How recreational marathon runners hit the wall: A large-scale data analysis of late-race pacing collapse in the marathon
Using pacing data from more than 4 million marathon race records, the study defined hitting the wall as a sustained period of significant slowing late in the race. 28% of male and 17% of female runners hit the wall (OR = 1.43). Wall-hitting was more common in the 3 years around a recent personal best (36% in the 3 years before a recent PB vs 23% in earlier years).
When they hit the wall, men slowed a little more than women and over a slightly longer distance, but these effect sizes were small. The minutes lost to the wall grew with runner ability.
Participants
Recreational marathoners represented by more than 4 million race records; compared by sex, age and ability. The abstract does not give the number of individual runners or the races included.
Protocol
Pacing-based definition of hitting the wall: a sustained period of slowing during the latter stages of the marathon.
The slowdown was expressed relative to earlier race pace (5km-20km), and its cost relative to an estimate of each runner's recent personal-best time.
Comparisons by sex, age and ability.
Key Results
Prevalence by sex: 28% of male runners hit the wall vs 17% of female runners (p < 0.01, OR = 1.43).
Around a personal best: 36% of all runners hit the wall in the 3 years before a recent PB vs 23% in earlier years (p < 0.01, OR = 1.31). Wall-hitting was more frequent in the 3 years immediately before and after a recent PB.
Size of the slowdown: relative slowdown 0.40 (men) vs 0.37 (women) compared with 5km-20km pace (p < 0.01, d = 0.15).
Length of the slowdown: 10.7km (men) vs 9.6km (women) (p < 0.01, d = 0.17). The author notes that the effect sizes of these sex differences are small.
Cost: the finish-time cost of hitting the wall (lost minutes) increased with ability (r2 = 0.91 for men, 0.81 for women, p < 0.01).
The abstract gives the typical onset as after the 20-mile mark (from its definition of the phenomenon), not as a measured result.
Practical Application
In more than 4 million recreational marathon records, 28% of men and 17% of women hit the wall (a sustained late-race slowdown), and when it happened they slowed for 9.6-10.7km, a relative slowdown of 0.37-0.40 against their own 5km-20km pace (observational, pacing-defined). Hitting the wall is common, not a rare disaster: about a quarter of men and a sixth of women in this dataset; plan the first half and the fuelling to avoid being one of them.
Runners in the 3 years around a recent personal best hit the wall more often (36% vs 23%). Inference, not tested here: a runner chasing or defending a PB is more likely to set an ambitious early pace, which is the obvious suspect; the paper did not measure intent.
Men were more likely to hit the wall than women (OR = 1.43), but once it happened the sex differences in how much and how far they slowed were small (d = 0.15 and 0.17).
The finish-time cost of the wall grew with ability, so a faster runner is not protected from paying for it.
Not from this paper; the base's own practice: start at or slightly below goal pace, take carbohydrate from early in the race, and arrive with muscle glycogen topped up; the abstract names energy-store depletion as the usual cause but the study itself did not measure fuelling or glycogen.
Limitations
Observational, from race records only: no data on fuelling, heat, course, injury or intent, so the cause of each slowdown is unknown.
Hitting the wall is defined by a pacing threshold, not by physiology; a runner who slowed for another reason counts the same.
Personal-best times are estimated, not recorded race-by-race for every runner.
The sex differences are statistically significant because the sample is huge; the author notes their effect sizes are small.
Written from the abstract and record metadata only (full text not read); the exact slowdown threshold and races included are not given here.
How this entry was checked and changed
Why this level
large observational dataset; hitting the wall is defined from pacing, not measured physiologically