Santa Fe Institute’s postdocs have completed the second annual 72 Hours of Science, a 3-day-long marathon of research, data analysis, and writing for publication. Riffing on the structure of the 48-hour Film Festival, 72 Hours of Science — 72h(S) for short — explores the limits of what novel science a group of diverse researchers can produce in a short, focused timeframe. In a rented house tucked in the hills outside of Santa Fe, this year’s group decided to explore the data-rich arena of world records.

Artemy Kolchinsky, a postdoctoral fellow, pitched the winning idea, which the group selected from 13 wildly different topics. While he knew he'd come with a strong idea in mind, but was still surprised when it became the overall favorite, he says. "It was fascinating to watch the idea acquire a momentum and trajectory of its own."

One of the first major changes in trajectory came as the group reviewed the existing literature. “We quickly realized that there was more that had been done on this topic than we’d known about going in,” says Eleanor Power, an Omidyar Fellow. “It was a good sign that we were on the right track, but we had to pivot on a tight timeframe — we digested what had been done and developed a new tack and set of questions to pursue.”

On the final morning, with two hours left before their 1 p.m. deadline, the house was quiet save for a fervent clicking of keyboards. Everyone had their laptops open, simultaneously editing the same Google doc or rendering the final set of figures. With seconds to spare, they posted the resulting 22-page paper to the arXiv preprint server. 

To complete this project, the team had divvied up the work into smaller groups. Some gathered records from sports and games, biological evolution, and technological development. Others took a deep dive into the data of marathon races. 

“If you think about records as extreme behavior, something pushing the boundary for a process, that allows you to think about records across lots of different domains. The question then is, for all of these different records, how do you start to put those into meaningful categories? The hope is that those records can tell you something about ultimate limits,” says SFI Omidyar Fellow Chris Kempes.

SFI Omidyar Fellow Dan Larremore was part of the group analyzing marathon data. “If you want to understand how records are set, you need to know about the mechanisms that are generating exceptional individuals,” he says. This deep dive into marathon records revealed that record-setting data alone can obscure the bigger picture. For instance, a line highlighting the times achieved by marathon winners shows incremental, periodic improvement over nearly 50 years, with new records set only nine times. Meanwhile, a snapshot of the top 100 times in those same races paints a different picture, with periods of rapid, consistent improvement and a narrowing of the time distribution among the top finishers overall. Looking at the broader dataset of people who came close to winning provides a better picture of the record-setting process. 

“I wasn’t sure if we’d get results and a conclusion — a storyline — but now, I’m actually proud of it,” says Keyan Ghazi-Zahedi, SFI’s newest postdoc, who arrived at SFI just a week before 72h(S). “It was on the one hand very professional, but I also had fun. That’s something really hard to balance.”

SFI President David Krakauer congratulated the team after the paper was published. “It’s an inquiry into the limits and possibilities of human performance through an actualized experiment on the limits of human concentration!”

Read the paper, "On the Records," on the arXiv.