Why NYC Marathon Age Grades Differ From the Official Page

I ran the New York City Marathon on 2 November 2025. 2:56:19, aged 48 and one week. The official results page gave me an age grade of 76.4%.

Then I logged the same race on RaceRecords, and it said 74.9%.

Same time. Same age. Same distance. A point and a half apart. And here's the uncomfortable part: neither number is wrong.

Why are the two numbers different?

Both scores use the same family of tables — the Alan Jones road tables that most major US races run on. The gap is vintage. My 2025 NYC result was graded against the 2015 edition of those tables, and RaceRecords grades against Alan Jones 2025, approved by USATF's Masters Long Distance Running Council in January 2025 (Alan Jones, Age-Grade Tables, 2025).

So it isn't two rival systems fighting. It's the same system, two editions apart.

Age grading answers one question: how close did you run to the world best for your sex and age at this distance? To answer it, the table needs a world best to measure you against. That world best is the part that changed.

What actually moved: the world record, not you

The men's marathon world record was 2:02:57 (Dennis Kimetto, Berlin 2014). Then Kelvin Kiptum ran 2:00:35 in Chicago in October 2023, ratified by World Athletics the following February (World Athletics, "Ratified: Kiptum's world marathon record", 2024). The 2025 tables adopted it as the open-class marathon standard (Alan Jones, Age-Grade Tables, 2025). That single number is the denominator every marathon age grade divides into.

And here's the thing about my official 76.4%. Work it backwards through the published editions and only one fits. The Alan Jones 2015 table — marathon standard 2:02:57, the record set by Dennis Kimetto in Berlin in 2014 — grades my run at 76.5%, which rounds to the 76.4% on my results page. The 2020 edition gives 76.0%, and the 2025 edition gives 74.9%. Neither matches. So my NYC results page was computed on the 2015 tables.

That's a decade-old edition. The marathon record had fallen three times since it was published. Eliud Kipchoge ran 2:01:39 in 2018 and 2:01:09 in 2022, then Kiptum went 2:00:35 in 2023. My official grade never saw any of them. It was still dividing into the 2014 world.

That's how the tables work. Alan Jones revises them in discrete editions, not the day a record falls. Between editions the standard sits frozen while the real record moves under it. A results page can be a full edition or two behind without anyone noticing, because the percentage still looks perfectly reasonable. Mine did. It was just measuring me against a world from 2014.

Here is my race run through both denominators — the 2015 edition my results page used, and the 2025 edition on RaceRecords.

Almost all of that gap is the world record. The age factor for a 48-year-old man over the marathon barely moved between editions — 0.9117 on the 2015 table, 0.9127 on the 2025 one. That 0.001 nudge is worth about a tenth of a point. The rest of the point and a half is the standard racing away from 2:02:57 down to 2:00:35.

So the mechanism is boring, which is the good news. Nobody rescored my aging curve in any meaningful way. The elite ceiling got 2 minutes and 22 seconds faster, and I got measured against the newer, harder ceiling.

Which number is correct — 76.4% or 74.9%?

Both are correct on their own terms. 76.4% is a true statement about where 2:56:19 sits relative to a 2:02:57 world best. 74.9% is a true statement about where the same time sits relative to a 2:00:35 world best. The score is only ever a ratio, and the ratio depends on the benchmark you pick.

But if the question is "which world record does the current table use," there's only one answer. Alan Jones 2025 standardises on Kiptum's 2:00:35, the record when the edition was published. The real record has since fallen again — Sawe's 1:59:30 at London 2026 — but the tables haven't caught up, so 2:00:35 is still the number RaceRecords divides into today.

RaceRecords always grades against the current published table, so every race I've ever logged sits on one comparable scale (RaceRecords methodology). That's a deliberate choice. The alternative — freezing old races on old tables — would mean my 2019 marathon and my 2025 marathon couldn't be compared, because they'd be measured against different worlds.

Will this happen with every race I import?

Yes, and it's predictable. Any road result you bring over from a results page grading on an older edition will read a little lower on RaceRecords, because RaceRecords is on Alan Jones 2025. The gap is largest at the marathon, where the world record moved most. It's smaller at the shorter distances, where the open-class bests barely changed between editions.

The size also depends on your time. The faster you ran, the more the shifting benchmark matters, because you're closer to the ceiling that moved. A mid-pack finisher might see 0.3 of a point. Someone pressing toward local class, like me, sees more.

None of it touches your finish time. 2:56:19 is 2:56:19 on every page that will ever exist. Only the derived percentage moves, and it moves the same way for everyone graded against the same table.

The gap also closes on its own eventually. Nothing stops a future NYC results page from adopting Alan Jones 2025 too, and the moment it does, its number and the RaceRecords number line up. Until then you don't have to guess which edition you're looking at. Log your race on RaceRecords and you'll see the current-table score straight away, with the edition printed on a chip next to it — so you can check your official percentage against the current one instead of wondering why they disagree.

What should I read off my age grade, then?

Read the trend, not the third decimal. An age grade is only meaningful against another age grade computed the same way. My 74.9% is directly comparable to every other race on RaceRecords, because they all sit on Alan Jones 2025. It is not directly comparable to a 76.4% from a page on an older table, any more than two thermometers on different scales.

So when the two numbers disagree, don't ask which site is broken. Ask which world record each one is dividing into. Then use whichever set is internally consistent for the comparison you actually care about — your race this year against your race last year.

And this isn't only an NYC problem. Every race that prints an age grade picks its own table edition, and they don't all pick the same one. My Berlin result might be scored on one version, my Chicago result on another, my local half on a third. Line those official percentages up side by side and you're comparing thermometers on different scales again — the gaps between races are partly real, partly just table vintage. There's no way to tell which is which from the results pages alone.

That's the case for putting them all through one engine. Log every race on RaceRecords and each one is graded on the same current table, so a difference between two of your races is a difference in how you ran — not an accident of which organiser updated their software when.

For me that's the whole point of grading a lifetime of results against one current table. See where your last race lands on your own age-graded history.

What happens when the record falls again?

It already has. Sebastian Sawe ran 1:59:30 at the 2026 London Marathon, the first marathon world record under two hours. Alan Jones 2025 doesn't know about it yet — the current table still standardises on Kiptum's 2:00:35, and the next revision will adopt 1:59:30.

When it does, my 2:56:19 drops again. To isolate the effect of the record alone, here's my race scored against five marathon world records — holding the age factor fixed at 0.9127, so the only thing changing row to row is the benchmark.

Same 2:56:19. Same 48-year-old. Same age factor on every row, on purpose — so you can watch the percentage fall from 76.4% to 74.3% on the strength of the record alone. I never ran a step differently. That's not deflation, it's honesty. The ceiling I'm measured against keeps rising, so the same climb reads as a slightly smaller fraction of it.

And 74.3% is only today's answer. The record could fall again before Alan Jones publishes his next revision — the tables always lag the track by months, sometimes years. So the bottom row isn't a floor. It's just the newest world best that the maths happens to have caught up with. Whenever a faster one lands and gets adopted, my number ticks down once more, and my race stays exactly what it was.