Even Pacing Was Associated With Faster Marathons

Thomas Solomon, PhD

Marathon pacing can shape how the final kilometres feel and how quickly the finish line arrives. This study examined nearly 79,000 Boston Marathon records to see which pacing patterns travelled with faster performances.

Reference: Dong et al. Pacing strategy patterns and performance outcomes in marathon running: A large-scale analysis of split time data. PLoS One (2026) DOI: https://doi.org/10.1371/journal.pone.0352808.

Medical informationThis article is for general educational purposes only and is not medical advice. Supplements and exercise/recovery interventions can cause side effectsAn additional effect of a treatment that is not the main intended effect. It may be harmful, harmless, or helpful, while an adverse effect is specifically harmful. and may interact with medicines, nutrients, or medical conditions. Speak with a doctor or suitably qualified healthcare professional before making changes if you have a medical condition, take medications, are pregnant or breastfeeding, or are unsure what is safe for you.

Study snapshot

A quick, practical summary for runners and coaches.

Quick answer

This cross-sectional studyA cross-sectional study is a type of observational study where the exposure and outcome are measured at a single point in time, giving a snapshot of a population—what’s happening right now. Cross-sectional studies are used in health surveys, prevalence studies, or for hypothesis generation, and can show prevalence (how common something is) and associations (but not cause and effect). E.g., What percentage of runners currently report using recovery supplements, and is use linked to age or training volume? analysed 78,912 finishing records from 3 Boston Marathons. Relatively even pacing was associated with faster finishes and less severe late-race slowing. The study was large, but it cannot show whether pacing itself caused the performance differences.

Key takeaways

  • Even pacing was associated with the fastest marathon times.
  • Fitness, preparation, and experience may partly explain the result.
  • A controlled opening effort remains sensible, but perfect splits are not essential.

How confident should we be?

Evidence confidence: Moderate

The huge dataset, official split times, and detailed analyses strengthen confidence in the overall pattern. The observational design and limited information about the runners prevent a causal conclusion.

Bottom line

Aim for a controlled start and a reasonably steady effort. Do not treat this study as proof that forcing identical splits will make every runner faster, especially on hilly courses.

Read the deep dive below to get a practical interpretation, the wider evidence, some actionable decisions, my thoughts, my rating of perceived scientific enjoyment, and the full study details (research question, study design, participants, methods, results, and the strengths & limitations).

Running science research reviews for endurance runners

The deep dive

The details behind the headline result, including the practical meaning, full findings, limitations, and my interpretation.

Practical meaning

What does this research mean for runners and coaches?

The clearest finding was not that runners must hit every kilometre at precisely the same pace. It was that large, progressive, or erratic slowdowns tended to occur alongside slower marathon performances.

That is an association, not proof that pacing alone caused the difference. Faster runners may also have been fitter, better prepared, more experienced, and more realistic about their target pace. The dataset could not properly untangle those factors.

For runners

For road marathon runners, the results support starting under control and avoiding an early pace that feels suspiciously easy only because the race has just begun.

The practical target is a steady effort, not necessarily an identical pace. The Boston Marathon is a point-to-point course with substantial elevation changes. Hills, wind, congestion, drinks stations, and temperature can all produce sensible fluctuations.

Even the study’s most consistent pacing group slowed slightly. Their average second half was about 5 minutes slower than their first. A modest positive split is therefore not automatically a pacing disaster.

The findings are less directly applicable to trail and ultra runners. Uneven terrain, long climbs, descents, technical sections, aid-station stops, and changes in altitude can make even pace unrealistic. For those runners, controlled effort and sensible energy use matter more than keeping each kilometre identical.

For coaches

The split pattern may be more useful than the final positive split alone.

A gradual fade, a sharp late collapse, and repeated pace fluctuations are different patterns. They may reflect different combinations of fitness, pacing judgement, fuelling, terrain, conditions, or race-day problems.

As a practical interpretation, coaches could review how an athlete’s pace changes across the opening section, halfway point, and final third of a marathon. The study did not test whether this type of review improves performance, but it may help identify repeated pacing problems.

Runners in the slower finish-time categories showed less even pacing. That does not necessarily mean that they lacked discipline or experience. Lower fitness and less fatigue resistance may simply make an even race harder to produce.

Evidence in context

What does the wider evidence say?

