REF-820 · Smyth, 2018 · Journal of Sports Analytics, Vol. · Paper

Fast starters and slow finishers: A large-scale data analysis of pacing at the beginning and end of the marathon for recreational runners

B+moderate evidenceMeasured on recreational to trained athletesWritten before October 2026, being re-checked

An analysis of 1.7 million recreational marathon runners, focused on pacing at the start and at the end of the race. Starting or finishing too quickly went with poorer finish times: fast starts tended to be very fast and led to endurance problems later, while fast finishes suggested overly cautious pacing earlier in the race.

Women tended to pace more effectively than men, but could also be more cautious, which cost them minutes overall. The author presents the findings as a way to quantify the cost of uneven pacing at the start and end of a marathon.

Participants

1.7 million recreational marathon runners (race records from big-city marathons). The abstract does not give the races, years, ages or the male-female split.

Protocol

Retrospective analysis of pacing (split) data from big-city marathons.
Two race stages examined: the start and the end of the marathon.
Runners compared by how fast they started and finished relative to the rest of their race, and by sex, against finish time.
The abstract does not give the definitions of a fast start or a fast finish.

Key Results

Fast starts went with poorer finish times; they "tend to be very fast", leading to endurance problems later in the race.
Fast finishes also went with poorer finish times, read by the author as a sign of overly cautious pacing earlier.
Women paced more effectively than men, but were sometimes more cautious, which cost them minutes overall.
The abstract gives no effect sizes, percentages or minutes for any of these comparisons.

Practical Application

Across 1.7 million recreational marathon runners, both a fast start and a fast finish went with slower finish times: the fast starters paid for it later, and the fast finishers had held back too much earlier. Aim for an even race: go out at goal pace, not faster, and treat a big sprint over the last kilometres as time left on the course rather than a sign of a well-judged race.
For men especially: the abstract reports that women paced more effectively, so a male runner's default risk is the fast start.
For cautious runners: the abstract says women could be too cautious, costing minutes. Inference, not tested here: if the second half is much faster than the first, the next race's opening pace can be a little bolder.
Not from this paper; the base's own practice: set the opening pace from a realistic goal rather than a hoped-for one, and check the first split early; the late-race collapse that follows a fast start is described in REF-775.

Limitations

Observational race records: the pacing pattern and the finish time come from the same race, so a slow finish time may cause the "uneven" profile as much as the reverse.
Definitions of fast start and fast finish, and the size of the time costs, are not in the abstract.
Course, weather and runner ability cannot be fully separated in race data.
Not indexed in PubMed; written from the Crossref abstract and record metadata only (full text not read).
How this entry was checked and changed
Why this level

very large observational dataset; associations only, and the abstract reports no effect sizes