Which parkrun 5K series was most consistent in finish times?

Observed in SEL data: Huddersfield parkrun had the lowest coefficient of variation of instance medians, CV 1.4444% across 229 instances (across-instance median 29:04, median field 574).

Which parkrun 5K series was the most variable?

Observed in SEL data: among well-populated series, Rondebult parkrun showed the highest observed CV of instance medians at 17.8259% across 83 instances after guarding out series with thin participation or high missing-time shares.

How many parkrun series and result records were analysed?

Observed in SEL data: the analysis ranked finish-time consistency across 1,039 parkrun series with at least 30 observed instances (of 1,047 series seen in 2020-01-04 to 2026-06-13) using 43,032,659 observed result records.

What share of records had valid finish times?

Observed in SEL data: 94.75% of observed result records carry a valid finish time; 2,258,480 records (5.25%) are missing or zero and were excluded from finish-time metrics.

Does larger participation relate to finish-time consistency?

Observed in SEL data: across 1,008 ranked series the relationship between median participation and CV of instance medians is -0.454 (Pearson), indicating larger fields tend to show lower observed CVs.

Does this study explain why some series are more consistent?

No — this analysis is descriptive only. It reports observed consistency across series and does not explain causal factors or make claims about individual runners.

parkrun · Consistency

Which parkrun 5K series show the most consistent finish-time patterns?

Identity-linked results Descriptive, unadjusted Grounded against live data Full audit trail

The evidence base

43,032,659

observed result records across 1,039 parkrun 5K series, recorded between 2020-01-04 and 2026-06-13. Every figure in this study traces back to them — observed, not inferred.

1,047

parkrun series observed

201,915

observed event instances

1,039

series ranked (≥ 30 instances)

94.75%

records with a valid finish time

Why consistency is worth measuring

What the shape of a field tells us

Some recurring parkrun 5K series show very stable finish-time distributions from week to week. Others vary much more.

That matters because the shape of an event's field tells a story about the event itself — not about any individual runner. This study shows which parkrun series have the most stable observed field patterns, where finish-time distributions shift more from one instance to the next, and how participation size relates to stability.

It is descriptive only. It describes what was observed — not why a series is stable, and never a claim about any single runner.

We ranked finish-time consistency for the 1,039 parkrun series with at least 30 observed event instances, drawing on 43,032,659 observed result records. Records with a missing or zero finish time are set aside for finish-time metrics only.

The steadiest series, and the most variable

Coefficient of variation of instance medians

Consistency is measured by the coefficient of variation (CV) of each series' instance medians — how much the typical finish time drifts from one event to the next. A lower CV means a steadier field.

At the steady end sits Huddersfield parkrun — a long-running Yorkshire fixture whose Saturday-morning field barely shifts. Its instance medians vary by a CV of just 1.4444% across 229 instances, with an across-instance median of 29:04 and a median field of 574. At the other end, among well-populated series, Rondebult parkrun is the most variable.

Most consistent vs most variable seriesCoefficient of variation of instance medians — among 1,039 ranked parkrun 5K series CV of instance medians

Huddersfield parkrunmost consistent · 229 instances · median 29:04 · field 574 1.4444%

Rondebult parkrunmost variable (well-populated) · 83 instances 17.8259%

Most consistent (lower CV) Most variable (higher CV) Shorter bar = steadier field

Bar length shows each series' coefficient of variation of instance medians. Huddersfield's field is among the steadiest observed; Rondebult's varies most among well-populated series.

Observed, not explained. Rondebult is named the most variable only after guarding out series with thin participation or a high share of missing times. We describe the pattern; we make no claim about why a series is steady or variable.

Bigger fields tend to be steadier

Participation and variability, across 1,008 series

Across the 1,008 series eligible for ranking, larger fields tend to show lower variability. The observed relationship between a series' median participation and the CV of its instance medians is negative — as fields grow, week-to-week drift tends to fall.

Participation and consistency move togetherMedian participation × CV of instance medians — 1,008 ranked series Pearson r

−0.454Pearson correlation

A moderate negative relationship: larger fields tend to be more consistent, with lower variability from one event to the next. Median participation across ranked series is 150, ranging from 20 to 1,290.

20 min field150 median1,290 max field

The relationship is observed, not explained — it does not tell us why larger fields drift less. No individual series points are shown.

How this study was done

Methodology

The analysis focused on all parkrun event series in the results dataset from 2020-01-04 to 2026-06-13. It examined finish-time records, excluding instances with missing or zero times, to compute consistency metrics across various event series with a minimum of 30 occurrences.

Consistency is assessed only for the 1,039 parkrun series with at least 30 observed instances in 2020-01-04 to 2026-06-13; 8 thinner series are excluded.
Ranked lists additionally require median participation ≥ 20, a high-missing instance share ≤ 50%, and valid records ≥ 100, so tiny or noisy series cannot dominate; 31 threshold-passing series failed a guard.
2,258,480 of 43,032,659 observed records (5.25%) have a missing or zero finish time and are excluded from finish-time metrics only.
Week-to-week change is the per-series median of absolute differences between chronologically adjacent instances (median finish time, and the mean absolute change across the five finish-time band shares).

What this study can and cannot say

Limitations

This study describes the consistency of parkrun event series across their observed event instances. Consistency here is a property of a series' observed finish-time distributions from one instance to the next, not of any person.

It counts observed result records, not verified unique runners; no individual runner is tracked across events, and it makes no retention, progression or personal-history claim.
Parkrun only: non-parkrun series are excluded; this describes parkrun 5K series, not running events in general.
Records, not people: all metrics are computed from observed result records — a result record is not the same as a person, and no individual is tracked across events.
Not representative: SEL holds a syncing, partial-coverage sample of parkrun results; series coverage and instance counts are lower bounds, and absent series or instances may shift any series' apparent consistency.
Descriptive, not causal: consistency is described, not explained — course profile, seasonality, event age, local attendance and source-coverage gaps may all affect a series' observed finish-time stability.
Consistency metrics depend on the chosen instance threshold and finish-time bands; the design's sensitivity checks bound but do not eliminate this dependence.

This article is generated from a Sports Evidence Lab analysis with a complete audit chain: research design v8 (approved) → data grounding pass → deterministic analysis run → reviewed article draft v9. Every figure traces to a recorded query.