Welcome to the Science of Sport, where we bring you the second, third, and fourth level of analysis you will not find anywhere else.

Be it doping in sport, hot topics like Caster Semenya or Oscar Pistorius, or the dehydration myth, we try to translate the science behind sports and sports performance.

Consider a donation if you like what you see here!

Did you know?
We published The Runner's Body in May 2009. With an average 4.4/5 stars on Amazon.com, it has been receiving positive reviews from runners and non-runners alike.

Available for the Kindle and also in the traditional paper back. It will make a great gift for the runners you know, and helps support our work here on The Science of Sport.

Wednesday, November 04, 2009

Are marathon times getting faster or not?

A Further Analysis of Marathon Performances

As US fans continue to drink in the first American victory in New York since 1982, we thought we would follow up the race report with a further analysis of marathon times.  If you read the comments to the race report, Joe Garland and cassio598 mentioned that in the early years of the race the course was four laps through Central Park, and that to examine the progression of times we should rather analyze the winning times since 1976.  The graph in question is this one here:

So in 1976 the current race course was adopted, the one that passes through all five boroughs, and since then the average winning time has deviated year-to-year by only about 90 s and has gone below 2:08 only once (in 2001 when Jifar set the current course record).  The blips, for your information, are from 1984 when the ambient temperature was in the upper 70s F with 60-90% humidity, and 1990, when the max temperature on the course was 79 F.

One of the readers, cassio598, suggested analyzing the data from 1976 to the present, because then it appears that the winning time creeps lower over time.  And the verdict?

The regression line for these data does indeed move in a downward direction, however the coefficient of determination ("r-squared") is only 0.16, which does not suggest a very good fit.  But the bigger issue is that since 1981 there have been six times that are very close to 2:08, and so it is hard to conclude that the times are getting faster when almost 30 years ago the winning time was one minute faster than the winning time this year.

The nature of the course:  A peculiar finding

Perhaps the really interesting comparison comes from looking at the the "current" average times of all the big city marathons---London, Boston, Berlin, Chicago, and New York---and see how the times of this decade compare to the "average" times from a decade ago.  So what you see below is the average winning time from 1990-99 (blue column) compared to the average winning time from 2000-09 (red column) for all of the above-mentioned races.

Recall that Chicago, London, and Berlin are flat courses, while NYC and Boston are not flat.  You can see that the winning time for all the flat courses has decreased by about 2.0% while NYC and Boston have remained more or less the same.  How is that?  Even when scrutinizing the Boston data, it comes out that in 1983 the last American to win, Greg Meyer, ran 2:09:00.  It also turns out that the average winning time for the Kenyans (and a Kenyan has won Boston many many times!) is 2:09:32.

One key factor is pace-makers - at least for the last few years, neither Boston nor New York have employed pace-makers, whereas Berlin, London and Chicago have.  Berlin, in particular, has evolved into a staged record attempt every year.  Fast times are designed on these flatter courses, and that is certainly a factor to take into account.  However, it does not explain why the Kenyan dominance, particularly in Boston, has been sustained with slower times than in the 1990s (and, as in the case of Meyer, the early 1980s!)

This all leads me to believe that the less than flat nature of these two courses may "neutralize" the speed of the East Africans.  How this works we have not figured out just yet, but how can it be that while the average best American marathon time is so many minutes slower than the best African times, yet on courses like these we had runners over 20 years ago running just as fast as the Africans are running now?

Conservative race strategy?

Perhaps the hilly nature of Boston and New York dictate a more comfortable (i.e. ~2:09) pace for the first half, as no one wants to blitz the first 20 km in 57 min only to have the wheels come off during any bumps in the second half.

One question we asked ourselves after New York was, "What if Wanjiru ran New York. . .?"  Because if he ran aggressively like he ran Beijing, London and Chicago, at halfway he might have had only two runners with him and not seventeen others as was the case in New York.  With all due respect to Meb, if the 21.1 km split was 1:03:00 and not 1:05:11, he and most of that pack would likely not have been there to contest the race around Manhattan.  Having said that, however, it will be really interesting to see how Meb goes next year, because he did set a half marathon PB this year with a 61:00 back in April.

The fastest of the fast:  Berlin

One thing that also comes out of the graph above is that Berlin has been the fastest course around for years.  We all know why the average time since 2000 is about 2:06:20, but even in the 1990s Berlin was almost one minute faster than London and Chicago.  Seeing as how the margins are so tight now, it seems that Berlin is perhaps the course most likely to produce the next marathon world record.  Previously, when the record was "softer" and still above 2:05, the turns, bumps and cobbles of courses like London were not really obstacles.  Now, however, the record is so close to the limits of the current crop of potentials that even having to take several extra 90 degree turns will be meaningful.

