The New York City Marathon, 2017

An Analysis of Participation & Performance at the 2017 New York City Marathon.

barrysmyth
Nov 15, 2017 · 11 min read

TLDR;

  • We examine the race records of more than 50,000 runners from the 2017 New York City Marathon.
  • Who were the participants from this year’s race? Where were they from? How old were they? Was ot their first marathon or had they run before?
  • We also examine how gender, age, experience, and ability influence performance (finish-times, pacing, hitting the wall) at this year’s 40th anniversary event.

On Your Marks, Get Set, …

This year’s New York City marathon got underway on the morning of November 5th. A little over two hours later Geoffrey Kamworor of Kenya won the men’s race in 2:10:53, beating countryman Wilson Kipsang by 3 seconds for his first major victor. Meanwhile, Shalane Flagan took the women’s title in 2:26:53, beating Mary Keitany, who was coming off three straight NYC wins, and marking the first time a US woman has won this event since 1977.

How did the other 50,000+ runners get on? What follows is my usual analysis of the data generated by this year’s runners as they navigated through NYC’s five famous boroughs; similar reports exist for this year’s Boston, Berlin, Chicago, Dublin, and London marathons.

The race data available from New York includes 5km split-times, as well as gender and age information. In addition, separate data that we have collected over the years (>2.5m race-records from more than 100 marathons) makes it feasible to match runners across multiple races to identify whether a particular runner is a new marathoner (first-timer) or a regular racer (repeater). This means that we can consider how gender, age, ability, and experience come to influence how race day unfolded for NYC’s runners.

It’s the Taking Part…

This year, more than 50,000 runners crossed the line, with just over 58% male, and with 137 different countries represented. Italy, France, and Great Britain provided 1,500 or more particpants each, to the field on the day; the chart below shows the number of men and women for the top-20 countries (by participation levels) excluding the United States itself.

In the graph below we see the age distribution for men and women. Notice how there is a greater proportion of young females (20–35 years old) than similarly aged males. For example, approximately 40% of females are aged 35 or younger, compared to just 28% of males.

In the 30–35 years-old range there is a noticeable decline in the percentage of female runners for a period of five years or so, and beyond the 40 years-old mark the percentage of males consistently exceeds that of females. This decline in female runners in their early 30s is not unique to NYC, and it’s a pettern that repeats for large marathons around the world. It may be due to lifestyle factors that hinder female runners from participation during their 30’s, for example.

First-Timers vs. Old-Timers

We can also divide runners up in to those who appear to be completing their first marathon (first-timers) and those that have completed multiple marathons (repeaters). Strictly speaking, our ability to identify first-timers and repeaters is limited to the runners in our larger dataset of >2.5m marathon race-records. If we have no previous record for a runner then they are considered to be a first-timer. Otherwise, they are considered to be a repeater. These classifications are not perfect, because even this larger data-set is obviously incomplete and many runners will have raced in marathons that are absent. In other words. we will tend to over-estimate the number of first-timers and under-estimate repeaters, but hopefully not by too much, and in any event it should be good enough for the trend analysis that follows.

In NYC this year, a majority of young runners were first-time marathoners. As might be expected this percentage decreases steadily with increasing age, but it is encouraging to still see plenty of (apparent) first-timers starting later in life too. Females in NYC are slighty more likely to be first-timers than males from their 30s onwards.

Try, Try, Try Again

For repeaters, the number of marathons completed increases steadily with age, as we might expect. A typical 30 year-old in NYC this year was running their 3rd marathon, while a 50 year-old runner was probably on their 4th marathon, on average; the gap between men and women is in this regard is less in NYC than some other marathons, such as Chicago.

Finishing Strong

Now that we have a sense of who the runners are, and their experience levels, let’s take a look at their finish-times and, later, other aspects of their performance. The chart below shows the number of male (blue, above the line) and female (red, below the line) runners who finished at various finish-times; the height of each bar corresponds to the number of runners cross the line at that time. We can see very small numbers crossing the line for the fastest finish-times (shortly after 130 minutes for men) followed by a sharp increase in finishers from about 180 and 210 minutes, for men and women respectively.

The insert shows a magnified view of the top-placed men and women, from the 2 hour to the 3 hour mark. We can see how the mens race was fairly competitive with less than a minute separating the first three across the line and eight runners finishing within 3 minutes of the winner. Shalane Flanagan’s win was somewhat more emphatic, beating Mary Keitany by just over a minute.

The most popular finish-time for men is just before the 240 minute mark; in fact there is a series of spikes immediately before this landmark time. This time is also popular for women. The ‘spikey’ nature of this chart reflects how runners, in NYC, as elsewhere, tend to target various landmark finish-times, such as 180, 210, 240, 300 minutes, and also intermediate times such as 195, 225, 255 minutes etc. The finisher-rate through these times favours the minutes directly before a landmark time.

The Wisdom of Age

As we might expect age has a significant influence on average finish-times. As the chart below shows, as we get older, we also get slower, but only after our mid- to late-30s, at least for men. For example, 20 year-old men have an average finish-time of about 260 minutes. By the mid-30s this comes down to about 250 minutes, but as these runners leave their 40s their times gradually slow again.

Experience makes a difference. When we compare the average finish-times of first-timers to repeat runners we see the same pattern with age, but first-timers areslower than repeaters across all ages. For example, female first-timers in their mid-30s have an average finish-time of about 290 minutes, compared to about 275 minutes for females of the same age who are not first-timers. The same holds for men and this experience dividend persists across all ages.

