2:01:39 in Berlin

A Comparative Data Analysis of Eliud Kipchoge’s World Record at the 2018 Berlin Marathon

The build-up for this year’s Berlin marathon was a little more exciting than usual. There was something in the air and a hint that perhaps we might witness something special. In the time since Eliud Kipchoge’s Breaking2 attempt on the Monza race-track in Italy, which secured him a place in the history books with 2 hours and 25 seconds over the marathon distance, he has seemed inevitable that he would go on to break the currentmarathon world-record, secured by Denis Kimetto in 2014 with a time of 2:02:57. And so, as Sunday morning approached the talk seemed less about if the current world-record would fall, and more about by how much.

Since 1998 there have been 7 new male marathon world-records set on Berlin’s fast and flat course, 3 of them since 2010. Would we see a another? Sure enough, as if to order, Eliud Kipchoge served the latest world-record to Berlin crowds, crossing the line in an incredible 2:01:39, smashing the previous record by 1 minute and 18 seconds; we haven’t seen a record breaking margin of more than a minute since Derek Clayton lowered the world record from 2:12:00 to 2:09:36 in 1967. It was all the more impressive because the 33-year-old Kenyan achieved it while running the last 17km on his own after the last of his pacers, not to mention the rest of the field, drifted away long before the finish.

2:01:39 in Context

Let’s begin to try to put Kipchoge’s 2:01:39 finish-time into context. It corresponds to an average pace of 4:38 minutes per mile or about 2:53 mins/km, which is the equivalent to running a 17.4 second 100m sprint. Try it; it’s non-trivial. Now imagine doing it 421 more times to get to the finish-line!

What about the rest of the field who started with Kipchoge? At the time he crossed the finish-time how far has the average runner left to run, assuming they continued at their average race-pace? 21.6 kms! Imagine that, the average runner still has more than half the marathon to run by the time Eliud Kipchoge crossed the finish-line; it’s a little less for men (20.9 kms) and a little more for women (23.2 kms).

Figure 1. The fraction of (a) male and (b) female runners with a given distance yet to complete at the time Eliud Kipchoge crossed the finish-line.

For example, in Figure 1 we see the fraction of male and female runners with a given remaining distance at the point at which Eliud Kipchoge crossed the line. In summary, while Kipchoge was celebrating his win, approximately 50% of men and more than 70% of women still had at least half a marathon to complete. Notice, the spike in Figure 1(a) around the 20km mark. This represents the significant cohort (approximately 12%) of men targeting a 4-hour finish; the same spike is there in Figure 1(b) for about 10% of women, but it is somewhat overshadowed by a larger fraction of women targeting a 4.5 hour finish, and who have 25 km still to run when Kipchoge finished.

2:01:39 in Comparison

How does Kipchoge’s new record compare with previous WRs? The great thing about Berlin is that it attracts the best of the best and so not surprisingly quite a few WRs have been established there. This means we can perform a like-for-like comparison by comparing Kipchoge’s race to (in our dataset) the previous 3 world-records, all of which have been produced in Berlin:

Figure 2 shows the pacing (in decimal mins/km) for all 4 WRs across each of the 5km segments of the race (5 km, 10 km, …, 40km), and the final 2.195km segment. In each case the dashed line reflects the average pace for the runner in question.

Based on the timing data released by the Berlin Marathon, Kipchoge ran the first 5 km in 14 mins, 24 seconds, or anout 2:53 mins/km, and he ran between 2:52 mins/km and 2:56 mins/km until the final 2.195 km stretch, which he dispatched at just under 2:48 mins/km pace, faster than any of last three male WRs managed in Berlin.

Figure 2. The race segment pacing (mins/kms) profiles for each of the last four male WRs at the Berlin Marathon.

As an alternative perspective, the equivalent graphs below show the race segment pacing based on the projected finish-time that a given pace would deliver. Kipchoge kicked off his first 5 km at just over 1:21:30 pace, significantly faster than recent WR races. For the penultimate 35–40 km segment he was running at a sub-1:21:00 equivalent pace, significantly faster than Kimetto, Kipsang, or Musyoki at the same stages in their WR races, before accelerating to just over 1:57:00 pacing for the final 2.195 km segment, a pace that none of the recent world-record holders could come close to in their finishes.

