Kenenisa in 26:17 Form Could Run 2:00:36? So Say MIT Guys

Homunculus wrote:

From the report:

Predictions for elite athletes allow for a more precise estimation of quotas and betting risk. For example, we predict that a fair race between Mo Farah and Usain Bolt is over 492m (374-594m with 95% prob), Chris Lemaitre and Adam Gemili have the calibre to run 43.5 (±1.3) and 43.2 (±1.3) resp. seconds over 400m and Kenenisa Bekele is capable at his best of a 2:00:36 marathon (±3.6 mins).

I'm not great at statistics, so i'm not sure if I'm reading this correctly, but does this mean a range for Bekele of 2:04:12 (perfectly reasonable) to 1:57:00 (completely insane). As I said, I don't have a great facility with statistics, but wouldn't a result like that undermine the model? I'm not sure the point of a model that suggests that Chris Lemaitre and Adam Gemili should be beating LaShawn Merritt and Kirani James.

Yes that's right. The confidence interval is + or - 3.6 minutes. It's difficult to tell the incremental value of this paper but it feels like very little.

One the one hand they are touting a very high fit for the model at statistical level, but on the other hand the predicted range for Bekele is 1:57-2:04, which at a practical level is not that helpful. They say one implication of their model is that it can help with pacing, but if Bekele were to use their model to set his pacing than and went out in 1:00, but was only actually capable of 2:04 (which is within the 95% confidence interval), he would most certainly crap the bed 30k into the race.

Could be true, but it can be difficult to model extremes due to lack of data. That makes the model flawed due to the assumptions / guesses about the data trends and difficult to extrapolate at the extremes. For example, how fast can the average (say mean of 20 runners) run a 10k if their 5K time is 11:37? Well, no one knows because there is no data on such 5k times. You might guess that there is some consistent curve or slope that continues down to that time, but you are only guessing. There could be some process that causes the correlation to change. There is very little data near the 5k world record. Even though Bekele ran it once, only a handful of people have even run within 10 secs of that time.

It is even possible that the ability to run a world record at 5k makes you slower at the marathon. Certainly there is a breakdown the model somewhere - no one thinks that Usain Bolt could run a world record marathon. A 5K world record may well reflect hyper developed lactate tolerance or some other ability that would not help you for more than a few percent of a marathon race. It is possible that the 5K world record holder could really only use his unique abilities in the last 2-3 miles of the race (approx 10% of the total distance).

The big question, besides the 26:25 in 2008 and the 12:46 in 2012 ..where is the next group that will take down any world's best...? I predict you'll never see another WR set on the track..NEVER.

ventolin^3 wrote:

rekrunner wrote:Ventolin's "two paramater" calculator would estimate 2:00:07 from the 5K and 10K times

I would never use that calculator for a M off 5k/10k or any other track combo

( albeit surprisingly good for Paula in '02 shape if you use 8'22 / 30'01 ! )

I did some tables for different track distances upto 10k

I can add track M

28'00 = 2"09'12

27'00 = 2"04'57

->

26'17 = 2"01'54

all run on a track

Vent, a quote from you in another thread:

"in a flat-out track 10k, fully rested, i'd expect wilson to go maybe 26'20/26'25 & kennster more like 26'45 off their M training"

Wilson recently ran 27:53 at Great Manchester Run for fourth. Do you still think based off of that performance that he could run 26'20/26'25 in a track/flat 10k just by training for the marathon (which he has been)?

If you do, I don't give you much credit for marathon to 10k predictions or vice versa.

Cue - "2/52 from London!!! w/ even pace and wabbit to 7k on track he'd have dusted the whole field!!!!" He was not beating Sambu that day, and Sambu is not running 26'20.

Uh, no.

26:17 is beyond mind-boggling but 2:00 is another planet fast. Even the top .1% have reason to celebrate when they break 1:00 in the half. Now, put 2 back-to-back!?

I get your point completely (even if it is somewhat exaggerated) -- I'm not going to defend any performance models (especially as I haven't looked at this one, and I might not ever) and I know too well that individual performances can vary and deviate widely from mathematical models.However, they built the model on 1.5 million individual performances over 10 events: 100 meters, 200 meters, 400 meters, 800 meters, 1,500 meters, the mile, 5 kilometers, 10 kilometers, the half-marathon (21 kilometers), and the full marathon of 42 kilometers. The model should incorporate effects like "fueling" because it should be represented in the input.When I said that the expected errors could be interesting, we see now that the estimate is 95% confidence interval for 2:00:36 +/- 3:36 (or something like that). This kind of precision should drive how much weight we put on the calculations.

Precious Roy wrote:

They are two completely different events. It is almost like saying Novak Djokovic beat Federer in straight sets, so he could shoot a 74 at the Masters. The marathon is all about fueling. Your flat out track speed does not mean anything in the marathon. Mo Farah crashed and burned in his Marathon debut. Tadese can't break 2:10. Kimetto and Kipsang are purely road racers with almost no track experience. But then you have Kipchoge who was a beast at 5k and never did many 10ks before moving up to the marathon. Thus, any attempt at finding an objective formula to predict a marathon performance is just silly. Each athlete is different. You just cannot tell who will excel at the marathon until they run one. It is a totally different event.

Um no, Kenenisa in 26:17 form could run 26:17. If you think that his 10000m world record will automatically translate to a marathon over 2 minutes faster than the current world record, you clearly know nothing about running. These conversions are so stupid it's actually kind of funny.

Star wrote:

If a 26:17 Bekele could run 2:00:36 doesn't that mean that the 26:22 Gebrselassie should have run 2:01?

