Dissecting the Chicago Marathon: Seven World Records, One Flat Course, and the Variables Nobody Counted

I open with terrain data, not with emotion. The 2026 Chicago Marathon is the 48th edition by the organisers' count — a figure I will return to at the end. What deserves attention first is the course.

Chicago's cumulative elevation gain is roughly 74 metres, with 72 metres of descent. The high point sits at 183 metres above sea level, the low point at 175 metres. The net spread is 8 metres across the full 42.195 km. Berlin, the course the media by default calls the fastest on the planet, has an elevation range of 21 metres. Chicago is nearly three times flatter than Berlin by range. That is measured data, and it is the starting point for everything that follows.

Context: a stage designed for the clock

The Chicago Marathon was founded in 2026. For nearly half a century it has positioned itself not as the place where Olympic or world titles are decided, but as the place where people race the clock. Its calendar slot falls in late October, when temperatures across the American Midwest have dropped to a level favourable to endurance. That is a systemic advantage, not random luck.

Structurally, Chicago belongs to the World Marathon Majors group. Traditionally that group comprises six races: Tokyo, Boston, London, Berlin, Chicago and New York. Sydney is a later addition. I spell this out because the source document I am cross-referencing lists Boston, New York, London, Berlin, Tokyo and Sydney — while omitting Chicago, that is, omitting the very subject of the piece. For anyone working with data, a broken link in a list is a signal to re-verify the entire chain.

The 2026 field exceeded 55,000 runners, the largest in the race's history. That is a commercial marker, not a competitive one. A rising mass-participation number says nothing about the depth of the elite group. Those two curves have to be drawn separately.

I also note a technical detail: the source document at one point gives the distance as 42.2026 km. The standard marathon distance is 42.195 km. This is almost certainly a typo, but it reminds me that every figure in a retrospective feature must be cross-checked before it is quoted again.

The data core: seven times in four decades

Chicago's central evidence chain consists of seven world records set across four decades: 2026, 2026, 2026, 2026, 2026, 2026 and 2026. Seven times in forty years is a sample large enough that it cannot be attributed to chance. Before I trust a reputation, I need to see the data behind it — and here the data exists.

In 2026, Paula Radcliffe ran 2:17:18 in Chicago. A year later, in London, she lowered it to 2:15:25. The Chicago mark was therefore not Radcliffe's final peak but her launchpad. Based on my experience tracking road races over more than twenty years, this pattern recurs often enough that I treat it as a working rule: a fast course does not produce the final record, it produces the conditions under which an athlete dares to try.

On 8 October 2026, Kelvin Kiptum ran 2:00:35 in Chicago. It was the third marathon of his career. He broke Eliud Kipchoge's world record of 2:01:09. At 23, Kiptum became the first man to break 2:01 officially. In February 2026 he died in a road accident. His career stopped at exactly the point where the performance curve was climbing.

On 13 October 2026, Ruth Chepngetich ran 2:09:56 in Chicago. She became the first woman to run a marathon under 2 hours 10 minutes, at an average pace of roughly 3 minutes 05 seconds per kilometre. Chepngetich had won Chicago in 2026, 2026 and 2026.

Then a different fact cut across that sequence. In October 2026, the Athletics Integrity Unit handed Chepngetich a three-year suspension after a positive sample for hydrochlorothiazide — a diuretic commonly used as a masking agent — taken on 14 March 2026. The 2:09:56 record remains on the books, because the positive sample post-dates the record.

I worship data, but I pray through verification. Here, verification yields a two-tier conclusion: on paper, 2:09:56 is a world record; in value terms, it is a performance carrying an asterisk that has not been removed. Readers are entitled to both tiers at once.

By 2026 the men's picture had changed entirely. On 26 April 2026, at the London Marathon, Sabastian Sawe ran 1:59:30 — the first men's world record under two hours in record-eligible conditions. Previously, a sub-two mark had only been achieved in an exhibition event ineligible for record ratification. In the same London race, Yomif Kejelcha ran 1:59:41 and Jacob Kiplimo ran 2:00:28.

Three men under 2:01 in a single race. The all-time list now contains three athletes faster than Kiptum, and all three came from the same morning in London. A season should be read as a sequence of probabilities, not a sequence of events. This is precisely a probability sequence that has shifted: not one mutant individual, but an entire cohort crossing the threshold together.

The counter-intuitive angle: a flat course does not generate records by itself

This is where I part company with the conventional telling.

A flat course is a necessary condition, not a sufficient one. Chicago has an elevation range of 8 metres — almost negligible across 42.195 km — but records are not born from terrain. They are born from four other variables: the quality of the lead group, the depth of the pacing group, race-day weather, and shoe technology.

Of those four, shoe technology is the largest and the least quantified. The source document mentions running shoes as one driver of marathon progress, but stops at the mention. The common estimate among analysts is that carbon-plated shoes deliver an advantage of one to four minutes at elite level. If that figure holds, most of the performance improvement of the past decade belongs to equipment, not to biology.

Put another way: when a course is called fast, one must ask whether it is fast because of terrain or fast because of shoes. Chicago is the case where both are true, which is why it is a good laboratory — and also why it is a laboratory prone to misreading.

There is one more point. The figure of seven world records almost certainly counts only mixed-race or outright records, excluding the separate women's-only world bests that existed historically. That is a definitional choice, and it inflates Chicago's headline number. It is not technically wrong, but it needs to be stated.

On Chepngetich, my conclusion is not aimed at her personally. It is aimed at how the system keeps its books. A record retained because the positive sample falls outside the time window is a record that is lawful but reputationally fragile. The biological passport could reopen that window. Until it does or does not, the number 2:09:56 should be read with a footnote.

And on Kiptum, I refuse to speculate. He was never named in any file. But a 2:00:35 in a third career marathon is the steep-progression pattern for which analysts always want more verification data. Noting that is the duty of anyone working with data, not an accusation.

Finally, the number 48. The race was founded in 2026. By 2026, counting the 2026 cancellation for the pandemic, the actual number of editions would normally fall around 49 to 50. The figure of 48 in the source may be correct if the organisers count differently, or it may be wrong. I leave it flagged as pending verification.

Takeaway: signals for the next cycle

Three signals I will be tracking in the coming cycle.

First, Sawe's 5 km splits at London 2026. If the second half is markedly faster than the first, there is still headroom above 1:59:30. If the second half is slower, that is a mark optimised to its limit.

Second, the regulatory framework for racing shoes. Any change to stack height or carbon-plate structure will rewrite the entire performance curve, and Chicago is where that curve surfaces earliest.

Third, the biological-passport review window for Chepngetich. A decision to reopen the file would turn 2:09:56 from an asterisked record into an annulled one.

Luck is the residual my model cannot explain — and I never round it to zero. But luck also keeps a file. My job is to read that file before it gets printed as a headline.