Dubai may become the new player on the record scene, and along with Berlin, has become one of Gebrselassie's two paced record attempts each year (as it will be in 2010).  So it may have the fastest time of all, but being such a recent race, and really only by virtue of the fact that Geb has chosen it, it hasn't proven itself over decades, yet.

The answer to the title of this post:  YES!

If we looked only at New York and Boston we might conclude that, no, times are not getting faster, because in those two races there seems to be a real limit in the form of the race course.  While a 2:07 in either of those races will almost guarantee you a win and probably the course record too, a 2:07 in London is only good enough to squeeze into the top five and probably a podium in Chicago, but no guarantees you will be victorious.  But obviously the times are getting faster, because we have seen a progression of world records in 5000 and 10000 all the way up to 42.2 km, and no one would argue with that.

What we have stated before is that the marathon class of 2009 has been particularly fast, and that actually warrants an analysis on its own---between Kibet's and Kwambai's 2:04:27 in Rotterdam, Wanjiru's 2:05:10 in London, Gebrselassie's 2:06:13 in Berlin, Wanjiru's 2:05:41 in Chicago, and even Vincent Kipruto's 2:05:47 in Paris, the average winning time this year is 2:05:28, which might be the fastest average for one year!



David K said...

Do you have any thoughts on what effect wind speed may have had on the winners' times over the years? Look at the average wind speed and direction for each race and factor in (somehow) how the winner won - from the front, from a pack, etc. David K

David O said...

Maybe there is an inverse relationship between the prize money and course records. Where the prize money is greater, the course record is slower, because the races are more strategic.

Giovanni Ciriani said...

Where can I get your data. I'd love to run a regression between temperature and times. I've made a table of temperature available (with humidity and wind where possible) at http://www.globussht.com/NYCmarathonweather.xls

Ross Tucker and Jonathan Dugas said...

Hi Everyone, thanks for the comments.

To David, the wind can play a role, but when you start to average 10 races then this effect starts to wash out, as one year it might have been a real factor but in some other years it might have been calm. So we make an assumption here that since we are averaging many races we minimize or get rid of this effect.

To David O, perhaps? I have not tried to tabulate the prize money each year. In theory you have a point, but I am not sure if that will translate into different times. But would be an interesting analysis!

To Giovanni, email us at sportsscientists@gmail.com. I have pasted the marathon times into your spreadsheet and can send it back to you. . .

Kind Regards,

Giovanni Ciriani said...

I made up your data, reading it from your graph. By adding year of race, Fahrenheit, humidity and wind the R-squared changes imperceptibly, from 0.159 to 0.161. Surprisingly weather conditions are not a factor.

Danilo Balu said...

I know that it's not necessarily the case, but to evaluate a race as faster over the years, don't you think you should take more than just the winner?
As long as you can have an outlier or a tactical race, maybe taking the first 5 or 10 runners... what do you think about it?
Cheers from Brazil, country of Ronaldo da Costa (2h06'05" in 1998)!

Ross Tucker and Jonathan Dugas said...

Hi Danilo,

Thanks for your suggestion about taking the top finishers and not just the winner. . .

I actually agree with you entirely that it would enhance the quality of the analysis. However I accumulated the data some time ago while trawling the web on a rainy day and collected only the winning time from all of the big city marathons, so I do not have that data right now.

But agreed, it would be a better representation of the "true" speed of the race.

Thanks again for commenting!

Kind Regards,

Antonio Perricone said...

Did you factor the extra 400m which were added in the mid 80'? I am not sure about the date, but I recall the NYRRC saying that they had "discovered" that the course was short.

Meg & Dave said...

I've been pondering this the last couple days and... It seems to me that marathon times shouldn't really get faster without evolutionary change. Humans' ability to run long distances efficiently is theoretically rooted in a pre-weapon "hunting" method of running animals to exhaustion. The more efficient we were, the more likely our children were to survive because we could quite literally bring home the bacon. You all know metabolic science way better than I, but in such a pure, long contested sport like marathon running, it would seem that fundamental evolutionary adaptation would be needed to actually change the times. Training techniques and technology would seem to have much less affect in running, a fundamental human function, than in other sports that require more adaptation. Obviously, some modern technology, simple expansion of the pool of contestants, and things like drafting and careful race strategy will bring times down to a certain extent, but from a physiological standpoint... without "performance enhancers" like drugs or carbon fiber legs, I just don't believe there is a system of periodisation or long term adaptive training that will substantially change marathon running.

Of course, what do I know? This is all speculation from a neophyte non-runner. :)

BridgeportJoe said...

Obviously, some modern technology, simple expansion of the pool of contestants, and things like drafting and careful race strategy will bring times down to a certain extent, but from a physiological standpoint... without "performance enhancers" like drugs or carbon fiber legs, I just don't believe there is a system of periodisation or long term adaptive training that will substantially change marathon running.