Geb ran 26:22 in 1998 and then a marathon in 2:06:35 in 2002.

His marathon PR came in 2008 at 2:03:59 which was a world record at the time.

It's not like Geb didn't try his best to run a fast marathon.

I don't trust that algorithm when you had a very similar human with very similar PRs try to do this and not come close to this projection.

Except that there are - unfortunately, I'm guessing - 4 years between the 26:22 and the first marathon. As a huge Haile fan, it SURE would have been great to see what he could have done in the marathon in the late 90's...or at least the early 2000s, after starting in the late 90s.

On the one hand, his fastest marathon falling TEN years after his fastest 10K is awfully impressive. On the other - because of aging, obviously - there's the painful "What if?" associated with backing up his marathon career 2 or 3 years? Would that have gotten him 2:02-something?

But analyzing his road/late career is kind of depressing (DESPITE the obvious success), so I'll stop....

It just occurred to me: Is it a pattern that guys with little/no track credentials run great marathons immediately, and guys like Tergat and Haile have to "warm up to the event" ?

4 years after his 10K PR but just two years after a 10K Olympic Gold medal.

And the fact the he improved over 6 years from there suggests he was not going to run a fast marathon in 1998 if he lined up for one.

Not sub 2:04 for sure.

It would be odd to say that his best marathon fitness would have been in 1998 but his 2002 marathon fitness was much slower than his 2008 marathon fitness.

The Dingo wrote:Wilson recently ran 27:53 at Great Manchester Run for fourth. Do you still think based off of that performance that he could run 26'20/26'25 in a track/flat 10k just by training for the marathon (which he has been)?

If you do, I don't give you much credit for marathon to 10k predictions or vice versa

what part of wilson running a M just few weeks before manchester you not comprehend ???

Homunculus wrote:Chris Lemaitre and Adam Gemili have the calibre to run 43.5 (±1.3) and 43.2 (±1.3) resp. seconds over 400m

these research guys are talking utter drivel

lemaitre was good for 19.85-basic in his 19.80 ( + some wind )pb & gemili couda gone 19.85 basic on a warm windless day

those are their best possible times at their best event

how on earth do they get ~ 19.85 basic = 43.2 or 43.5 ???

someone look up iaaf tables for what a 19.85 = for 400

i have my own "mathematical" tables & that ->

19.85 = 44.20

i would love to see ventolin interact with people on a daily basis...

"you call this a latte?!? DRIVEL!"

Star wrote:

4 years after his 10K PR but just two years after a 10K Olympic Gold medal.

And the fact the he improved over 6 years from there suggests he was not going to run a fast marathon in 1998 if he lined up for one.

Not sub 2:04 for sure.

It would be odd to say that his best marathon fitness would have been in 1998 but his 2002 marathon fitness was much slower than his 2008 marathon fitness.

I think I made it pretty clear that I was assuming a similar or same "figuring out the marathon" ramp up time, but wondering if he would have ended up faster had he started sooner? Certainly not that he would have stepped off the track in '98 and ran a 2:02.

And age does matter, and obviously Haile was getting pretty old by the late 2000s (and maybe even older than advertised). I would find it pretty surprising if a similarly marathon-experienced Haile in the early/mid-2000s, wouldn't have been faster than the 2008 model.

Brings up another issue of age and performance ability. No one expects folks to set PRs on the track in their mid and late 30s (what could Haile of run for 10K - off 10K training - in 2005...or 2008?). But we're not surprised in the marathon (starting at least with Lopes, and thru Tergat and Haile). Does that make sense? Is age really eroding the factors contributing to 26 mile performance that much less than those for 3 or 6 mile performance? Or would those guys simply have run faster had they moved to the marathon sooner? It's easier for me to believe the latter...and arguably, the plethora of fast marathon times from young folks seems to support this. But two things:

1) Sucks that the 10K on the track is approaching, or arrived at, irrelevancy.

2) Again, why is it that some can run 2:04 the first or second time out in their early 20's, and Tergat - and perhaps even more so, Haile - both certainly as much or more talented than any of these guys, took so long to run similar (or now slower) times?

Of course! A marathon...on a track.

ventolin^3 wrote:

rekrunner wrote:Ventolin's "two paramater" calculator would estimate 2:00:07 from the 5K and 10K times

I would never use that calculator for a M off 5k/10k or any other track combo

( albeit surprisingly good for Paula in '02 shape if you use 8'22 / 30'01 ! )

I did some tables for different track distances upto 10k

I can add track M

28'00 = 2"09'12

27'00 = 2"04'57

->

26'17 = 2"01'54

all run on a track

i didn;t use paula's 14'31 in commonwealth games in '02

she jogged around for 600m before blasting last 4400m on cool day, likely windy

use something like ~ 14'20 - 14'24 for flat-out back then when wr was qo or wo or bo at 14'27

Wow didn't know mit had such a good running program.

Well, you know what they say... garbage in garage out

This. Why does everyone on these boards think just because an individual attends an institution of 'perceived' prestige, the individual in question is of equal prestige? Or what they're saying is credible? Or vice versa, eg a product of everyone's favorite clown college UO is so "clearly" inferior? College admission is more a game of strategy and having the support and guidance to say and do all the right things the college wants to hear. Jump through the hoops. Less frequently is it based on merit. Knowing the system is often half the battle.

Soooo.... grade the body of evidence presented by the paper (hint - rubrics exist to assessing quality of research). Don't slander by propagating loose correlations of credibility and and perceived prestige (or lack there of).

Nah. 2:00:36? I don't think so.

The predictions for the elite athletes are made using the

fastest section of the database