Good-for-Age

It‘s not all bad news for older runners. Obviously age and experience are connected, after all older runners are more likley to have run multiple marathons. To better understand this relationship between age, experience and finish-time we can use the famous Boston Qualifier(BQ) finish-times. These are gender-based and age-based finish-times, which runners need to satisfy in order to be eligible to qualify for the Boston Marathon. For example, at the time of writing (October 2017) male runners in the 35–39 years old age bracket needed to finish no later than 3 hours and 10 minutes (190 minutes) to qualify. For similarly aged women the qualification time is 220 minutes.

The point is that by using these BQ thresholds we can evaluate how runners in different age-groups are performing, relative to one another, by calculating the percentage of runners achieving their BQ time within these groups. In the first two bar charts above we plot these BQ rates for men and women, separating first-timers from repeaters, to get a sense of how experience helps. Not surprisingly experience helps a lot. For men and women, the BQ rate for those running repeat marathons is always a lot better than the BQ rates for first-timers, across all ages. Getting older also tends to help, as we can see in the trend-lines in the 3rd (bottom) graph: as age increase so too does the average BQ qualification rate, for men and women, first-timers and repeaters. In NYC this year, female repeat runners appear to enjoy a larger BQ age dividend than men, since we can see a steeper increase in the BQ rate for older, female, repeat runners when compared to male repeaters.

A Race of Two Halves

It’s time to move from finish-times, and the end of the marathon, to pacing during the marathon. Since we have access to 5km split-times, it is natural to look at the average pace for men and women across each of the 5km segments. To make it easier to compare runners with different finish-times, it is more natural to look at relative paces for these segments; each relative pace (RP) denotes the degree to which the runner is running faster or slower than their average race-pace. For example, a relative pace of 0.9 for the first 5km means that the runner has run this segment 10% faster than their average pace (for the race as a whole).

In general, male and female runners adopt broadly similar pacing strategies. The first 5k segment tends to be the fastest (RP = 0.92) but then the pace gradually slows until the short (2.2km) final segment, when most runners manage to speed-up somewhat. While men and women both start out fast, women tend to moderate their early pacing more efficiently than men: they slow more, and sooner. As a result they slow down less than men during the latter stages of the race, and they speed-up more for the final segment. For men and women the 35–40k segment tends to the the slowest of the race. Somewhat uniquely, in New York we see a short-lived speed-up for many runners between 25k and 30k, probably as runners benefit from the short decline coming off the Queensboro bridge as they approach the 16 mile mark.

When we look at the pacing of those runners who achieve a BQ time (by definition, those runners who are running races that are considered to be very good for their age) we see a much more even pacing profile, and one that is almost identical for men and women. BQ runners still tend to start fast, but only about 3% faster than their mean race pace, and they still finish more slowly, but not as slow as regular runners; this is consistent with other marathons.

Doing the Splits

For most runners then, the NYC marathon is a race of two halves, a somewhat speedy first-half, followed by a much slower second. The relative difference between the first and second half is commonly used as an indicator of pacing discipline. Thus, a relative split of 0.1 (a positive split) means that the first half is 10% faster than the second while a relative split of -0.1 (a negative split) means that the first half is 10% slower than the second.

As the graph below shows, very few of New York’s runners, only about 5%, manage a negative split (relative split < 0) and with little difference between men and women. However, a significant difference exists between the sexes when it comes to positive splits (relative split > 0). Men tend to run fewer small positive splits (0 < relative split < 0.2) and more large positive splits (relative splits > 0.2), compare to women.

For example, about 2.25% of women run a 10% positive split (relative split = 0.1) compared to only 1.75% of men. In contrast, for 30% positive splits (relative split = 0.3), representing a very significant slow-down in the second half of the race, there are more than twice as many men (0.6%) than women (0.25%).

About the Wall

Large positive splits are likely to identify runners who hit the dreaded wall. For the purpose of this analysis we determine that a runner hits the wall if their relative split is greater than 0.33. In other words, they must slow by at least 33% in the second half of the race.

In the graph below we compare how often men and women hit the wall, using this 33% threshold, and based on their finish-time. Overall, NYC runners enjoy relatively low rates of hitting the wall, much lower than Chicago, for example, despite Chicago’s flatter course. As usual men are significantly more likely to hit the wall than women, across all finish-times and, generally speaking, the likelihood of hitting the wall increases with finish-time and peaksaround 320 minutes when about 12% of men hit the wall, compared to less than 4% of women.

Once again there is some good news for runners as they age, especially for men becsuse the rates at which men hit the wall gradually fall with age (see the graph below). In part this is probably due to marathon experience but a similar effect is seen, but not shown, for first-timers and repeat marathoners, so the decline is at least in part due to age alone. The effect is more pronounced for men than for women, but then men hit the wall much more frequently than women to begin with, and thus there is a far greater room for improvement among men; in fact, the rate of hitting the wall tends to increase somewhat for women from their 50’s. Regardless, women continue to enjoy far few incidents of hitting the wall than men, for virtually all ages.

Conclusions

The NYC marathon is among the largest and oldest, big-city marathons in the world. It is also one of the most diverse, attracting runners of all ages, from across the world, fast and slow, novices and veterans. This year’s event was no different with more than 50,000 runners from more than 130 countries. Hopefully this analysis helps to shed to some light on how the event unfolded for these participants this year.

barrysmyth

Written by

Professor of Computer Science at University College Dublin.

Running with Data

A collection of articles at the intersection between data science and endurance running.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade