"clean", "suspect", "miraculous" and "mutants"

[Le breton]

stats

When I read all this junk from people who took stats courses, I am glad I never did

 

[Ripper]

Thanks for the info and insight, and you are right, I don't know those climbs from personal experience, only from the power meter data my clients have provided and which validated my method of estimating W/kg, which covers most of the major French alpine climbs.

I certainly haven't suggested the weather changed in 1991, but who knows with global warming, perhaps can we blame that on Lance and Hein?

More seriously though, it is possible to to apply the maths to parse out from power meter data the likely impact of wind, and to show on what parts of a course it was negligible, beneficial and detrimental, and to what extent. You do need a little knowledge of the rider's morphological characteristics, and their power meter file from a quality power meter along with an accurate elevation profile, but it's certainly do-able via the technique of virtual elevation profiling.


Or it might indeed have a tendency to bias certain climbs in one particular direction as you pointed out, meaning the W/kg estimates for a given climb might tend towards one or other of under or over estimation.

But I totally agree, the data should really be looked in terms of the general trends over the longer term. Picking out one individual ride however will have a lower level of precision because of the uncertainty of the inputs - and that's really my point.

What is precisely known (we hope) are the rider's actual climb times, so why not simply chart those for each climb and avoid the precision problems when attempting to first estimate and then normalise to W/kg estimates? The trends will still be there if you plot each climb separately.

Not sure if it's being added or considered or already done, but a database of climbs, precise timing points, rider's times during official races, dates and comments on the conditions and race situation would be quite an interesting resource.

Interesting discussion.

And for the bolded - indeed it would!

 

[Galic Ho]

1/10

Thanks for the reminder to ride my bike more, read books in print more, and spend less time on the forums.

Fair enough. Was just giving you a quick refresher on your last paragraph.

But if you think I am wrong, I'd love to hear more about statistical analysis where there are 8 standard deviations. Seriously.

The rest of your post was spot on BTW. Last paragraph was just a bit too melodramatic, but I don't think that was your aim.

Seriously though...that stat talk you gave was wrong. Lemond is no doubt an outlier. A once in a generation rider. Someone who was born to ride. There are plenty of pro's with a VO2max in the low 80s and high 70s. There are only three I have ever heard of going over 90. Lemond, Hinault and I also heard talk Floyd hit 90. But that was just a rumour. JV also said his was really high, but he couldn't hit a high FTP for 60 mins at the end of 3 weeks in the mountains like those guys could. Your Lemond'ss and Hinault's are the guys outside the far right in the narrow margin of the bell curve.

Anybody's guess right now who naturally clean is that for this current lot of pros.

 

[Galic Ho]

When I read all this junk from people who took stats courses, I am glad I never did

It's actually really simple stuff. Relevant for psych, accounting, marketing and some math disciplines. It's pretty straight forward. Everything falls in a bell curve and most people fall within 1 standard deviation of the mean. The elite in any population are 3 SD's to the RIGHT of the mean. Very few make it there.

So this covers sports, athletic performances, academic performance, etc. Yes it can be boring but it's simple to understand. Just gotta know the basics and not mess Z and T tables up. Also gotta remember the basic rules...so many people break them it isn't funny, which are big NO NO's.

 

[Galic Ho]

Of course there is such a thing as 7 standard deviations from the mean in a normal distribution.

Not for total populations there ain't. But keep telling us that. If you go above 3 SD's the bands are so narrow it isn't funny because statistically, like I said, that leaves maybe if you're lucky 3% of a population. Ain't nobody human in any sport gonna be at the 7th SD, let alone 8th. Sure for income...but keep it relevant. This is human bodies, not economics. The scale and area of measurement are very limited.

Come on man, keep it real here for a second. VO2 max...I pretty much gave you the average. Name me one pro ATM who has one over 90, which is double the average males. Name one. Then name me one who can push out 85-90% of that at FTP for 40mins to an hour at the end of a 5 hour stage in the final week of a GT. You can't. Then tell me they're 7 SD's from the mean. They're not in terms of pure VO2max.

