Marathon Finish Time: London Open Data Derivation
No published paper provides age- and sex-stratified finish-time percentile tables for marathon finishers. The London Marathon open dataset publishes individual finisher records (age category, sex, finish time) under a CC-BY-4.0 licence, so percentiles were derived directly from the raw per-runner rows.
- Data: Per-runner records from the 2018, 2019, 2022 and 2023 editions. The 2020 race was cancelled (COVID-19) and the 2021 edition is excluded as an atypical, COVID-affected field (the only year with more female than male finishers and an inverted pace pattern). Elite athletes and non-finishers are excluded by the dataset authors.
- Quantiles: Empirical percentiles (P5, P25, P50, P75, P95) were computed for each sex within each native age bracket, pooling the four editions. No weighting is applied; every finisher counts once.
- Brackets: The dataset's native age categories are used as published. The source 80-84 and 85+ categories are collapsed into a single 80+ bracket to harmonise across editions.
- Suppressed brackets: The 75-79 and 80+ brackets had too few finishers (under 100 per sex) after the 2021 exclusion to publish stable percentiles, so they are omitted from the norms.
This is an internally derived dataset, not a peer-reviewed publication. The source and its limitations are summarised on the reference page.
Why only the London Marathon
London is used because it is the marathon results collection that is both openly licensed for reuse (CC-BY-4.0) and carries per-runner age, sex, and finish time. Finish times depend heavily on course profile, elevation, and race-day climate, so pooling different races would mix populations these percentiles are meant to separate. A single course is reported instead.
Large results collections do exist for other major marathons, but they are not openly licensed for republishing as derived norms. For readers looking to explore other races, the SCORE Sports Data Repository publishes a 2023 Boston Marathon dataset (age group, sex, and finish times for 26,598 runners) curated for statistics and data-science education; Berlin and New York City results circulate on community data platforms. These were reviewed but are not pooled into the norms here, both for the course-comparability reason above and because their licensing does not cover republishing them as a reference product.