Sports Evidence Lab
Evidence study 5 km · recurring events Observed, never predicted
Return timing · 4+ week breaks first finishes after break

How Timing Changes Affect First Finishes After a Break in 5 km Events

SEL's calibrated evidence links 600,422 5 km result records to 152,831 observed participants across 820 venues — high-confidence, same-venue only. We compare each participant's first finish after a four-week-plus break to their own pre-gap baseline.

Identity-linked results Descriptive, unadjusted Grounded against live data Full audit trail
600,422
Evidence5 km result records linked
152,831
Populationobserved participants
820
Reachrecurring venues
170,463
Episodesreturn-gap episodes measured
01 — What this study describes

A break, then a return — measured against your own recent self

Strong descriptive finding

Most running stories are told against other people. This one is told against yourself — your own recent baseline, taken just before a gap of four weeks or more.

From 600,422 high-confidence, same-venue 5 km result records, the resolver linked 152,831 observed participants across 820 venues. Within that linked population we found 170,463 return-gap episodes — a break of 28 days or more, then a return to the same recurring event series — drawn from 47,336 observed participants. Each first return is described against that participant's pre-gap baseline: the median of their last three valid pre-gap finishes.

Organisers and runners can use this descriptive evidence to see how a return result relates to an individual's recent baseline in a linked, high-confidence sample. It reports what was observed — it does not imply a cause and does not extend beyond the linked sample.

The return curveMedian first-return change vs each participant's pre-gap baseline, by gap band — 170,463 episodes return_gap_first_return_delta_pct_by_gap_band
0% · pre-gap baseline −0.36% −0.08% +0.33% +0.98% +2.61%
← shorter breakmedian Δ as % of baseline · seconds belowlonger break →
After short breaks the first return sits essentially on baseline (−0.36% at 28–55d). As the gap widens it rises to +2.61% after 365 days or more.
First-return change by gap band
Gap band Median Δ (s) Median Δ (%) Pre-gap baseline (s)
28–55 days−6−0.36%1,763
56–83 days−1−0.08%1,761
84–167 days+6+0.33%1,755
168–364 days+17+0.98%1,747
365 days++43+2.61%1,731
First-return delta = first return finish minus the pre-gap baseline (median of the last three valid pre-gap results). Values shown in seconds and as a percentage of baseline.
A short break barely shows. A long one shows up as seconds, then a couple of per cent — observed, against your own recent baseline.

Observed. These are descriptive, unadjusted medians over high-confidence linked return-gap episodes. We make no claim about cause — only what was measured at the first return after a break.

+2.61%median first-return change after a break of 365 days or more (+43 s)
The widest gap band

After a year away, the first 5 km back is a couple of per cent off baseline.

Across the 365-days-plus band the median first return came in +43 seconds on a median pre-gap baseline of 1,731 seconds. The shortest breaks barely register — 28–55 days returned at −0.36%, essentially on the baseline.

+2.61%
365 days+ · first return
−0.36%
28–55 days · first return
02 — A second cut of the same episodes

Faster baselines showed a slightly larger percentage change

Re-grouping the same return-gap episodes by each participant's pre-gap baseline speed gives a different view. Read as a percentage of baseline, the change at first return was a little larger for faster baseline bands and a little smaller for slower ones — from +0.84% under 20:00 to −0.08% at 35:00 and slower.

Median first-return change, by baseline-speed bandSame episodes, grouped by pre-gap baseline finish time median % vs baseline
Dots left of the dashed line sit below baseline (faster); right of it, above baseline (slower).
As a percentage of baseline, the median change ran from +0.84% (under 20:00) to −0.08% (35:00+).
Baseline band Episodes Baseline (s) Median Δ (s) Median Δ (%) Within 60s Within 2%
Under 20:003,1461,154+9+0.84%71.2%37.3%
20:00–24:5931,8111,396+9+0.64%60.6%32.7%
25:00–29:5959,3761,652+7+0.45%51.0%30.5%
30:00–34:5942,6931,926+2+0.11%42.6%28.6%
35:00+33,4372,354−2−0.08%31.5%25.6%
Faster baseline bands held closer to baseline: 71.2% of under-20:00 episodes returned within 60s, versus 31.5% at 35:00+. Within-2% shares are computed over complete post-return sequences (1,366 / 15,162 / 28,173 / 20,031 / 15,657 sequences).
03 — Complete sequences only

After the first return, a small recovery

Of the 170,463 return-gap episodes, 80,389 form a complete post-return sequence — three valid follow-ups. For those, the median recovery change after the first return was modest and broadly similar across gap bands, easing back toward baseline.

