Mark Kiptoo ran 2:15:24 at 50 in Frankfurt on 25 October (results). Three days later Athletics Illustrated wrote that his run "ranks at the 99.5th percentile."
It's a nice line. It's also the wrong word. Kiptoo finished 13th in Frankfurt, so of what, exactly, did he beat 99.5%?
Nothing. That 99.5 is an age-graded percentage, a ratio between his time and a standard in a table. Nobody else in the race went into it. A percentile is a rank inside a crowd. The two get swapped often, and the swap changes what the number says.
So is an age grade a percentile? No. Here is the difference, with 100,807 Berlin finishers to show how far apart the two numbers sit.
What does a 99.5% age grade actually measure?
It measures how close a time came to the age standard, as a ratio. Nothing else. On the Alan Jones 2025 road table, the men's marathon standard at age 50 works out to 2:14:32. Divide that by the finish time and multiply by 100.
Athletics Illustrated headlined Kiptoo at 2:15:13. That gives 8,072 seconds over 8,113, or 99.5%, exactly their number. The official result of 2:15:24 gives 99.36% (the figure I used in October's masters marathon ranking).
Either way the arithmetic has two inputs. His time, and a number in a table. You could compute it if he'd run Frankfurt alone at 3 a.m.
That's why it can't be a percentile. A percentile needs a field. It tells you what share of a group scored below you, and it moves the moment the group changes. (The formula itself is in what your age-graded percentage means. This post is about what the result is not.)
What percentile is a 70% age grade?
At Berlin, about the 92nd. Only 7.9% of 100,807 finishers from 2021 to 2023 graded 70% or higher. It's the dataset from the Berlin distribution post, re-run for this one as a straight percentile curve.
| Age grade | Share of Berlin finishers who reached it | Percentile |
|---|---|---|
| 40% | 92.8% | 7th |
| 50% | 64.6% | 35th |
| 55% | 45.3% | 55th |
| 60% | 28.1% | 72nd |
| 70% | 7.9% | 92nd |
| 80% | 1.2% | 99th |
| 90% | 0.1% | 99.9th |
Read the 55% row first. There the two numbers agree within a third of a point, which is a coincidence of this one field. Below it the percentile falls away faster than the grade: 40% is only the 7th. Above it the percentile races ahead. A 70% is already the 92nd.
So a 60% marathon, "local class" on the usual labels, already beats seven in ten Berlin finishers. A 70%, regional class, beats more than nine in ten.
My own 78.73% from Berlin 2023 lands at about the 98th percentile of these three fields. (That's graded at my exact age, against a field graded at the mid-age of each class.) I didn't make much of it at the time. Within my own band, men 45–49, 11.6% reached 70%, so there a 70% was nearer the 88th percentile than the 92nd. Same grade, different crowd, different rank.
Same grade, different rank by sex and age
A percentile depends on who is in the field, and an age grade doesn't. Same Berlin data, same tables, same three races, split by sex:
- Men: 70% was the 90.3rd percentile. About one man in ten got there.
- Women: 70% was the 95.7th percentile, roughly one woman in twenty-three.
By age it gets more interesting. Among Berlin men the share reaching 70% hardly moves, between 8% and 12% in every band from 20–29 to 70–74. Among women it sits around 3% to 5% in every band from 20–29 to 50–54. Then it climbs, to 9.6% at 60–64 and 11.7% at 65–69. A 70% for a woman in her forties is a rarer result in that field than a 70% for a woman of 67.
The 2026 NYC Marathon points the same way. On NYRR's own age-grade tables, Marathon Handbook found that 11% of runners in their 60s reached 70%. Fewer than 3% of runners under 30 did.
But that isn't proof the tables go soft with age. The Berlin men argue against a general drift, flat at roughly one in ten across five decades. For the women I can't rule it out from this data. Part of it, at least, is who is still racing at 65. The women who start a Major at that age are a heavily selected group, the same effect I saw in Berlin's women's medians.
So if you want to know how you placed, use place. If you want to know how good the run was for your age, use the age grade.
