distributed, whether that is one long climb or forty punchy ones
- What is underfoot, and whether you can run any of it
- Heat, and what the heat does to your stomach
- The night, and whether you sleep
- Altitude
- Time standing still at aid stations, which on a hundred adds up to a real chunk of the finish time
There is one more that is easy to miss. Riegel scales the pace you actually ran, so whatever course you enter rides along into every prediction. A mountain 50K carries mountain difficulty forward. A road marathon carries road forward. If you type in a flat road result and race a course with 20,000 feet of climbing, the projection is describing a race that does not exist.
Which is the honest way to hold the whole thing. A predicted time is one result you already ran, pushed through a curve. It says nothing about whether you are ready, and it is not a verdict on you. Use it as a starting window, put the slow end of that window against the cutoff chart, and go from there.
Turning it into a race-day pace band¶
Once the table is up, pick your goal distance, add the course's elevation gain if you know it, and the tool emails you a printable pace band as a PDF.
The band is even effort throughout. Up to the marathon it gives you a checkpoint every mile. Past the marathon it switches to every five miles, because a hundred rows of splits is not something anyone reads at mile 71. Each checkpoint shows the cumulative clock, which is the number that matters when you are comparing yourself to a cutoff.
If you enter elevation, the PDF says plainly that those are flat-course splits and that you will run the climbs slower and get some of it back on the descents. Even splits are not a mountain race plan. What the band gives you on a mountain course is the cumulative time to beat, which is exactly what the cutoff sheet is asking for anyway.
For pace work that is not tied to a prediction, the pace calculator solves for pace, time, or distance directly from any two of them.
If you are going past 100 miles¶
The table stops at 100 miles, and the pace band goes with it. That is deliberate rather than an oversight.
The formula was fitted on road racing up to the marathon, and 1.08 is already a patch stretched over the ultra distances. Pushing it to 200 miles would produce a confident-looking number for a race decided by sleep deprivation, night navigation, and whether you can keep eating on day three. None of that is in the equation, and there is no exponent that fixes it.
For a 200, the more useful inputs are the ones you already have: your last long finish on similar terrain, your back-to-back weekends, and the race's own cutoff structure. Cutoff math is a better planning tool at that distance than any prediction.
Predictions from the running you have already done¶
One race is a thin input. It is a single day, with whatever fitness, weather, and stomach you had that day, and it treats your fitness as a fixed thing rather than something that has been moving for the last two months.
Connect Strava and NavRun builds the same predictions from your actual training instead, and updates them as that training changes. When you add a race with its elevation gain, the prediction adjusts for the vert using Scarf's rule, where 1,000 feet of climbing costs roughly what half a mile of flat running costs, with the downhill benefit capped at 10% because that is what real-world data shows people actually get back. It also compares the race's vert per mile against the vert per mile in your own training, which is a more useful thing to know in July than any finish time is.
The free tool runs on what you type in. The dashboard runs on what you did.
Connect Strava and see predictions from your training ->
Questions¶
Is the race time predictor free?¶
Yes. No account, no card, nothing to install. The prediction table, the training paces, and the fatigue-curve read are all instant. An email address is only asked for if you want the pace band PDF.
How accurate is it for a 100 miler?¶
Treat it as a rough ceiling. Extrapolating from a shorter race to 100 miles gets a Medium or Low confidence label for a reason, and vert, terrain, heat, and how much you stand around at aid stations move a hundred-mile finish by hours. Use the number as a window, not a target.
Does it account for elevation?¶
The predictions do not. The formula only sees distance and time, so the course you enter is the terrain it assumes throughout. You can enter elevation gain when you request the pace band, and the PDF then tells you those splits are flat-course splits. Elevation-adjusted predictions are part of the signed-in dashboard, where a race carries its own gain and loss.
Can I use it for a trail race?¶
Yes, and it works best when you feed it a trail result. Enter a race with comparable terrain to the one you are aiming at and the prediction inherits that difficulty. Enter a road PR and aim at a mountain course and the number will read fast.
How many races should I enter?¶
One is enough for a prediction. Two at different distances is better, because the tool fits the fatigue exponent to your own results instead of using the default. Pick two on similar terrain so the fit describes you rather than the gap between a road course and a mountain one.
Does it predict 200 milers?¶
No. The table ends at 100 miles. Beyond that the formula is extrapolating well past anything it was fitted on, into a race decided by sleep and nutrition, so it would be a confident number with nothing behind it.
Related: