Could a fit Alan Webb break 3000M WR?

Renato Canova wrote:I can suppose that, in that period (July / August 2006), Shaheen could run 48"5 / 1'44"5 / 3'28" / 7'23" / 12'42" and 7'51" in steeple.

48.50/1'44.50 ->3'28.69 , 4'46.04 , 7'25.51 , 12'57.06 , 27'27.86

i must admit this doesn't look cherono's fingerprint

it actually looks more like ~ aouita of '85 ( but low-48/low-1'44 is better for him ), or possibly gus choge a couple of years ago when he ran high-1'44 & mid-3'29 ( he wasn't high-3'28 nor 12'56 anymore but it's just a ballpark suggestion for gus )

He could also run about 26'40" in 10000m

i'm incredulous any human currently coud have simultaneous

48.50/26'40 ->1'43.68 , 2'12.28 , 3'25.74 , 4'41.25 , 7'16.52 , 12'38.28

ventolin^3 wrote:

i'm incredulous any human currently coud have simultaneous

48.50/26'40 ->1'43.68 , 2'12.28 , 3'25.74 , 4'41.25 , 7'16.52 , 12'38.28

Maybe that is a sign that your calculator isn't as precisely honed as you would like it to be? Maybe it is a sign that using best guesses (even coaches' best guesses) isn't a good technique?

What is the confidence interval for your calculator?

eh ?

the only other snippet of an elite middle distance guy's 400 or 800 ability i have read was a few year's ago when geb before the '99 champs mentioned he had run a tt in 1'47 ( it may have been here, t&f news or iaaf forums )

he was in top shape in those years & 1'47 was more or less exactly as expected

if geb coud only go 1'47, i'm a little sceptical that cherono, a guy who is superb, but virtually 99% of sentient posters woud agree, woudn't be as fast as geb over any flat distance, coud go a mid-1'44

Mr. Obvious wrote:

Maybe it is a sign that using best guesses (even coaches' best guesses) isn't a good technique?

canova's guys often seem to have odd mix of speed/endurance

the one that didn't make any "sense" was kemboi who ran 26'30 & only had 13'00 that year

i coudn't see how that 26'30 was possible of 13'00, but then canova explained it himself & stated he reckoned kemboi was in big sub-12'50 shape in his 26'30

that made sense & a line of fit coud be found which was reasonable ( to nearest 0.25s for 400/800 )

52.50/1'49.75 ->3'33.86 , 4'50.07 , 7'25.53 , 12'44.43 , 26'28.23

doing it to 0.05s gives more precision :

52.45/1'49.65 ->3'33.67 , 4'49.82 , 7'25.15 , 12'43.80 , 26'26.98

52.45/1'49.60 ->3'33.49 , 4'49.53 , 7'24.61 , 12'42.66 , 26'24.08

maybe canova coud cast his eye on those ?

ventolin^3 wrote:

Can you answer the mathematical question? What is the confidence interval for your calculator? With two points of data, since that is what you use. Can it be adjusted to use more points?

it goes with 2 data points

take it or leave it

What is the confidence interval?

go read the site :

the ultimate times for other distances can be predicted

use your brain ( if you have one ) & work out what that implies for your "confidence interval"

Mr. Obvious wrote:

What is the confidence interval?

To admit the existence of a confidence interval is to admit the imprecision of the formula.

It's more convenient to assume any deviations from the expected/predicted time are due to inaccuracies in the input numbers or non-optimality of conditions/fitness on the particular day the athlete competed at the distance being predicted.

kartelite wrote:To admit the existence of a confidence interval is to admit the imprecision of the formula

no

if it says on the site :

"based on statistical analysis" i'm sure confidence intervals woud be mentioned

if no such mention is made, confidence intervals are not relevant

kartelite wrote:

Mr. Obvious wrote:

What is the confidence interval?

To admit the existence of a confidence interval is to admit the imprecision of the formula.

It's more convenient to assume any deviations from the expected/predicted time are due to inaccuracies in the input numbers or non-optimality of conditions/fitness on the particular day the athlete competed at the distance being predicted.

All formulas like this would have a confidence interval. That is really not up for debate. Kartelite, I believe you know this from your econ (and maybe other, I can't remember) background. It is basic statistics. Even I know it and my statistics is limited to one undergraduate course and a little real world experience. According to Malmo (who I respect and has demonstrated he has a very firm grasp of statistics) Ventolin^3 also has a very strong grasp of statistics.

