Well, well. Kenyan marathon runner Sabastian Sawe has officially broken through the fabled “sub-2-hour” marathon barrier.
On a reportedly perfect Sunday, 26 April 2026 in London, the 31-year-old Sawe ran through the finish gate on the Mall in front of Buckingham Palace’s gilded architectural flourishes in an official marathon time of 1 hour 59 minutes and 30 seconds.
This betters the marathon world record by a whopping 65 seconds, the largest single improvement since 2018. The previous world record was held by the late Kelvin Kiptum, also of Kenya. Kiptum’s 2 hours and 35 seconds, set in Chicago in 2023, now somehow seems an entire era away.
In fact, saying Sawe broke 2 hours is something of an understatement. Such was the brilliance of the moment, that Sawe pushed the second-placed Yomif Kejelcha of Ethiopia below the sub-2-hour mark as well, just 11 seconds behind Sawe.
Keep up with the latest headlines on WhatsApp | LinkedIn
But as we absorb all of this, it’s hard not to wonder, “What next?”
My interest as a data scientist and economist (and fellow runner) lies in analysing the historical progression of the men’s and women’s world marathon records.
If sub-2 was the driving force behind the marathon in the last decade, what’s left to aim for?
Read more: Marathon under 2 hours is closer than ever – scientist shows how Kenya’s Kiptum tests human limits
Humanity seems obsessed with the limits of human creativity, ingenuity and performance. We award extravagant prizes for world firsts and remember the greatest achievements through bronze statues in prominent squares the world over.
But can we actually calculate these limits? Is there a “maths” of human endeavour?
Historical world record progression
Back in 2018, the legendary Eliud Kipchoge of Kenya ran 2 hours 1 minute and 9 seconds in the official Berlin Marathon. At the time, many dared to dream Kipchoge could be the one to take the men’s marathon below 2 hours.
In fact, a year later, Kipchoge appeared to do just that – running a phenomenal 1 hour 59 minutes and 40 seconds in a tightly orchestrated “breaking 2” display in Vienna.
However awe-inspiring, the Vienna effort would never make it into the official marathon books. The run was contrived in a number of ways, fully understood and acknowledged by Kipchoge and the organisation around him. This was never about the record, but instead, it was, he said, about proving that limits are there to be broken.
Around the same time, I had been working on a statistical approach to modelling the progression of marathon world records over the last few decades. I was intrigued to apply learnings from technological change in economics to the question of human performance.
There are all kinds of factors that feed into a world-record marathon performance. These range from training methods, nutrition, supplementation and biometrics, to performance analysis, and of course, clothing and shoe technology.
However, my approach, drawn from the economics of innovation, is founded on the idea that while performance gains can be made in any of these areas at any time – providing innovation rates stay steady over time – then the next world record marathon performance should be somewhat predictable.
Back then, I estimated that the official men’s marathon would break the sub-2 barrier around May 2032. That is, assuming a pretty rare 1-in-10 chance on any given marathon day of it happening.
Since then, we’ve had Kipchoge himself break his own record at Berlin in 2022, then Kiptum in Chicago in 2023, and now Sawe in London.
At each point, I’ve adjusted my predictions, since the model can use the new world record marks to improve its accuracy.
My most recent prediction, made in October 2023 for a runner similar to Kiptum, would be that the official sub-2 would go down in March 2027. From the perspective of a prediction exercise starting with data from the 1960s, Sawe was just a touch early!
How likely was Sawe’s run?
Using my original modelling framework, if we include data only up to Kiptum’s Chicago run in Oct 2023, the likelihood of a sub-2 on 26 April 2026 is estimated to be 1 in 4.29 (just less likely than 1 in 4 odds). In other words, pretty likely!
However, this is the likelihood of a run of just under 2 hours – 1 hour 59 minutes 59 seconds to be precise.
But Sawe went well under 2 hours, so what were the odds of his actual run?
If I use my framework to calculate the odds of Sawe’s actual time on that day, given the sweep of historical world records since 1960, I find the likelihood of 1 hour 59 minutes, 30 seconds on 26 April 2026 to be 1 in 7.4 (around 2 in 15) – that’s pretty rare.
Clearly, a lot of things had to click for the performance that played out in London. And indeed, the backstory already includes:
- the timing of Sawe’s fitness meshing perfectly with the London event;
- the importance of getting fuelling and shoe technology right;
- the “just so” conditions in London on Sunday (something that was absent in Berlin during Sawe’s previous attempt on the record); and, of course,
- the competitive environment that saw Sawe pushed by the second-best-of-all-time Kejelcha until the final few hundred metres.
So then what’s next?
My statistical framework uses an assumption that, over time, performance gains get harder and harder to achieve. Any of us who have aimed to improve on our local park run time will know all too well how hard it becomes to eke out more performance gains after the initial euphoria of the first week or two’s improvements is over.
In my model, if we follow the improvement process out for very long time periods, we can estimate the eventual limits of human performance. That is, an estimate of the best possible human marathon time ever. I call it the “limiting” time.
In 2019 when my findings were first published, based on men’s world record times up to and including Kipchoge’s world record of 2 hours 1 minute and 39 seconds set in 2018 in Berlin, the limiting men’s marathon time came out to be 1 hour 58 minutes and 5 seconds.
In 2023 I updated this forecast to include Kipchoge’s next world-record time of 2 hours 1 minute and 9 seconds (also set in Berlin, 2022) and Kiptum’s astonishing Chicago run of 2 hours 35 seconds (2023). At that time, and following the “Kiptum line” – a runner like him closer to the 1 in 4 odds line – the new limiting marathon time dropped to 1 hour 55 minutes 40 seconds.
As I remarked then, Kiptum had given the limits of human performance a real bump.
After Sawe obliterated the men’s 2-hour barrier, rerunning my model sees the limiting time once more drop, but this time, not by quite so much.
The new limit comes out to 1 hour 54 minutes – a full 5 minutes 30 seconds faster than Sawe produced in London. In performance gap terms, there is still around four and a half percent of performance gains to be made.
Naturally, there are a lot of inherent assumptions. And such is the exercise that new data points (new world records) tend to have a significant impact on forecasts. Furthermore, we are talking here about the limits of human endeavour – potentially hundreds of years into the future.
The tiniest deviations in a line of forecast today can have outsize impact on a point thousands of days into the future.
Which is a long way of saying, when Sawe’s Italian coach, Claudio Berardelli, hinted that Sabastian might go faster on a better suited course like Chicago or Berlin, I for one, will not be surprised.
The statistical arc of human endeavour in the marathon keeps bending upwards. There is still much to be inspired by.
Simon D. Angus, Professor, Department of Economics & SoDa Laboratories, Monash Business School, Monash University