The New York City Marathon Is an Engineering Marvel
Marathons don’t happen by themselves; they require months of planning and an expertise in engineering and crowd science. That’s doubly true for the world’s largest race.
Outside's long reads email newsletter features our strongest writing, most ambitious reporting, and award-winning storytelling about the outdoors. Sign up today.
The first thing Ted Metellus did when the 2021 New York City Marathon wrapped up, apart from getting a decent night’s sleep, was to start planning for this year’s race. “It’s on my mind all the time,” says Metellus, the race director and vice president of the New York Road Runners (NYRR), which puts on the race. This was last February. Stretching across five boroughs in the country’s largest, densest city, with as many athletes on course that would normally fill the stands at a sporting event—though it also has its many thousands of spectators—the marathon is a complex negotiation of time, space, and people. “It is one of the single largest mobilizations of resources in the city,” says Metellus.
By sheer numbers, it is impressive. Last year, even with a reduced pandemic field of more than 21,000 runners, some 41,240 gallons of Poland Spring water—and 1.4 million paper cups—were dispensed to runners on course (in addition to the 45,000-plus bottles given out at the start); along with 30,000 Honey Stinger gels. Some 122,760 pounds of clothing was shed by the runners at the start, then collected and given to Goodwill. There’s a medical station every mile, even therapy dogs and psychologists at the start. “Sometimes people just need a moment to kind of settle themselves in when you’re getting ready for an event of this scope,” says Metellus.
This year, the marathon is back at full strength, with more than 50,000 participants expected to throng the streets of New York this Sunday. The job of understanding how that small city’s worth of people will travel the 26.2 miles from Staten Island to Central Park—and to ensure it happens with as little friction as possible—falls largely to Marcel Altenburg, a senior lecturer in crowd science at the Manchester Metropolitan University. Born in Germany, and a former Captain in the Bundeswehr, Altenburg went to Manchester to pursue a degree in crowd science, a discipline, he notes, that got its real start in 1989 with the infamous Hillsborough Stadium disaster, when 97 people were killed in a crush caused by overcrowding. He stayed on and became a lecturer.
Since then, he’s been involved with numerous high-profile events, from presidential inaugurations to rock concerts to football championships to, most recently, managing the massive crowds that queued to pay respect to Queen Elizabeth. And, of course, any number of marathons, from Berlin to Chicago. In 2016, he began working with the NYRR on the world’s largest race. On Sunday, at 8 A.M., when the first wave of athletes—the professional handcyclists—set off on the course, Altenburg will be at the start village, looking to see if his exquisitely calculated script plays out as calculated.
The starting process is itself massive: It will easily take longer to dispense the five waves of runners, in 15 “corrals,” across three starting points, than it will take the best runners to complete the race (it takes 18 minutes, less than a pro can run a 5K, just to dispense each group). And the start, from a planning perspective, is everything. “It is the last moment we can influence the race,” he says. “It’s the last time someone listens to you—the last time we can tell them, stay right, wait for a second. From then on, the race is on them.”
It’s a bit like a water tap. You can control the source, but once the water is flowing, you cannot easily call it back. When he started working with the Road Runners, he had a revelation. “We were convinced that the way we start impacts everything on the course,” Altenburg says. “That everything on the course is of our own making.”
Once you had accurately modeled the start, you could predict, with unprecedented accuracy, everything that happened afterwards. After backwards engineering previous years’ data, Altenburg advised that changing to 15 corrals, from 12, would allow better control. He told Metellus: “If you give us the start, we can predict the finish, and the whole 26.2 miles in between.”
Breaking the race up into five minute windows, Altenburg projected that the largest finishing wave would consist of 1,366 runners. There were 1,367. “I know who the guy was,” Altenburg says, laughing. “He was from Mexico.” But his overall estimate was 99.93 percent accurate. The code had been cracked, his “Start Right” predictive algorithm born. Now, any contingency that might arrive—even a global pandemic that suddenly required six-foot social distancing—could be modeled.
The rolling-start, as opposed to the “open start,” is now fairly de facto at most major international marathons, but some races, Altenburg notes, “are in love with this big crowd picture at the start of the race—a guy shoots a gun, and everyone start at the same time.” But that’s no longer possible in the largest events. “The races are bigger,” he says, “and the cities are definitely not getting bigger.”
