Image generated by an AI large model
How fast can humans ultimately run 100 meters? Since Usain Bolt set the world record of 9.58 seconds, this question has continued to attract widespread attention. On the 100-meter track, every 0.01-second improvement represents a new breakthrough, making the “limits of human speed” a classic topic in sports science.
Professor Yi He from the College of Science at Eastern Institute of Technology, Ningbo, in collaboration with John H. J. Einmahl, a leading international expert in extreme value statistics and professor at Tilburg University in the Netherlands, re-estimated the ultimate limits of human 100-meter performance based on a more rigorous theory of extreme value statistics. The study shows that, in a statistical sense, the “speed ceiling” for human 100-meter sprinting is approximately 9.49 seconds for men and 10.20 seconds for women. Recently, the research findings were published in Extremes, an authoritative international journal in the field of extreme value theory, under the title “Accurate Estimates of Ultimate 100-Meter Records.”
Why were previous estimates inaccurate? The method has now been “upgraded.”
Image generated by an AI large model
The 100-meter sprint is a typical “extreme value problem.” Researchers are not concerned with the average performance of ordinary people, but with the ultimate limit of the world’s top sprinters.
In the past, traditional research methods faced two major challenges. First, the available data were sparse, as such studies usually relied on limited historical world records or a small number of elite performances, resulting in relatively small sample sizes. Second, the assumptions were often too idealized, typically presuming that all top athletes shared the same “talent ceiling.” In reality, however, individual differences, or heterogeneity, do exist, and each athlete may have a different “ceiling.” These factors led to considerable uncertainty in previous predictions.
To address these issues, Professor Yi He’s team incorporated “individual differences” into the formula and used heterogeneous extreme value statistics to estimate the limits of 100-meter sprinting in a way that more closely reflects real-world competition.
Defining the Boundary of 100-Meter Limits Based on Large Samples
To improve the reliability of the estimates, the research team compiled a large volume of 100-meter performance data from 1991 to 2023, covering top sprinters worldwide.
Specifically, the men’s dataset included 25,244 performance records from 5,618 athletes, while the women’s dataset included 11,654 performance records from 2,528 athletes. Rather than using only each athlete’s single best performance, the study incorporated multiple annual best performances from the same athlete, with up to five annual best performances included per athlete. Only by including multiple performances from each athlete can researchers statistically characterize individual heterogeneity and obtain accurate estimates of statistical error.
Based on these data, the study derived the 95% confidence lower bound for the ultimate limits of human 100-meter performance.
The ultimate limit for the men’s 100-meter sprint is 9.49 seconds.
This means that scientists are 95% statistically confident that it is almost impossible for humans to run faster than 9.49 seconds. This limit is only 0.09 seconds faster than Usain Bolt’s world record of 9.58 seconds.
The ultimate limit for the women’s 100-meter sprint is 10.20 seconds.
Similarly, from a statistical perspective, it would be extremely difficult for women’s 100-meter performance to break the 10.20-second mark. Compared with the world record of 10.49 seconds held by Florence Griffith-Joyner, this suggests a theoretical improvement margin of 0.29 seconds.
Compared with previous studies, another key feature of this research is its significantly improved precision. By using multiple performance records from the same athlete, the study more accurately characterizes the impact of athlete heterogeneity on statistical estimation, reducing the asymptotic variance of the statistical estimates by approximately 35%. This means that the new results are not only more precise, but also closer to the complex realities of real-world competition.
A New Statistical Tool for Studying the Limits of Athletic Performance
This study does not claim that world records will never be broken. Rather, it provides a statistically meaningful boundary for the question of “how much faster humans can still run.” It transforms public discussions about speed, talent, and limits into questions that can be tested through data and scientific theory, offering a new perspective for research in sports science and related fields.
Eastern Institute of Technology, Ningbo is the first affiliation of the paper. Professor John H. J. Einmahl of Tilburg University in the Netherlands and Professor Yi He of Eastern Institute of Technology, Ningbo are co-first authors of the paper. Professor Yi He is the sole corresponding author.
Paper information:
https://doi.org/10.1007/s10687-026-00537-8
About Professor Yi He’s Research Group
Professor Yi He’s research group mainly focuses on extreme value theory, high-dimensional statistics, and financial econometrics.
In the field of extreme value theory, Professor He is committed to developing extreme value analysis frameworks for heterogeneous big data, proposing new statistical laws and theoretical systems, and advancing the efficient prediction of natural disasters and extreme climate events.
In high-dimensional statistics, he applies random matrix theory to study non-sparse models. Through rigorous theoretical analysis, he develops optimal methods for statistical inference and prediction, with the aim of improving the predictive power for complex economic systems.
In financial econometrics, he focuses on building data-driven statistical inference frameworks and studies the dynamic features and structural changes of financial time series, providing methodological support for the accurate identification and real-time monitoring of financial risks.
Contact: yihe@eitech.edu.cn