Your first phrase answers your own question. The reason that we've seen times drop so much in the past ten years or so (basically about four minutes, after taking decades to drop that much prior to the mid-90s) is that Kenyans and Ethiopians started to flock to the marathon en masse.

Is there another paradigm shift that might drop times a comparable amount? Who knows. But there could be. And even if not, because everyone has different optimal training regimes, we should still see a few dozen seconds cut off here and there as genetically-gifted runners get closer and closer to their own perfect mix.

Amby Burfoot said...

In this wonderful paper (abstract below) Mark Denny showed that dog and horse times have effectively reached a plateau, even though these animals are bred for speed-distance. It's reasonable to expect the same for humans, he says, and I agree. I don't expect Gebrselassie's time to drop much in the coming years/decades, and I don't expect to see Radcliffe's record bettered for a long, long time. In fact, I won't "see" a new women's record at all, because I won't live long enough.

J Exp Biol. 2008 Dec;211(Pt 24):3836-49.
Limits to running speed in dogs, horses and humans. Denny MW.

Are there absolute limits to the speed at which animals can run? If so, how close are present-day individuals to these limits? I approach these questions by using three statistical models and data from competitive races to estimate maximum running speeds for greyhounds, thoroughbred horses and elite human athletes. In each case, an absolute speed limit is definable, and the current record approaches that predicted maximum. While all such extrapolations must be used cautiously, these data suggest that there are limits to the ability of either natural or artificial selection to produce ever faster dogs, horses and humans. Quantification of the limits to running speed may aid in formulating and testing models of locomotion.

Ross Tucker and Jonathan Dugas said...

Hi Amby

Thanks for the mail!

Yes, we've seen that study. In fact, last year when I did my "2008 review", this paper featured in my Number 1 story of 2008 - world record limits. Here is the link:


I agree that barring some medical miracle, we will be perhaps the last generation that sees records set frequently. By 2050, we may be right down to one per decade.

Then again, as swimming has shown, technology can always intervene!

Hope you enjoyed New York! Great racing, full of surprises!


Anonymous said...

One note that I didn't see mentioned concerning New York and Boston is that over the past decade the really big players, (Geb, Tergat, Khannouchi, Wanjiru, etc) have chosen to only run the fastest courses. Since they know times would be slower in New York and Boston, they don't bother with them, especially if they would rather gain a $$ bonus for a world record.

Unknown said...

In response to what Amby posted, the biggest difference is in the mind - animals bred for speed don't know if some other animal across the world covered six furlongs a fraction of a second faster. Does Hall run 2:06 if he's never heard of Geb? Does Tyson Gay run 9.7 if Bolt hasn't run 9.6?. There is a physical limit to what humans can do. But what we thought was a physical limit quite often vanishes when someone opens the door and shows us what's possible.

hgoforth@cox.net said...

Hal Goforth says,
I totally agree with Amby and Stephen.
World class runners will all tell us (and have if one is a student of the sport) that the mind is the ultimate limit to physical performance. We(in the US at least)have gone through a time when the discomfort of demanding training and effort required for peak racing performance has not been as common as the late 70-mid 80's. In those years, it was the norm (to train to the limit and race to limit) and each time Boston raised the qualifying standard, the real competitive runners met that mark regardless of the effort required. Based on my knowledge of exercise physiology and experience racing 34 Boston marathons, I can't predict an imminent physiological limit (at least not in the next few decades. I know a bit about the human body, it's muscles and metabolic components. These still have much to show us without doping,and with the application of optimal training, recovery, hydration, nutrition, and the critical importance of mental drive. The mind is by far the the most powerful limiting factor to performance. It still has more to show us and it will continue to improve performances. However, injuries, their prevention and recovery may ultimately limit runners performance. Because the highly competitive mind may fail to protect the body.

Greg said...

@Antonio - The NYC course was found to be short after Salazar's "world record" run in 1981. The course measurers had followed the middle of the road instead of the tangents as they should have. I believe the finish line was moved back in time for the '82 race.
To Danilo's and others' points: I agree that the results could be much different if we looked beyond the winners. Berlin has become nothing more than a paced time trial for Gebresallasie. Similarly, the flat courses and pacesetters at Chicago and London mean these courses virtually self-select for fast times, whereas NYC, falling so near Chicago's date, and Boston, usually the same weekend as London, self-select for great, but not necessarily fast, runners.
It should also be said that the Chicago, London and Berlin races - having been founded after NYC and Boston - have had greater variance in their prize and appearance money purses, and therefore the relative speed and depth of the elite fields they have been able to attract, than NYC and Boston.

Anonymous said...

You mentioned that the Boston times run by Kenyans today are not much different than Greg Meyer's 2:09 in 1983. So how come Americans are not winning Boston?