With chip timing, he adds, “they don’t need to be on the start line—you get to start two hours later and still get your finishing time.” Key to this, he says, is making the departure point narrower than anything they’ll experience on course—and keeping the “water tap” open only to 70 percent. It’s a bit like “ramp metering,” those traffic lights that tell you when you can enter the highway. In essence, you go slower to go faster.
What differentiates a marathon from other crowd-management scenarios is its dynamic nature. While it is, essentially, a rolling queue, it’s a queue, says Altenburg, “in which everyone is constantly changing the order of everything.”
Compared to even a large event like the Queen’s funeral, which saw upwards of 250,000 people, “a marathon is, to be honest, 50,000 times more complicated.” With something like a soccer match, the crowds may be massive, but the behavior is generally constrained. “I need to get them in, that’s a big task. I need them to sit down, then they go to the loo, then they go home.” These are all big steps, he says. “But in a marathon, they never sit once.” They are arriving “by all means of transport,” then circulating around the start village, then they get on the road, then they’re finishing, grabbing their poncho, and trying to find their family or friends. “Fifty-five-thousand people are making their way in shorts, and everybody’s got their own story, everybody with their own pace.”
Marathons, in effect, cannot be understood as a system. Armed with huge amounts of computing power, data from previous races, and a hope that people more or less run at the pace they have said they are going to run, Altenburg needs to calculate every single runner. “The ideal experience is that I see the same 100 people throughout most of my race,” he says. “The organizer is going to great lengths to minimize the number of overtakes on the course.”
Being constantly overtaken, or by contrast constantly having to “zig-zag” past groups of other runners, is not only stressful, he says, but can be unsafe (the algorithms provide for an ideal of three square meters for each runner, a number that was briefly increased during the era of social distancing). The professional field, says Altenburg, will “immediately stretch,” while runners further back may spend more time together. But people are not data points, they will do the unexpected. They are chaos. I speak here from experience. When I, eager and undertrained, participated in the event in 2017, I ran a fairly brisk half-marathon, passing many runners—which was often not easy on narrow Brooklyn streets—before slowing in the second half, and essentially blowing up at the end. While my finish was statistically average, I was, at a more micro level, often an outlier.
And then there is the city itself. “It’s the same race every year, the name is the same, but New York is a living organism,” Altenburg says. Roadways are altered, massive construction sites arise, new bike lanes are built; all things that might not affect the individual runner, Altenburg says, but could have system-level implications. During the pandemic, on-street dining emerged, and many structures have remained, further constraining the streetscape (for some, NYRR asks for temporary closures). Every five years, the course is painstakingly measured.
Working with the city’s Department of Transportation (DOT), the NYRR conducts a number of course inspections in the months leading up to the event, flagging potential obstacles. “We do not allow steel plates on the roadway,” the DOT’s Jessica Colaizzi told me. “We stand very firm on that and we work closely with contractors to make sure the plates disappear.”
Then there are things that are outside of anyone’s control, but must still be factored in, like weather. This year’s event is promising higher-than-normal temperatures. “When the temperature goes up by five degrees Fahrenheit, we run a different simulation,” Altenburg says. Medical resources can be shifted to potential trouble spots (temperatures above 70 degrees, as Sunday is promising, are associated with an elevated risk of heat stroke).
For the professional field, Altenburg says, this will hardly matter—they’ll be finished by the time warmer temperatures set in. But for everyone else, this could have an impact. And not only, Altenburg says, for the slower, later-starting runners. “You might not be a professional, but someone who knows what they’re doing, and wants to break the three-hour barrier. That’s exactly when it’s hot, when you’re going to your limits.”
While each runner runs their own race, Altenburg has observed some aggregate trends about the New York City Marathon over the years. “People always speed up when they hit Manhattan,” he says, “even though the advice is don’t—you still have eight miles to go!”
But another trend is that as the race gets larger, it is actually getting slower. “They are attracting a lot of people who see it as a bucket list race,” he says. “It’s slower because it’s more inclusive. It’s an amazing race, and you want to do it at least once.” Altenburg himself has run it, in 2015. But in his head, and on his computers, he’s always running it. “As a scientist, it’s bananas. I absolutely love it.”