Figure 3. Race segment pacing expressed as equivalent marathon finish-times.

2:01:39 by the Splits

Kipchoge, and Kimetto before him, secured their world-records with negative splits, running the second half of the race more than 30 seconds faster than the first half. In contrast Kipsang and Musyoki secured their records with slight positive splits, running the second half about 10 seconds slower than the first; see Figure 4. Since 2010 about half of the Berlin winners have run negative splits, averaging a 30-second difference, and half have run a positive split, averaging a 45-second difference.

Figure 4. Halfway splits for the last 4 world-records at the Berlin Marathon.

Figure 5 summarises the fastest and slowest segments for each of the four runners in their WR races. The bar charts indicate the fastest/slowest paces (again in mins/km), while the line graphs indicate the position of the relevant segment (fastest or slowest) in the race. For example, in Figure 5(a) we see that Kipchoge’s fastest segment (at just under 2:48 mins/km ) was the final segment of the race. While none of the other record holders matched Kipchoge’s fastest pace they did all also run their fastest segments late in the race.

Figure 5. The fastest and slowest segments for recent Berlin world-record races.

Kipchoge’s slowest pace, 2:56 mins/km, was faster than the slowest paces of the other recent WRs at Berlin; see Figure 5(b). In fact, for Kipchoge, this slowest pace was faster than the average race-pace for the Kipsang and Musyoki world-records! Even when he was running his slowest, he was still running faster than 2013 (and therefore 2011) world-record times.

In Figure 5(b) it is also interesting to note the difference between the WRs in terms of where in the race the runners were running their slowest paces. Kipchoge and Kimetto, ran their slowest sections early in the race (10–20 km for Kipchoge and 15km-20km for Kimetto) and were largely able to maintain and even improve their pace later in the race. In contrast, Kipsang and Musyoki ran their slowest sections much later in the race, around the halfway mark for Kipsang and the 35–40km section for Musyoki.

2:01:39 from Behind

Finally, let’s replay these four Berlin WRs in the same (virtual) race, to get a better sense of how the lead pack might have looked had these incredible runners toed the line together at the height of their performance, to run their WR races. Obviously we cannot account for the additional competitive tension that this might have introduced, but we can at least compare their pacing and timing information to get a better sense of how this lead pack might have developed.

Figure 6. The number of seconds behind Kipchoge for Kimetto, Kipsang, and Musyoki as the race unfolds.

Using Kipchoge’s WR run as the baseline, Figure 6 shows the number of minutes each recent world-record holder was behind Kipchoge at the end of each 5 km race segment (and at the finish-line) in this virtual race. Kipchoge leads from the start, stays in front, and gradually but steadily extends his lead after the 15km mark.

Kipsang stays in touch with Kipchoge during the first 15km of the race, getting to within just over 7 seconds by the end of the 15 km mark. But after this, begins to drop back. Meanwhile, Kimetto, after spending the first 20 km in 4th position (by 20 km he is 40 seconds behind Kipchoge), moves into second place and starts to recover some ground to get within 35 seconds of Kipchoge by the end of 35 km. Despite the beginnings of what might have been a late surge by Kimetto, Kipchoge is too strong and extends his now unassailable lead all the way to the end, finishing a full 78 seconds ahead of Kimetto, just over 103 seconds ahead of Kipsang, and almost 119 seconds ahead of Musyoki.

To put this another way, we can estimate how far behind (in distance rather than time) each of Kimetto, Kipsang, and Musyoki would have been when Kipchoge crossed the line, based on their average race paces. The results of this are shown in Figure 7: Kimetto would have finished 446m (almost half a kilometer) behind Kipchoge; Kipsang would have been just under 600m behind; and Musyoki would have been 675m back. Not even even close!

Figure 7. An estimate of how far behind each of the previous world-record holders whould have been when Kipchoge crossed the finish-line.

Summary

By any objective measure Kipchoge’s Berlin world-record race was nothing short of stunning. He obliterated Kimetto’s 2014 record, set on the same course, and his incredibly disciplined negative split is all the more impressive because he did it largely on his own, after dropping the last of his pacers shortly after the halfway point. It seems right and proper that on September 16th 2018 the fastest marathoner also ran the fastest marathon.