I am just not sure why Ventolin^3 would not show the confidence interval? I mean, he is posting that the formula can meaningfully project from differences as little as .05s in the 400/800m. I am just asking for a little math and science to back that up.

Mr. Obvious, confidence intervals apply in statistical analysis when we're trying to predict the true value of a population parameter based on a sample of individual (and it's nice if they're IID, but of course that doesn't usually happen) observations.

A "formula" is a rule showing a relationship between one or more variables and constants. Something like the area of a circle, i.e. a = pi*r^2 is a formula, it doesn't allow for error. If there was some uncertainty over the value of 'r' then we could construct a "confidence interval" for 'a' given the probability distribution of 'r'. I think this is what you're referring to.

Ventolin is using a formula, i.e. fixed mathematical rule. To get a CI for the predicted value, you would need pdf's for both input values (then there's technically an intermediate step of calculating a joint pdf, not done by hand, obviously) and then you'd be able to use the formula (actually it would now be multi-dimensional, not planar) to calculate the range of predicted values.

Obviously, moving from a simple algebraic formula to a probabilistic model is not simple arithmetic but it's not really all that hard. The main issue would be deciding on the pdf's of the input variable. For example, instead of saying Bekele has 48.7 400m speed, we'd have to construct a pdf to best-estimate his 400m speed; say, normally distributed with mean= 48.7 and standard deviation of 0.4 - this would mean we have a 95% CI for his 400m speed of around (47.9, 49.5). Then we could take his 5k time, at his peak we know he could run at least 12:37, so construct a pdf that has that as an upper limit, and maybe we'd get something like (12:33.4, 12:31.1), then interpolate a 1500m 95% CI based off the 400m and 5k ranges.

Mr. Obvious wrote:

All formulas like this would have a confidence interval. That is really not up for debate. Kartelite, I believe you know this from your econ (and maybe other, I can't remember) background. It is basic statistics. Even I know it and my statistics is limited to one undergraduate course and a little real world experience. According to Malmo (who I respect and has demonstrated he has a very firm grasp of statistics) Ventolin^3 also has a very strong grasp of statistics.

Ventolin is a physician. He has a firm grasp of statistics. Katelite has a pHD in math. I d not have a firm grasp of statistics. What I do have is a firm appreciation for the mistaken assumptions that the higher math types are likly to make.

To clarify:

You would need more than the 95% CI's for the 400m and 5k times to get the 95% for the 1500m. That's where the joint distribution function comes into play.

I haven't been around for a while. Could Alan Webb break the 3000 record?

Based on what I've seen, and I admit that Renato would have way more information, but I would estimate that in the summer of 2006 that Shaheen could run about 49.8 for 400, and perhaps 3:30 also. I would be shocked, shocked, I tell ya, if he could have run 48.5, because if you combine that kind of speed with 12:48, I would guess you're looking at a guy who could have broken Komen's 3000 record, and although Shaheen could probably have run under 7:25, I don't see how he could have beaten 7:20. I will add that I enjoy Renato's posts, but sometimes a coach (no matter how good he is) is not the most objective appraiser of his athlete's abilities.

I actually don't have a phd in math (or anything for that matter) but I do have a couple masters degrees. Actually the reason I didn't finish my phd has a lot to do with the inability of many scientific types to not blindly follow their models, as malmo alluded to. So don't lump me in with them, I've been in that world but not of it.

MAPIV wrote:

Based on what I've seen, and I admit that Renato would have way more information, but I would estimate that in the summer of 2006 that Shaheen could run about 49.8 for 400, and perhaps 3:30 also. I would be shocked, shocked, I tell ya, if he could have run 48.5, because if you combine that kind of speed with 12:48, I would guess you're looking at a guy who could have broken Komen's 3000 record,

The problem with that logic is that Komen was a 3:29/12:39 guy. There has been only one human being in the history of the world with those gawdly credentials. Ya think that might be the reason he has the 3000m record indoors and out, plus the two fastest two mile times ever? YA THINK?

Commmon sense is telling me you don't have to be a detective to figure this one out.

Malmo, perhaps you're misunderstanding me. I don't see any disagreement between what I wrote and what you wrote. I really don't believe that Shaheen could have run 48.5, and I don't believe he could have run as fast as Komen.