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.

Sunday, September 20, 2009

Berlin 2009: Haile Gebrselassie 2:06:08

Berlin Marathon 2009: Live splits and first report

Haile Gebrselassie has won the Berlin marathon, but no world record this time. His time was 2:06:08, which was the result of a dramatic slowing of the pace over the final 10km, and particularly, the final 5km. The record was on, all the way to 35km, but the wheels came off, and a 15:57 5km split between 35km and 40km saw Geb end well short, in what is actually his slowest marathon in 5 attempts.

The numbers tell a story - the table and graph below illustrate just how the race unfolded. I will look much more closely at these numbers in the coming days, and look to compare this race to the 2007 and 2008 races, where the record did fall.

But for now, take a look at the table and graph:

In the graph below, I've plotted the 5km interval times, and in green, the projected marathon time based on the split time. The dashed line at 14:41 shows the pace that was required at the start in order to sneak inside the world record - it represents a pace of 2:56.2/km, and it's clear that Geb was under it (by some margin) for 30km, and then over it (by an even bigger margin) for the final 12km.

How the race unfolded

It's pretty obvious from the numbers how this race unfolded. The early pace was pretty much bang on - they'd requested something around 61:35 at halfway, and so when the mark was reached in 61:44, it suggested the record was on - if anything, they were slightly slow.

Then the pace really picked up - the 15km interval between 15 and 30km was covered in 43:46 (the 10km interval was 29:10!), which projects a marathon time of 2:03:04. That searing pace accounted for Duncan Kibet, who was dropped before halfway, in a very disappointing outing for him.

It meant that Gebrselassie would have only pacemakers for company, and when the last pacemaker fell out just after 32km, the race against the clock was on.

Unfortunately, the clock would be the winner. You'll see from the chart that Gebrselassie fought bravely. He fought to hold the pace together between 30 and 35km, but that split was considerably slower than anything before, and at 14:53, it suggested that the record was starting to slip away. He was still on course, based on his 35km time, but the true story is that had he maintained 14:53 pace for the rest of the race, he would have missed the record by 1 second!

As it was, we didn't see that sprint against the clock, because the split from 35km to 40km was where it all ended. 15:57 for the interval, and the record was very clearly gone.

In the end, the time of 2:06:08 meant that the final 7.195km were covered in 23:31, a pace of 3:16/km, and when you consider that the pace required was 2:56, then you appreciate just how big the slowdown at the end was.

The pacing strategy: Too fast in the middle?

Overall, it shows just how fine the margins are between a good day out and a blowout. It's hypothetical, of course, but what would have happened if Gebrselassie had covered the 15km stretch in the middle in 44:00 instead of 43:46? He'd still have been on course for the record - you will see in the table above that his projected time was well under the 2:03:59 he ran last year.

He hit 30km a full 41 seconds faster than last year, and given how close to the limit he was, this was simply too quick (easy to say in hindsight, I know, but I felt it was ambitious even before). Could have have given up 30 of those seconds and finished faster? Probably not, I think he the record would have eluded him anyway, but the fast pace pushed him out to the very slow finish, I have no doubt.

Of course, perhaps it was just a bad day. There are many ifs and buts, but we saw one thing confirmed today - this record of 2:03:59 is pretty close to the limit for Haile Gebrselassie, and probably ever other man currently racing. That's not to say it can't be broken, and maybe Sammy Wanjiru will be able to edge it down by a few seconds in Chicago, but the days where we expect to see up to a minute cleaved off a time are gone.

I remember back in 2007, when he broke the record for the first time, people waxed lyrical about how it was a matter of time before 2:03 would be broken, and even 2:02. Some people even spoke about a sub-2 hour marathon. I think what Berlin 2009 has shown is that 2:03 is mighty close to the limit, at least for the current generation, and unless everything is perfect, even running 2:04 is a mountain.

As for the pace, I think it was too ambitious. Given the tiny margin for error, setting off at 14:30 pace, and then hitting halfway in 61:44, projecting a 2:03:28 and a record by 30 seconds, that's too ambitious. Of course, some may disagree, but I think the middle of this race was too fast - 2:03:04 pace for 15km in the middle, and that time is paid back with interest at the end!

Nevertheless, a great attempt, very courageous, and maybe there'll be another for Gebrselassie. Age may count against him, and this is the second marathon in a row where he's fallen away at the end, and it's his slowest in 5 outings, but he showed enough in the first 30km to suggest he's still in the kind of condition to run 2:04-something. Whether he has another record in him, time will tell...


Donate and support The Science of Sport

Thank you for taking time to read and follow our coverage of the Berlin Marathon at The Science of Sport. We hope you've enjoyed the insights we've tried to provide! Your time and energies are greatly appreciated!

We run this site as a "labour of love" and will continue to do so, but of course, any support is greatly appreciated. So, if you would like to donate and support our continued efforts to bring you the insights and analysis, please consider donating to our site. You can do so in one of two ways:

  1. If you are reading this on our site, then simply scroll up to the top of the page, where you will find the DONATE button at the top right of the page.
  2. If you are reading this in our email newsletter, click here to be taken to our site, where you will see the DONATE button at the top right of our homepage.
Thank you once again, and keep on reading!



