Nicolas Sitter
AGENTIC|
Projects

The archive

Marathon de Paris — every edition

12 editions, 520,581 classified finishers, 2013–2026. Every figure here is read from the results the timekeeper published.

Editions
12
2013–2026
Finishers
520,581
classified, all editions
Women
25.3%
131,798 of 520,581
Biggest field
57,516
2026
First namesGenerationsWomenRegulars
The size of the field

Classified finishers per edition. The axis starts at zero, so the height of a point is the size of the field.

020,00040,00060,0002013201520172019202120232026Classified finishers per edition, 2013 to 2026

The size of the field. Use the left and right arrow keys to read each point.

Show the numbers
YearFinishers
201338,686
201438,526
201540,262
201641,791
201741,099
201842,377
2019—
2020—
202127,114
202233,599
202350,778
202453,899
202554,934
202657,516
The winning time

Fastest finisher, over the same 42.195 km every year — so unlike a mountain race, these times are directly comparable.

2:04:002:05:002:06:002:07:002:08:002013201520172019202120232026Winning time per edition, 2013 to 2026

The winning time. Use the left and right arrow keys to read each point.

The scale starts at 2:04:00, not zero — these are finishing times, and a clock plotted from zero is a flat line. Read the gaps between points, not their height above the axis.

Fastest of the archive: Elisha ROTICH in 2021, 2:04:21 (2:57/km).

Show the numbers
YearWinner
20132:05:38
20142:05:03
20152:05:49
20162:07:11
20172:06:10
20182:06:25
2019—
2020—
20212:04:21
20222:05:07
20232:07:15
20242:05:33
20252:05:25
20262:05:18
The median finisher

Half the field finished faster than this, half slower.

gun time4:00:004:30:005:00:005:30:002013201520172019202120232026Median finishing time per edition, 2013 to 2026

The median finisher. Use the left and right arrow keys to read each point.

The scale starts at 4:00:00, not zero — these are finishing times, and a clock plotted from zero is a flat line. Read the gaps between points, not their height above the axis.

Show the numbers
YearMedian
20134:38:08
20144:42:54
20154:51:19
20165:08:41
20175:01:48
20185:09:45
2019—
2020—
20214:03:22
20224:00:15
20234:00:38
20244:06:41
20254:06:42
20264:08:47
Women in the field

Share of classified finishers.

0%10%20%30%40%2013201520172019202120232026Women as a share of classified finishers per edition, 2013 to 2026

Women in the field. Use the left and right arrow keys to read each point.

Show the numbers
YearWomen
201320.4%
201420.5%
201523.8%
201624.2%
201723.9%
201824.2%
2019—
2020—
202121.3%
202224.2%
202325.7%
202427.3%
202529.8%
202631.8%

Where the timing breaks down

Not every edition supports every page. These are the ones whose data would not carry the analysis the others get, and what went wrong in each — stated here rather than left for a reader to infer from a missing table.

2023results only

Results only. Every runner has all ten mats and not one chain adds up: the intervals sum to 97.4% of the finish time, which by the feed's own segment speeds is 40.1 km of 42.195 — the km-30 leg measures 3.9 km where the course says 5. Publishing them would put every late-race split about three minutes early.

The clock changed

The editions from 2013 to 2018 are ranked on gun time — measured from the starter’s gun, however long a runner waited to reach the line. From 2021 the official classification switches to chip time.

It is not a detail. In a field this size the later pens cross the start line more than an hour after the gun, so the two clocks disagree by more than the gap between the median finisher and the top hundred. Each edition here is ranked the way its own results were published, established by testing which time field reproduces the provider’s own ordering — not by trusting the flag in the feed, which is honest on some editions and wrong on others.

Missing years

2019Raced, but the timing provider cannot serve it: the race and its results both answer HTTP 500 on every attempt, while the editions either side of it answer 200 from the same host in the same second. The event resolves and reports 47,495 entrants — the results themselves are what is unreachable. Absent at source, not missing from this import.
2020Cancelled — the pandemic. The 2021 edition ran in October rather than April, which is why its field is the smallest of the modern years.

Every edition

YearFinishersWomenWinnerWinning timeMedianClock
201338,68620.4%PETER SOME2:05:384:38:08gun
201438,52620.5%Kenenisa BEKELE2:05:034:42:54gun
201540,26223.8%MARK KORIR2:05:494:51:19gun
201641,79124.2%Cybrian KOTUT2:07:115:08:41gun
201741,09923.9%Paul LONYANGATA2:06:105:01:48gun
201842,37724.2%Paul LONYANGATA2:06:255:09:45gun
202127,11421.3%Elisha ROTICH2:04:214:03:22chip
202233,59924.2%Deso GELMISA2:05:074:00:15chip
202350,77825.7%Abeje AYANA2:07:154:00:38chip
202453,89927.3%Mulugeta UMA2:05:334:06:41chip
202554,93429.8%Benard BIWOTT2:05:254:06:42chip
202657,51631.8%Yemaneberhan CRIPPA2:05:184:08:47chip

What this archive will not tell you

How many people abandoned. Every edition here lists classified finishers only — there are no DNF rows to count, so an abandon rate computed from these numbers would come out as zero for every year.

About this project and its data. The results are Sport Innovation / TimeTo’s — the official timekeeper of the event. They were fetched once after the event and republished here with segment times, recomputed leaderboards and a page per athlete that the official results do not offer. The timing is their work; the analysis is mine, and any mistake in it is mine too. Course details come from Amaury Sport Organisation (A.S.O.), the organiser. Each edition is ranked as its own official results were — gun time on the older years, chip time from 2019 — established per edition against the published ordering. Athletes who asked to be anonymous in the official results stay anonymous here.

520,581 finishers across 12 editions · built 2026-09-07