Average half marathon time

Runsy's synthesis: the average half marathon finish time is around 1:59:40 for men and 2:19:13 for women. The one source publishing an overall figure gives a median of 2:14:59.

The men's figure is the most confidently sourced number anywhere on this site. Two studies, built from different samples and reporting different statistics — one a median, one an average converted from pace — land fifteen seconds apart at 1:59:48 and 1:59:33.

The women's figure from the same two studies is nine minutes and forty seconds apart. We do not have a confident explanation for the asymmetry, and rather than average it away quietly we are putting it at the top of the page. If you are a woman trying to work out whether your 2:18 is above or below average, the honest answer is that it lands squarely inside the range where the published sources cannot agree.

There is no single dataset that reports finish-time distributions for road running. So instead of picking one number and calling it the average, we did this:

1. Collected every published central figure we could trace to a named source with a stated sample size.
2. Discarded anything we could not trace to its origin. A magazine quoting a statistic is not a source; we followed each figure back to the study that produced it, and dropped it when the trail went cold.
3. Took the unweighted mean of the surviving figures. Unweighted, deliberately: the samples overlap heavily and are not independent, so weighting by sample size would just hand the answer to whichever study counted the most rows.
4. Published the range as well as the midpoint. Where the sources disagree, the disagreement is the finding.

The result is Runsy's synthesis of published data. It is not a measurement, and we do not present it as one. It is the middle of what the published record actually says, with the edges shown.

Two things the arithmetic cannot fix

We are averaging means with medians. The percentile study reports a 50th percentile; the others report an average. Finish-time distributions have a long slow tail, so the mean sits later than the median. Mixing them widens our range and pulls the midpoint slightly late. We say so rather than quietly picking whichever suits.

Some figures are converted from a published pace, not a published time. Where a source reports average pace per kilometre, we multiply by the race distance. That is arithmetic on their number, not a new number — but the mean of a set of paces is not identical to the pace of the mean time, so treat converted figures as the coarser of the two kinds. Each one is labelled below.

Every number below is one we could trace to a named publisher who states what they measured and how many results they measured it over. Nothing here is modelled, estimated or carried over from a secondary write-up.

• RunRepeat, Compare Running Finish Times
  Measures: Half marathon finish-time percentiles, overall and by sex. Sample: 35M results across all distances, 28,000+ races. Years: “last 20 years”, no end date given. Reuse terms: Publisher states material may be reused with attribution to the original report
• RunRepeat / IAAF, The State of Running 2019
  Measures: Average half marathon pace by sex and by nation. Sample: 107.9M results across all distances, 70,000 events, 193 countries. Years: 1986–2018. Reuse terms: “You may use material from this report as long as you refer to this original report.”
• Nikolaidis, Cuk, Rosemann & Knechtle (2019), IJERPH 16(10):1777
  Measures: Mean running speed by age group and sex, Ljubljana 2017 half marathon. Sample: 7,258 half marathon finishers (plus 1,853 marathon finishers). Years: 2017. Reuse terms: Peer-reviewed, open access under CC BY 4.0

The first two are the same publisher; the third, Performance and Pacing of Age Groups in Half-Marathon and Marathon (Nikolaidis, Cuk, Rosemann and Knechtle, IJERPH 2019), is independent of them but is a single race in a single year.

The Ljubljana figures sit outside the synthesis on purpose. They are one European city race in one year and both are markedly faster than the global numbers, which is what a single well-organised destination half marathon looks like. They are here as a third reading, not a third input — and the fact that a real race can be eight minutes quicker than the global median is itself worth knowing before you compare yourself to any average.

One further note on the mean-speed conversion: the paper publishes mean running speed in metres per second (3.14 for men, 2.77 for women), and the time equivalent of a mean speed is not the same as the mean of a set of times. Treat those two figures as approximate.

A likely error in one of the published tables

The percentile study's men's half marathon column reads 1:58:16 at the 40th percentile, 1:59:48 at the 50th and 2:09:58 at the 60th. A 92-second step followed by a ten-minute step is not a shape finish-time distributions have. We think one of those cells is wrong, we do not know which, and we have reproduced them as published rather than silently correcting somebody else's table. It is a reason to hold the men's median a little more loosely than its two-source agreement suggests.

• An age-by-sex table of half marathon finish times. The peer-reviewed source we have publishes mean running speed by age group for one race, not finish times for a representative field. Converting one city's speeds into a table headed “average half marathon time by age” would be presenting a single race as the world. We did not.
• An explanation for the nine-minute women's gap. We can see it; we cannot account for it from what is published. Saying so is more useful than picking a side.
• A corrected version of the percentile table. We flagged the implausible cell rather than editing another publisher's data.
• Anything from parkrun or its community scrapes and dataset mirrors: their terms prohibit reuse.

For a comparison that adjusts for your age and sex against a maintained published standard rather than against whoever entered, an age-graded percentage is the sounder instrument.