Earlier reviews and large race-result studies suggest that relatively even pacing is generally associated with faster marathon times, whereas aggressive starts and substantial late-race slowing are linked to poorer performances. However, most evidence is observational, and course profile, weather, ability and race goals can influence the pacing pattern recorded (Sha et al. 2024). An analyses of 190,228 New York finishers and 1.7 million recreational runners also found that faster runners maintained steadier speeds and that unusually fast starts were associated with slower finishes (Santos-Lozano et al. 2014, Smyth 2018).

How this study fits: This study broadly confirms those associations. Its main contribution is methodological: detailed splits from 78,912 Boston finishers were grouped into naturally occurring pacing patterns, distinguishing variable pacing from different degrees of positive splitting. It does not show that even pacing itself caused faster times.

Practical decision

Should runners change anything?

Maybe. The study reinforces existing pacing advice rather than establishing a new pacing formula.

Consider this if

  • You regularly start faster than your planned pace.
  • Your pace deteriorates progressively after halfway.
  • You choose marathon pace from a hopeful race predictor rather than recent training.
  • Your early race effort feels more like a 10 km race than the start of a marathon.

Do not overreact if

  • Your pace changes because of hills, wind, crowding, or aid stations.
  • Your second half is modestly slower than your first.
  • Your race goal is completion rather than the fastest possible time.
  • You run trail marathons or ultras where even pace is rarely realistic.

A sensible next step

Review the pacing pattern from your most recent marathon or long race simulation. If you repeatedly start quickly and fade heavily, practise a more restrained opening during selected long runs or marathon-specific sessions. This is a practical interpretation, not an intervention tested by the study.

TIP: Never make any major changes to your training or lifestyle habits based on the findings of one study, especially if the study is small or provides low-quality evidenceA low quality of evidence means that, in general, studies in this field have several limitations. This could be due to inconsistency in effects between studies, a large range of effect sizes between studies, and/or a high risk of bias (caused by inappropriate controls, a small number of studies, small numbers of participants, poor/absent randomisation processes, missing data, inappropriate methods/statistics). When the quality of evidence is low, there is more doubt and less confidence in the overall effect of an intervention, and future studies could easily change overall conclusions. The best way to improve the quality of evidence is for scientists to conduct large, well-controlled, high-quality randomised controlled trials.. Check whether other trials confirm the findings. If there is a meta-analysisA meta-analysis quantifies the overall effect size of a treatment by compiling effect sizes from all known studies of that treatment. on the topic, look at the effect sizeA standardised measure of the magnitude of an effect of an intervention. Unlike p-values, effect sizes show the size of the effect and how meaningful it might be. Common effect size measures include standardised mean difference (SMD), Cohen’s d, Hedges’ g, eta-squared, and correlation coefficients., the variability between studies, and the quality of evidenceCertainty of evidence tells us how confident we are that the published results accurately reflect the true effect. It’s based on factors like study design, risk of bias, consistency, directness, precision, and publication bias. High certainty means that the current evidence is so strong and consistent that future studies are unlikely to change conclusions. Whereas, low certainty means more doubt and less confidence, and that future studies could easily change current conclusions..

Expert interpretation

My thoughts

Running science from Thomas Solomon at Veohtu

This study gives us an unusually detailed look at pacing across nearly 79,000 marathon finishes. Using 9 intermediate checkpoints was much more informative than simply comparing the first half with the second.

The central association is convincing. Relatively even pacing travelled with faster finish times across age groups, sexes, and performance levels. Severe fluctuations and large late-race slowdowns travelled with slower times.

The causal explanation remains less tidy.

Good runners tend to do several things well. They train consistently, assess their fitness accurately, choose realistic goals, prepare for the course, and manage race-day effort. Even pacing may help them perform well, but it may also be the visible calling card of being well prepared in the first place.

There is also some mathematical overlap between the pacing profiles and finish time. Both were calculated from the same split data. A runner who slows heavily will necessarily finish later than they would have without that slowdown. The analysis describes the pattern very well, but it cannot tell us how much time the same runner would save if instructed to pace differently.

The Boston course adds another wrinkle. A runner can hold an excellent, steady effort without producing perfectly steady splits. The hills do not care what your spreadsheet predicted.

Still, the advice is practical, low-risk, and consistent with previous marathon research: do not try to bank large amounts of time early. The marathon charges interest, and the repayment terms are dreadful.