Median post-return recovery change, by gap bandComplete post-return sequences only — three valid follow-ups n = 80,389 sequences
Within complete post-return sequences, the share returning within 60s of baseline ran 53.5% (28–55d) → 34.4% (365d+); within 2%, 34.5% → 19.6%.
04 — An early, coverage-limited read

Within this linked population, one journey shape dominates

Exploratory · limited coverage

The journey-pattern mix is dominated by a single shape. It is an early read on the linked population — not a headline claim about all runners — and the mix will shift as coverage grows.

Journey pattern types, observed participantsShare of the linked population journey_pattern_types
Small-cell suppression applied: 3 cells withheld (threshold 10 observed participants). Cross-venue patterns are not yet calibrated and are not analysed here.

How we ran this study

Methodology · in full view

We used SEL's article‑eligible, high‑confidence observed‑participant 5 km linkage (resolver sel-repeat-profile-resolver-v2). Return‑gap episodes were defined where an observed participant had a gap of at least 28 days at the same recurring event series and at least three valid pre‑gap finishes so a pre‑gap baseline (median of the last three valid pre‑gap finishes) could be computed. The dataset temporal coverage runs from **2020-01-04** to **2026-06-06**. Episodes were assigned to five gap bands (28–55 d, 56–83 d, 84–167 d, 168–364 d, 365+ d) and to five baseline‑speed bands (under‑20, 20–24:59, 25–29:59, 30–34:59, 35+). First‑return deltas were computed as first_return_time minus pre_gap_baseline (seconds and percent). Post‑return recovery metrics use only complete sequences with three valid post‑return finishes.

Coverage: 152,831 observed participants / 600,422 linked results; 820 distinct recurring event series (814 reported for return-gap episodes). Source provider: clubrunnr.

What this study can and cannot say

Limitations · every one kept

This study uses observed participant identity: confidence‑scored, evidence‑based linkage of public result records that may belong to one person. It is NOT verified or legal identity and NOT user‑account matching.

Not representative of parkrun as a whole: SEL holds a partial, club‑oriented sample of public results.

Return‑gap episodes: same recurring event series only; a return‑gap episode is a 4+ week break then a return, with the pre‑gap baseline taken as the median of the last 3 valid pre‑gap results. Post‑return recovery is described only for complete post‑return sequences (3 valid follow‑ups).

  • This study uses observed participant identity: confidence-scored, evidence-based linkage of public result records that may belong to one person. It is NOT verified or legal identity and NOT user-account matching.
  • Only calibrated high-confidence same-venue parkrun linkage is used (stage6-policy-v1, article-usable); medium/low confidence, cross-venue and road-bridge linkage are excluded. All figures are aggregates over observed participant clusters — no individual is identified, named or described.
  • Linked population only: results that could not be linked at high confidence (ambiguous, rejected, thin evidence) are absent, so every share describes the linked population, not all parkrunners.
  • Bounded coverage: the resolver has processed a bounded sample of the corpus; counts are lower bounds and pattern mixes will change as coverage grows.
  • Aggregate flags, not biographies: journey flags are heuristic thresholds over linked appearance patterns; they are never claims about an individual.
  • Not representative of parkrun as a whole: SEL holds a partial, club-oriented sample of public results.
  • Return-gap episodes: same recurring event series only; a return-gap episode is a 4+ week break then a return, with the pre-gap baseline taken as the median of the last 3 valid pre-gap results. Post-return recovery is described only for complete post-return sequences (3 valid follow-ups).
These are descriptive, aggregate measurements from a bounded, high-confidence linked sample. They do not verify legal identity, they do not represent all runners, and they do not support causal claims about why timing changed. A result record is not the same as a verified person; nothing here is prediction, coaching or training advice.

Provenance. This article is generated from a Sports Evidence Lab analysis with a complete audit chain: research design v1 (approved) → data grounding pass → deterministic analysis run (completed 2026-06-13) → reviewed article draft v2. Every figure traces to a recorded SQL query.

The one figure this study cannot show

Where does your own return sit on this curve?

We've placed the bands against 152,831 observed participants. The number left is yours — how your first finish after a break sits against your own pre-gap baseline. It stays personal in, aggregate out, always.

01

Create your profile

A name and an email. Two minutes, no card.

02

Add your 5 km history

So we can see your own pre-gap baseline at your home event.

03

Find where you fit

Your return-gap episodes, placed against this study's bands.

Your data builds your profile and the anonymous cohorts — never a prediction sold back to you.