DrTim said...

Awesome! Well up on last year ... C'MON!!!

Anonymous said...

1:27:44 is WR for 30 km!

DrTim said...

We should start a guessing comp ... I punt for 2:03:50 and a new WR :)

Andrew said...

Looking good for a WR, my guess is 2:03:52 :)
Disappointed than Kibet has fallen back. Also, doesn't the pacemaker share the WR for 30km, since they both went through the mark together?

DrTim said...

Don't blow Haile ... DONT BLOW. C'MON!

Sticking to 2:03:50 ...

Ross Tucker and Jonathan Dugas said...

Hi all

Welcome, thanks for comments. Sorry I can't reply in detail - trying to draw graphs and watch at the same time.

He needs to pick the pace up - the last 5km, if continued at that pace, has him at 2:04:05. So it's slipping away, unless he lifts it slightly...it's gonna be mighty close


DrTim said...

Last 2k in 6:13 ... he's gorn. Shame. :(

Ross Tucker and Jonathan Dugas said...

I think the record is gone - he's slowed badly.

I think he might only run 2:04:30.

Andrew said...

Ja, I agree. Think the record is gone.

DrTim said...

Right now you're trying to plot a 16min - 5k ... doh.

Ross Tucker and Jonathan Dugas said...

Well, outside 2:06, that's a dramatic slowdown.

Just shows how close to the limit 2:03:55 is, and he was too quick from 10km to 30 km.

Pity, brave effort, but too quick at halfway, and those middle kilometers cost him big...

DrTim said...

I'm no rocket scientist like you guys but my verdict ... too fast for the temperature ... still I'd be pretty darn happy with a 16min - 5k! Next year.

Anonymous said...


Frans Rutten said...

I doubt if it were exclusively the less favourable conditions to the end to blame. Projected end temp was 17gr C.

It's the 3rd time that Gebr fades towards the end: Amsterdam 2006?, Dubai 2008 and now Berlin 2009. There were/are certainly weather conditions at hand, but in each of those races it wasn't the whole story.

Note for today the fastest 10K segment from 20K to 30K was 29:10. After that he faded dramatically with 30:50 and 7:34. Halfs of 1:01.44 and 1:04.24.

Kiprop with a conservative 1:04.18 ended with 1:02.46
So did Terfa with 1:04.18 and 1:03.23.

Apparantly the marathon WR is so good, that each slight mistake in pacing and/or effect of weather components ruines any attempt.

And for all, no matter how good conditions or pacing will be: in the end the athlete must have the potential of breaking the WR.
We don't know if the potential of Gebr 2008 equals Gebr 2009.

Breaking a WR at random is nowadays definitely out of the question.

Anonymous said...

I was impressed by the last pacemaker (Sammy Kosgei). I think he had a 2:05/2:06 marathon in him today. Can he be the next Abel Kirui (a pacer that transforms himself in a very good contender)?

Anonymous said...

It is possible to put up a graph of each KM splits and/or mile splits?

Ross Tucker and Jonathan Dugas said...

Hi anonymous

If you have those numbers, sure, I'll gladly graph them.

But as far as I know, they don't report the kilometer times.

I will be in Chicago in October, and we are planning a study on pacing there and so we will record times every kilometer, but it's not the standard practice, unfortunately.


djconnel said...

It seems one should be able to estimate a simple model for endurance at a given pace, for example, take GH's best @ 10k and compare to his best @ marathon, fitting k and v₀ to
v/v₀ = (t₀/t)^(1/k),
for some chosen t₀, where t is the duration, then assuming a fixed reserve E (0, 1), integrate over this race
(1 / t₀) ∫ (v/v₀)^k dt,
and compare with an "ideal pacing"
v =
v₀ (t₀/t)^(1/k) =
v₀ (v t₀/d)^(1/k),
to see if his effort was consistent with the ability to have tied his PR.

Another pacing model is the CP model. This would also work. But it may not be valid at times as long as two hours. It predicts a certain pace is sustainable forever.

Anonymous said...

Just to give an idea of how much difficult a sub 2 hour marathon looks:
According to Ken Nakamura the (probably) fastest ever 5k splits recorded for the marathon were:
0k-5k: 14:08 London 2009
5k-10k: 14:22 London 2009
10k-15k: 14:27 Dubai 2009
15k-20k: 14:36 Berlin 2009
20k-25k: 14:35 Berlin 2009
25k-30k: 14:35 Berlin 2009
30k-35k: 14:17 Fuluoka 2008
35k-40k: 14:29 Berlim 2008
40k-finish: 6:10 Berlim 1998
If you sum the splits (of course it's a little non-sense since these were recorded in different routes, although, most are pancake flat): 2:01:39 (2:02:49 Berlin only). Almost convinced that will take 5-10 years just to see a sub 2:03 effort...
Interesting all fastest splits (with the exception of the last - 40k-finish)were recorded in the last 2 years. The marathon "revolution" so to speak.

Ross Tucker and Jonathan Dugas said...

Hi anonymous

That's fantastic data, thank you so much! Will definitely make use of it in a future post!


