What is a negative split, and does it actually help?

Updated August 26, 2026 · By the Runked team

Three pacing strategies, and what the evidence shows

"Negative split," "positive split" and "even split" describe the shape of a runner's pace across a race, separate from the finishing time itself. Wikipedia's negative split entry defines a negative split as "a racing strategy that involves completing the second half of a race faster than the first half," achieved through a deliberately slower start followed by a gradual or sudden increase in speed. The same entry defines an even split as one where "the runner must run the same split for every lap (or other distance considered a split) to hit the time." The table below lines up each strategy against what named studies of real marathon data actually found, at three different levels of the sport.

What large recreational marathon fields show

Robert Deaner and colleagues gathered results from 14 U.S. marathons run in 2011, comparing the second-half pace with the first half across 91,929 performances. Published in Medicine & Science in Sports & Exercise, the 2015 study found men's pace slowed by an average of 15.6% and women's by 11.7%, a statistically significant difference (p<0.0001). Across that large field, slowing in the second half, a positive split, was the pattern both sexes showed on average, women less severely than men. A 2025 reanalysis of the same dataset by Matthew Tenan and David Borg, published in the Journal of Sports Sciences, found the sex-linked gap in slowing depends on a runner's age and finish time rather than holding uniformly across the whole field, so the 15.6%/11.7% figures describe a field-wide average that any single runner's own result can land well away from.

What championship marathons show

Brian Hanley collated finishing and split times for 673 men and 549 women across nine Olympic and IAAF World Championship marathons, published in the Journal of Sports Sciences in 2016. The study found medallists of both sexes ran even-paced from the 10km mark onward, while runners who fell off the lead pack around halfway recorded the largest slowdowns. Overall, men still covered the second half in 109.4% of their first-half time and women in 106.4%, both positive on average, but women slowed less, a statistically significant difference, and were far more likely to run a negative split: 12.9% of all women did, against 4.7% of all men. Fewer women dropped out of the race entirely. Hanley notes that these championship fields differ structurally from the big-city marathons Deaner studied: mass-field races typically use pre-arranged pacemakers to help runners hold pace, an aid unavailable to competitors racing for medals rather than time.

What marathon world records show

Jose Joaquin Diaz, Eduardo Jose Fernandez-Ozcorta and Jordan Santos-Concejero analyzed every men's marathon world record set between 1967 and 2014, 15 records in total, splitting each race into eight 5km segments plus a final 2.2km segment. Published in the European Journal of Sport Science in 2018, the study found that, averaged across all 15 records, the first half was run slightly faster than the second, a small but measurable positive split. Splitting the group told a sharper story: record holders from 1967 to 1988 started significantly faster than those from 1988 onward, then their pace dropped sharply after 25km. Record holders from 1988 onward ran the opposite shape, negative overall, with their fastest 5km segments falling between 30km and 40km. Diaz and colleagues' own conclusion did not simply endorse the negative shape: they wrote that a pacing strategy with very little speed change across the whole race "may be the way to go in the future," even though they describe a negative pace distribution as the option earlier research proposed as most efficient.

Two recent individual world records, reported directly by World Athletics, illustrate both outcomes. Eliud Kipchoge's 2022 Berlin world record (2:01:09) split 59:50 for the first half and 61:19 for the second, a positive split; World Athletics notes that halfway time was identical to his half-way split in the unofficial sub-two-hour run in Vienna three years earlier. Kelvin Kiptum's 2023 Chicago world record (2:00:35) split 1:00:48 and 59:47, a negative split. At the recreational end, negative splits stay uncommon even among fast finishers: Munoz-Perez and colleagues' 2020 analysis of 2,295 finishers of the 2017 Berlin Marathon who finished in 3:00 or faster (men) or 3:30 or faster (women), published in PLOS ONE, classified over 90% of that field as even-paced by their statistical method, with negative pacing in just 0.12% of men and 0.16% of women.

How to apply it

RunnersConnect coach Jeff Gaudette recommends a mild version of a negative split for marathon goal-pace running, described as the 10-10-10 method: run the first 10 miles 15 to 20 seconds per mile slower than goal pace, hold goal pace for miles 11 through 20, then hold pace or pick up slightly for the final 10K. Gaudette's stated reasoning is that starting faster than goal pace risks the kind of late-race fade the studies above describe.

Runked's own worked example below applies Gaudette's ranges with real arithmetic, for a runner targeting a flat 4:00:00 marathon at 9:09 per mile.

A flat 9:09 per mile for all 26.2 miles finishes just under 4:00:00 (3:59:44). This version, using the top of Gaudette's 15-to-20-second cushion and a modest final push, finishes 2 minutes 24 seconds slower: the 200 seconds spent easing through miles 1 to 10 is only partly repaid by the 56 seconds gained back over the shorter final stretch. Closing that gap on race day means trimming the opening cushion, finding more speed in the closing miles than this example uses, or both, the same arithmetic that explains why Diaz and colleagues found so few real negative splits among the world records they studied.

Common mistakes with negative splits

• Treating the opening cushion as minutes instead of seconds. Gaudette's method calls for a 15-to-20-second-per-mile cushion; a much bigger opening cushion needs a much bigger closing kick to break even.
• Assuming any negative split beats any positive one. Diaz and colleagues' 50-year analysis found the historical average across all 15 world records studied was a small positive split, and their own conclusion favored steady, low-variation pacing over a negative shape for future record attempts.
• Copying an elite runner's splits directly. Kiptum's 60:48/59:47 and Kipchoge's 59:50/61:19 both came from paced, monitored record-attempt courses. Hanley's championship data, run without pacemakers, show a different pattern, and Deaner's mass-field data show a different one again.
• Ignoring effort in favor of the split alone. Hanley found the runners who slowed most were the ones who dropped off their pack around halfway; a split target without an effort check, RPE, heart rate or the talk test, can't flag that happening until it already has.

What to do next

Runked's race time predictor sets a realistic goal pace before any splits plan is worth building, since a pacing strategy only makes sense against an achievable target. What counts as a good marathon time for a given age and sex gives that target useful context. The marathon taper guide covers the weeks immediately before race day, when the pacing plan gets finalized. Compare your own splits to your own past races, not to a world record or someone else's plan; in Runked the rank starts everyone at Bronze and only moves on your own runs. See how the running rank works.

Frequently asked questions