Could it be that elite American marathoners elect to run the faster London course as their spring marathon in order to secure an Olympic standard? Ryan Hall's London times in the last few years would have out him in the hunt for Boston.

Do you think their is more "gaming" by American marathoners than in the past to achieve an Olympic standard that is making them shy away from slower courses?

Ross Tucker and Jonathan Dugas said...

Hi Anonymous

I think that for a long time, the standard of running in the US was just not good enough to compete. They certainly weren't competitive in the London/Berlin races either, so I don't think it's a case of choosing the faster races. They just were not good enough - I don't have the numbers on hand, but I suspect you'd struggle to find US men running faster than 2:10 in the period in question.

What we're seeing now is a revival, and Hall, Meb, Ritzenheim lead a new generation that are good enough. Earlier this year, Hall took third, and we've now seen a win (Meb) and fourth (hall) in New York. So I think it less to do with choosing races than the fact that they weren't up to it.

We should see more Americans being competitive from now on, a return to the level of the 80s, of sorts. What Meb, Hall and others will do is create more prestige and the next generation should be even better, so by 2015, perhaps they'll be consistently challenging the Kenyans.

The US situation back then is very similar to the UK situation today, by the way. They've slowed down, not that the Kenyans have moved the event out of reach.


Mark B said...

I'm guessing but the Boston statistics for 200x have probably been skewed by a run of particularly difficult weather. Boston 2004 was extraordinarily hot (~90 F), 2005 was unusually hot for Boston in April (~75 F), and 2007 had a very strong prevailing wind on the point to point course.

Anonymous said...

Another very interesting post.

It might be interesting to look at the times of the tenth placed finisher or some similar position as a proxy for the depth of quality in the field.


Anonymous said...

Although technically the winning times have improved a little over the last couple decades, isn't it misleading to say enthusiastically "the answer is YES!" and "obviously because 5000m and 10000m records have come down".

Wouldn't a more telling title or conclusion be that the best times have barely budged much and are more rarely bettered in recent years and any trend of faster running over the last few decades may be completely indiscernible from the data especially when other factors like weather cause much greater variability in results? Your own data show no statistically significant improvement for the NYC marathon. Boston has been slower and the other three have had 2-4 minute improvements between the '90-99 average and the '00-09 average, but it is hard to say if that is significant from the data. After all. humans have on average one breast and one testicle, but that fails to highlight that men have two testicles and women none. Even the 5K and 10K improvements have been smaller and more infrequent with more records dropping in the '90s and very few this decade. Of course, you have written a post about this so I won't belabor the point, but which is it? Are marathon times getting faster or are they reaching a performance plateau?

I do have a couple questions from the article and comments on it. Why would the best American time 30+ years ago be better than the best African winning times in recent history on the same course? It seems contrary to popular thought. The second question relates to the popular idea of drafting since one commenter talks about wind etc. It seems like drafting is useful for auto races and bike races where the speed and apparent wind is so high relative to any ambient wind and the "racers" (car or bike) are designed to be aerodynamic and maintain a laminar flow of air over them like a sail or a wing. Running, however, is slower and messier with flailing arms and legs that break up the wind and may offer very little shadow or drafting benefit. With the exception of watching Bekele who seems to be dancing with the runner ahead of him, it seems few runners even get close enough to benefit from any kind of real drafting. It seems to me the whole concept could be nonsense or completely misunderstood and improperly utilized. Have you guys posted anything on drafting or do you know of any really good research on the topic?

fischöl said...

I have read the article based on the Marathon races and its activity graph chart.I agree with that It seems to me the whole concept could be nonsense or completely misunderstood and improperly utilized.The conclusions may be drawn from the given charts which can give correct results for the race strategies.

Hywel Care said...

I'm not sure that the course difficulty is as great a factor as you conclude. The relative standing of each marathon in terms of the ability to draw the stars of the sport i.e. those who are real contenders to win in the majors in the fastest times (Lel, Wanjiru, Gebreselassie class of athletes). If we look at the spring marathons, London now draws a far more stellar field than Boston whereas 20 years ago the reverse may have been true, and the relative changes are perhaps attributable to this fact not the course difficulty (although London's course changes of recent years removing the cobbled sections and some of the tight turns will exaggerate the drop in times as similar courses aren't being compared.). I think the same pattern is evident in the autumn, though perhaps not so much to the detriment of New York in the way it has been to Boston, NY has remained prestigious and has seen it's wining times come down as a result. Berlin & Chicago, however, have gone from marathons not in NY's league in terms of prestige to taking places in the top 5 marathons in the world, hence the exaggerated drop in their times. I suspect their improvements will level off over the next decade when they simply maintain a position at the top tier of marathoning on the existing courses so the room for improvement is less dramatic.