Adriano said...

This is where I found the data (I'm anonymous)(http://www.trackandfieldnews.com/asset/special/2009/nakamura/Fastest_Marathon_Splits.pdf). All merits for Ken (I just make some calculations and conclusions)!

Adriano said...


D said...

While I agree with you that Wanjiru is perhaps most likely to break the WR, if anyone can, in the current generation, I have a gut feeling that says that Kenenisa Bekele will have an equal chance if and when he decides to run a marathon.

Perhaps you can try and write about why he can or can't

FĂ©lix said...

An rule of thumb in team sports is never change a winning formulae. Why on Earth did Gebrselassie changed his so drastically after trimming 27 sec. off from a mythical record the year before??? I can't understand why he didn't try to run a conservative first half like last year with a ~10-20 sec faster time, and then put the gas on for the last half like last year?? Why did he choose to run so fast and steady for the first half? I would love to learn what were the pacing strategy for previous records.

Kinga Mikolajczyk said...

I really enjoyed reading your post; you presented unique data tables to visualize Haile Gebrselassie'€™s progress during the marathon and showed how he failed to break his record after such a strong 30 km run. I have not found similar analysis of the race anywhere else in the blogosphere so I enjoyed analyzing it. The professional but accessible approach that you present gives me the impression that it is easy to understand your point of view; in my opinion, however, an overly ambitious pace was not the only factor that made Haile Gebrselassie slow down in the last seven kilometers of the marathon.

I disagree with your analysis that the factor of his age was a significant factor in his performance but believe greater emphasis should be placed on the psychological aspect of the sport, especially because Dunkan Kibet was forced to quit the competition at 32-kilometer point. I think that the fact that Kibet quit made Gebrselassie feel more confident about his position so he slowed down slightly, maybe even without noticing. I am sure that Gebrselassie was aware that he was the one of the top contenders who had a real chance to win a gold medal and the fact that he scored such good time intervals in the first half of the marathon made him believe that he was actually capable of setting a new world record. However, as you noticed in the post, he clearly slowed down on last seven kilometers of the marathon. This might have been caused by the fact that his pacing strategy failed as you argued but I a€™m convinced that it was not the only reason. I am sure that he is a great marathon runner but at the same time think that public pressure and excitement over having a chance to break a world record could have made him under-perform, ultimately. The anxiety could have overwhelmed him in the final kilometers and maybe that was the reason why he ran so fast in the first part of the race. The expectations from fans could simply have overwhelmed him resulting in a significant difference between earlier intervals and the last ones. I think that he was physically prepared for the marathon, but the mental pressure and anxiety took over him so he was finally unable to break another world record in Berlin. What do you think? Could it be a possible reason?

BridgeportJoe said...

That's not to say it can't be broken, and maybe Sammy Wanjiru will be able to edge it down by a few seconds in Chicago, but the days where we expect to see up to a minute cleaved off a time are gone.

This is a good observation. Also, Geb's effort showed that once you get this fast, it's basically impossible to hold 2:04 pace by yourself. Remember, he had company most of the way in 2008, and the only other men to run sub-2:05 (I think) did so in a photo finish in the same race.

It's for this reason that I think we may see the WR drop close to 2:03:00 (and maybe slightly under it) when Bekele takes up the distance. If he can stick to Wanjiru and then kick, we'll have a situation where pacesetters take two great runners through maybe 30-35 km at WR pace, and then they'll race it in.

BridgeportJoe said...

By the way, good to see that you did decide to race me and 40,000 others in Chicago is a few weeks. What are you shooting for?

I just got back to Denver from high/low training (i.e., visiting relatives at sea level) where my trial long run (24+ at 6:58 going out and 6:51 coming in) suggest 2:55 is right on. I'd love to hit the NYC auto qualifying standard.

Unknown said...

Just a quick late comment on this. Geb had no chance of smashing this record because he step away from what works for him. He should have held firm 14:30 5k's for the first two splits, but instead tried to run a single pace with a quick out 5k in the beginning.

5k splits help one break down a race, but the half-marthon negative split is not as relevant as it might seem.

In 2008 Geb first 25% effort and last 25% effort was supported by 50% at what for him would be sub speed pace. He coasted in the middle after timing time in 2008, and here he had no coast, so he was bound to run out of gas at the end. Course 2:06 is a fantastic time.

BCC said...

Another poster- "I am sure that he is a great marathon runner"

Yes, I think that's a fair assessment!

Frankly, it's comforting to see an elite runner with a pace curve that's the same shape as normal people (e.g. slowdown after 20 miles).

I just ran my first 13.1 in 8 years this past weekend. Prior 2 halves had both been negative-split, awesomely paced runs at or below 6min pace (that's good for me); this time, I hit it hard in the middle and subsequently did a Haile over the last 5k. It's easy to criticize ex post facto, but it happens. By the time it's apparent that you've been too aggressive, it's too late- you've already overdrawn the account.

Personally, I prefer real head-head races rather than WR attempts, but hats off to Haile for gunning for it.

DrTim said...

We should start a guessing comp ... I punt for 2:03:50 and a new WR :)