Does 100% mean the best ever run at your age?
No, though it's close enough that the shorthand sticks. Marathon Handbook's NYC piece describes age grading as comparing a finish time with "the best ever run by someone of the same age and sex." Still, the same article gets it right a few lines down. It says 100 "matches the age-group world standard."
The standard is the precise part. According to the Alan Jones README, the 2025 standards were built from single-age bests supplied by Tom Bernhard. That method, in use since 2002, fits curves to the data. A curve through the bests is not the bests, so some records sit above it and some below.
- Mark Kiptoo, 2:15:24 at 50. The fastest marathon on record by a 50-year-old, pending ratification. On Alan Jones 2025 that's 99.36%, 52 seconds outside the curve.
- Mariko Yugeta, 2:59:02 at 67, the W65 world record. On the same table she scores 100.92% (the highest masters marathon I've graded). More in the marathons over 100%.
- Clare Elms, 4:56.77 for 1500m at 62, the W60 world record. On WMA 2023, the track table, it grades 98.9%. I unpacked that one in why a world record can age-grade below 100%.
So 100% is a model's best estimate of what's possible at your age. It isn't a record book. But it isn't a ranking either, and a best-ever run can still miss it.
Grade or place: which to quote
Quote both, and label them. Plenty of results pages show a percentage next to a place, and the two answer different questions.
- Age-graded percentage tells you how good the run was against the standard for your age and sex. It travels. You can set it against your own race from five years ago, or against a friend twenty years older, and the comparison holds.
- Percentile or place tells you where you finished in that crowd, on that day. It doesn't travel. Change the crowd and the rank changes with it, even when the run is identical.
The test I use is simple. If the number would change because different people showed up, it's a rank. If it wouldn't, it's a grade. Kiptoo's 99.36% would be 99.36% on an empty street. His 13th place would not.
Either way, a 99.5 that reads as a percentile is still a remarkable run. It just deserves the right word. See where your last race lands on the calculator, then check it against the field you actually ran in.
Sources
- Athletics Illustrated, Christopher Kelsall, "Mark Kiptoo's 2:15:13 marathon at age 50; equal to Kelvin Kiptum's former world record of 2:00:35," 28 Oct 2026 — retrieved 2026-11-11, https://athleticsillustrated.com/mark-kiptoos-21513-marathon-at-age-50-equal-to-kelvin-kiptums-former-world-record-of-20035/
- world-track.org, "2026 Mainova Frankfurt Marathon results" — Mark Kosgei Kiptoo 2:15:24, 13th. Retrieved 2026-10-28, https://world-track.org/2026/10/2026-mainova-frankfurt-marathon-results-yimer-dinki-win/
- World Athletics, athlete profile, Mark Kosgei Kiptoo (born 21 June 1976) — retrieved 2026-10-28, https://worldathletics.org/athletes/kenya/mark-kosgei-kiptoo-14208497
- Marathon Handbook, Mark Kiptoo M50 marathon world record, 26 Oct 2026 — retrieved 2026-11-11, https://marathonhandbook.com/mark-kiptoo-over-50-marathon-world-record/
- Marathon Handbook, Michael Doyle, "The Best Performances at the 2026 New York City Marathon, Adjusted for Age," updated 9 Nov 2026 — retrieved 2026-11-11, https://marathonhandbook.com/nyc-marathon-2026-age-graded/
- Alan Jones, Age-Grade Tables (road, 2025 edition, approved 10 Jan 2025 by the USATF Masters Long Distance Running Council) — retrieved 2026-11-11, https://github.com/AlanLyttonJones/Age-Grade-Tables
- moralescastillo, Berlin Marathon results dataset (1974–2023), 2021–2023 subset — retrieved 2026-05-22, https://github.com/moralescastillo/datasets
- Percentile curve and band shares: author's analysis of 100,807 Berlin finishers graded with Alan Jones 2025 at the mid-age of each 5-year class (scripts/berlin-ag), 2026-11-11