The most useful next step would be an intervention study that assigns runners to realistic pacing plans while measuring fitness, training history, race experience, fuelling, weather, and course conditions. That would help show whether changing pacing itself improves performance.

My Rating of Perceived scientific Enjoyment

owlRPsE: 7/10

I experienced moderate scientific enjoyment because the enormous real-world dataset, official split times, transparent methods, and useful sensitivity analysesA sensitivity analysis tests whether a study’s results stay the same when the researchers change some of their methods, assumptions, or data inputs. E.g., researchers might reanalyse their data without studies that had a high risk of bias. Sensitivity analyses help show how robust the findings are and whether small changes could lead to different conclusions. produced a detailed picture of marathon pacing. The observational design and limited control of fitness and race-day factors kept the result interesting rather than decisive.

down arrow

Read on for further details about the methods, results, strengths, limitations, and conflicts of interest.

Research question

What did the researchers ask?

The authors aimed to identify common pacing patterns among Boston Marathon finishers.

They also examined whether those patterns were associated with finish time, performance level, age, sex, and severe late-race slowing.

The authors expected relatively even pacing to be associated with faster finishes. They also expected strong positive splitting and hitting the wall to be more common among slower runners.

Study design

What type of study was this?

This study was a retrospective cross-sectional observational studyAn observational study is where researchers observe what naturally occurs without intervening — no treatment is assigned. I.e., the researchers watch and learn, but don’t interfere. Observational studies are used in epidemiology and can have different study designs, including cross-sectional, case-control, and cohort study designs..

The researchers analysed race results that had already been collected. They did not assign runners to different pacing strategies.

This design can describe real-world pacing patterns and show associations. It cannot prove that even pacing caused faster performance.

The split times had a sequence within each race, but the researchers did not follow individual runners over several races to see whether changing their pacing changed their performance.

Participants

Who took part?

The dataset initially contained 79,638 completed Boston Marathon records from 2015, 2016, and 2017. After excluding records with missing, inconsistent, or implausible split times, the researchers retained 78,912 finishing records.

The records included 43,041 male runners and 35,871 female runners. The runners were aged 18 to 84 years, with a meanThe average of a set of numbers, calculated by summing all the values and dividing by the total number of values. age of 42.4 years.

The medianThe middle value in a set of ordered numbers; if there is an even number of values, it is the average of the two middle numbers. finish time was about 3 hours 46 minutes. The Boston Marathon qualifying standards created a more performance-selected sample than many mass-participation marathons.

The runners’ training history and previous marathon experience were not reported. The paper also did not report whether some runners appeared in more than 1 race year.

Methods

What did the researchers do?

Who? 78,912 Boston Marathon finishing records
What? Split-time and pacing-pattern analysis
How long? 3 race editions from 2015 to 2017

The researchers used publicly available timing records from the Boston Marathon.

The dataset contained cumulative times at 5 km, 10 km, 15 km, 20 km, halfway, 25 km, 30 km, 35 km, 40 km, and the finish.

The researchers converted these records into pace for each race segment. They then expressed each segment relative to the runner’s own average pace. This removed much of the effect of absolute running speed and allowed the researchers to compare the shape of each pacing profile.

A statistical clustering method identified 4 common patterns:

  • Even pacing
  • Mild positive splitting
  • Strong positive splitting
  • Variable pacing

Positive splitting means that a runner completed the first half faster than the second.

The researchers compared finish times across the 4 patterns. They also examined whether pacing patterns differed according to age, sex, and performance level.

The authors defined hitting the wall as completing the final 7.195 km at a pace more than 20% slower than the runner’s average pace over the first 35 km.

This was a split-time definition. The researchers did not measure glycogen depletion, symptoms, perceived effort, or whether each runner personally felt that they had hit the wall.

Main findings

What did the study find?

Even pacing Fastest average finish
Late-race slowing Least common with even pacing
Fast opening More wall cases

The researchers identified 4 pacing patterns

The even-pacing group contained 37,827 records, representing 47.9% of the sample.

The mild positive-split group contained 25,135 records, or 31.9%. The strong positive-split group contained 9,600 records, or 12.2%. The variable-pacing group contained 6,350 records, or 8.0%.

The pacing groups were not completely interchangeable with traditional first-half versus second-half classifications. When the researchers used only the halfway ratio, 24.0% of the runners had even pacing, 75.6% had positive splitting, and 0.4% had negative splitting.

This difference shows that the method used to define pacing can materially change the result. Segment-by-segment data picked up patterns that a single halfway comparison could miss.

