It is the question everyone asks and almost nobody answers honestly. There are real published numbers, and there is a real problem with all of them — one that a single statistic makes impossible to ignore.
Start with the number that should change how you read the rest
In a study of 32,342 half marathon finishers in Hangzhou, the fastest male age group was 61 to 65. The fastest women were 66 to 70.
That is not a discovery about ageing. It is survivorship. Almost nobody lines up for a half marathon at 65 unless they have been running for decades and are still good at it. The slower 65-year-olds are not in the data because they are not in the race.
Every finish-time statistic below has that problem. They describe people who entered and finished a race — a group selected three times over.
5K: what the published data shows
The best whole-population figure comes from an analysis of 223,224 Australian parkrun participants over three years: an average personal finish time of 33.0 minutes (SD 9.1). Notably, 76.4% of them were “few-timers” who ran only a handful of events, averaging 33.4 minutes.
A UK survey study reports a median of 29:20 — but from a 4.4% survey response rate among people already registered with parkrun, whose authors note responders “might be considered ‘keen’ parkrunners with fewer health conditions than the general population.”
US paid-entry road races run slower: an industry dataset covering 2009–2018 gives an overall average of 39:02 (men 35:22, women 41:21). That analysis explicitly excludes parkrun and elite runners. It is not peer-reviewed and its published percentile table is internally inconsistent, so treat the averages as indicative only.
What does not exist: after repeated searching, there is no peer-reviewed table of 5K median or percentile times broken down by both age group and sex. Any article presenting one is using industry data, and should say so.
Half marathon
From the Hangzhou dataset, converted from the published speeds: men averaged roughly 2:03 to 2:12 across age groups, women roughly 2:12 to 2:26. An industry dataset of 35 million results gives medians of 1:59:48 for men and 2:24:03 for women, though it documents no methodology.
One caution worth recording: the most-cited academic half marathon dataset reports mean marathon speeds faster than its half marathon speeds — implying an average marathon time under three hours across 10,205 female finishers. That is not physically plausible, and we have not used its figures here.
What about a beginner’s first race?
Here is the honest answer: nobody has measured it. Searches for first-time, novice or debut half marathon finishers return injury studies and biomechanics, not finish times. Any “beginners typically run 2:10 to 2:30” figure is coaching folklore.
The closest real data concerns Couch-to-5K. Among graduates who went on to parkrun, the mean time was 34:36 against 29:52 for other parkrunners. But note who is missing: a separate evaluation found only 27.3% of Couch-to-5K starters completed the programme. The people who found it hardest are absent from every dataset on this page.
A better question than “what is a good time?”
Two things are more useful than a population median.
Age grading. World Masters Athletics publishes standards converting your time into a percentage of the best-ever performance for your age and sex. Above 60% is described as local class, above 70% regional, above 80% national. One caveat usually missed: an age grade is not a percentile — a 60% grade does not mean you beat 60% of runners. It compares you to the record, not to the field. 5K is one of the four anchor distances, so its grades are calculated directly rather than interpolated.
Your own previous race. The Riegel formula — T₂ = T₁ × (D₂/D₁)1.06 — predicts one distance from another. A study of 2,303 runners found it “well-calibrated for races up to a half-marathon”, while dramatically under-predicting the marathon, where it ran at least ten minutes fast for half of runners.
So for the specific case of predicting your half from your 5K, Riegel is reasonable. For predicting a marathon from anything, it is optimistic. A pace calculator uses it, with that caveat built in.
And for India
There is no published Indian 5K or half marathon finish-time distribution. parkrun does not operate in India, and the major Indian events do not publish aggregate statistics in a citable form. UK, Australian and US figures cannot be assumed to transfer — different populations, different race cultures, different weather.
The genuinely useful version of the answer: a good time is one that is faster than your last one, at an effort you could repeat. The medians above describe who turns up. They are not a target, and they are certainly not a standard.
References: Grunseit AC, et al. Health Promot Int, 2024. doi:10.1093/heapro/daae098 · Haake S, et al. Br J Gen Pract, 2022. doi:10.3399/BJGP.2022.0001 · Yang S-J, et al. Int J Environ Res Public Health, 2022. doi:10.3390/ijerph19137802 · Relph N, et al. Health Promot Int, 2023. doi:10.1093/heapro/daad108 · Vickers AJ, Vertosick EA. BMC Sports Sci Med Rehabil, 2016. doi:10.1186/s13102-016-0052-y