- Accepted Paper
Stochastic dynamics and first-passage statistics in marathon pacing
Phys. Rev. E - Accepted 23 October, 2026
DOI: https://doi.org/10.1103/1wch-rv2h
Phys. Rev. E - Accepted 23 October, 2026
DOI: https://doi.org/10.1103/1wch-rv2h
Marathon pacing often undergoes a pronounced deterioration during the late stages of a race, a phenomenon known as "hitting the wall." We investigate this deterioration pattern as a distance-dependent stochastic process using split-time records from more than half a million runners across four World Marathon Majors and a finite-resolution Kramers–Moyal framework. The inferred drift reveals effective states at progressively slower pacing levels as the race unfolds, with marked sensitivity to population composition. Persistent pacing deterioration is naturally characterized as a first-passage process whose hazard varies strongly and nonmonotonically with race distance. A coarse-grained Langevin reconstruction reproduces both the aggregate late-race pacing evolution and the distribution of first-passage distances, while model ablations show that stochastic fluctuations and heterogeneity of pacing states both contribute to the observed dynamics. Transfer to an independent marathon population retains the main aggregate features. These results provide an effective stochastic description of late-race pacing deterioration in a heterogeneous, evolving population.
If the author has provided any supplemental materials with this article they will be available upon publication of the version of record.