Even pacing was associated with faster finishes

The even-pacing group had the fastest average finish time at about 3 hours 37 minutes.

The mild positive-split group averaged about 3 hours 54 minutes. The variable-pacing group averaged about 4 hours 17 minutes. The strong positive-split group averaged about 4 hours 36 minutes.

The even-pacing group finished an average of 59.5 minutes faster than the strong positive-split group. The statistical difference was large.

That does not mean an individual runner would automatically save about 60 minutes by changing their pacing. The groups probably differed in fitness, preparation, experience, goals, and other unmeasured characteristics.

Faster runners paced more evenly

Even pacing appeared in 82 out of every 100 records from runners finishing in under 3 hours.

It appeared in about 65 out of every 100 records from runners finishing between 3 hours and 3 hours 30 minutes. The frequency fell to about 44 out of every 100 among runners finishing between 3 hours 30 minutes and 4 hours 30 minutes.

Only about 20 out of every 100 records from runners taking longer than 4 hours 30 minutes showed even pacing.

The same broad pattern remained when the researchers ranked male and female runners separately. This reduces the chance that the findings arose only because the fixed performance groups contained different proportions of men and women.

Severe late-race slowing varied greatly between patterns

Overall, 11,973 records met the study’s definition of hitting the wall. That was about 15 out of every 100 records.

Only 65 of the 37,827 even-pacing records met the definition. This was about 0.2 out of every 100.

About 21 out of every 100 mild positive-split records met the definition. The figure was about 12 out of every 100 in the strong positive-split group and 85 out of every 100 in the variable-pacing group.

The lower incidenceThe number of new cases of a condition or disease that develop in a specific population during a defined time period. in the strong positive-split group than in the mild positive-split group looks counterintuitive. The paper did not establish why this occurred. The wall definition focused on the final part of the race, whereas the pacing clusters used the full race profile.

Among the records meeting the wall definition, final-segment pace was about 33% slower than the earlier average. This equated to an average slowdown of 1 minute 48 seconds per kilometre.

The pattern remained when the researchers used less and more strict definitions of hitting the wall. This strengthens confidence that the main finding did not depend entirely on the chosen 20% threshold.

A fast opening was associated with more cases of hitting the wall

The researchers also examined the opening 10 km.

Among runners who completed this section more than 5% faster than their eventual average pace, about 27 out of every 100 met the wall definition. Only about 2 out of every 100 runners with an opening pace close to their eventual average met it.

This was still an observational comparison. The runner’s eventual average pace is also only known after the race, so the 5% threshold should not be treated as a ready-made pacing rule.

The authors concluded that even pacing was the pattern most consistently associated with favourable marathon performance. They also concluded that pacing advice may need to consider the runner’s age, sex, and performance level.

What helps my confidence in the findings?

The strengths

  • The study included 78,912 complete finishing records, making it much larger than most marathon pacing studies.
  • The researchers used official chip-timing data from 9 intermediate checkpoints.
  • The segment-by-segment analysis provided more detail than a simple first-half versus second-half comparison.
  • The researchers used a data-driven method to identify pacing patterns.
  • Sensitivity analyses supported the main patterns across alternative performance groups, race-year adjustment, clustering methods, and wall definitions.

What limits my confidence in the findings?

The limitations

  • The observational design cannot show that even pacing caused faster finishes. Fitness, training quality, preparation, and experience may explain part of the association.
  • The Boston Marathon qualifying standards and distinctive point-to-point course limit how confidently the findings can be applied to other marathons.
  • The dataset lacked training history, previous marathon experience, race-day weather, and knowledge of the course. It also did not include many other potentially important factors, such as fuelling, injuries, race goals, congestion, or tactical decisions.
  • The analysis included finishers with complete and plausible split times only. It did not examine runners who withdrew or failed to complete the race.
  • The study’s definition of hitting the wall did not measure glycogen depletion or the runners’ subjective experiences. Some slowdowns may have occurred for tactical or mechanical reasons.

Funding and conflicts

Who funded the study?

The authors reported receiving no specific funding for the research.

The authors declared that they had no competing interests. Their listed affiliations were academic institutions, and they did not report any relevant commercial relationships.

Learn to train smart, run fast, and be strong.
© 2026 Thomas Solomon. All rights reserved.
Icons from Icons8.
Follow @veohtu
Join the club on