You're forgetting, like most do, this is about POPULATION. If we narrowed this even further among pro's, it'd been an even more of a stickler.

Anyway, I've said enough. Lemond's VO2max is UPPER UBER ELITE LIMIT. Only Bjorne Daehlie has one recorded higher and his sport uses both upper and lower body power. If Lemond was a cross country skier, his would have been as high. If you tested enough people, you'd find Lemond and Hinault fall just outside the 3rd SD. Most cycling pros fall just on or over 2 SD's. But if you want to kid yourself, they're 7 times better...by all means, laugh away.

 

[Eshnar]

Not for total populations there ain't. But keep telling us that. If you go above 3 SD's the bands are so narrow it isn't funny because statistically, like I said, that leaves maybe if you're lucky 3% of a population. Ain't nobody human in any sport gonna be at the 7th SD, let alone 8th. Sure for income...but keep it relevant. This is human bodies, not economics. The scale and area of measurement are very limited.

Come on man, keep it real here for a second. VO2 max...I pretty much gave you the average. Name me one pro ATM who has one over 90, which is double the average males. Name one. Then name me one who can push out 85-90% of that at FTP for 40mins to an hour at the end of a 5 hour stage in the final week of a GT. You can't. Then tell me they're 7 SD's from the mean. They're not in terms of pure VO2max.

You're forgetting, like most do, this is about POPULATION. If we narrowed this even further among pro's, it'd been an even more of a stickler.

Anyway, I've said enough. Lemond's VO2max is UPPER UBER ELITE LIMIT. Only Bjorne Daehlie has one recorded higher and his sport uses both upper and lower body power. If Lemond was a cross country skier, his would have been as high. If you tested enough people, you'd find Lemond and Hinault fall just outside the 3rd SD. Most cycling pros fall just on or over 2 SD's. But if you want to kid yourself, they're 7 times better...by all means, laugh away.

since we're talking about population, may I point out that the 3% of the world population is still an awful lot of people.
If it's true, as you said - and I have no reason to believe it's wrong - that there are only a handful of people to ever go beyond the 3SD mark, then you should consider the chance that this is not a Gaussian distribution, thus all this chat about the 3SD makes little sense.

 

[Le breton]

It's actually really simple stuff. Relevant for psych, accounting, marketing and some math disciplines. It's pretty straight forward. Everything falls in a bell curve and most people fall within 1 standard deviation of the mean. The elite in any population are 3 SD's to the RIGHT of the mean. Very few make it there.

So this covers sports, athletic performances, academic performance, etc. Yes it can be boring but it's simple to understand. Just gotta know the basics and not mess Z and T tables up. Also gotta remember the basic rules...so many people break them it isn't funny, which are big NO NO's.

Exactly what I was talking about. You think that everything can be described by a "normal" bell curve!!!
Of course not.
VO2 for example, which is obviously truncated on the low side.
Someone came up with another example : wealth.
Maybe if you took the log of the $/£ value of said variable it might look gaussian, but certainly not the $/£ value itself.

A very large number of phenomena are random in nature, therefore they can't be properly described by a gaussian distribution.

Many distributions look gaussian until you get to the tail, then you need to study that tail on its own, a problem often encountered in physics.

I am appalled reading that some people who had a stat course come out of it thinking that gaussian distributions is all there is to know, appalled that they think that there is nothing beyond 3 s.d.

 

[Le breton]

Not for total populations there ain't. But keep telling us that. If you go above 3 SD's the bands are so narrow it isn't funny because statistically, like I said, that leaves maybe if you're lucky 3% of a population. Ain't nobody human in any sport gonna be at the 7th SD, let alone 8th. Sure for income...but keep it relevant. This is human bodies, not economics. The scale and area of measurement are very limited.

Come on man, keep it real here for a second. VO2 max...I pretty much gave you the average. Name me one pro ATM who has one over 90, which is double the average males. Name one. Then name me one who can push out 85-90% of that at FTP for 40mins to an hour at the end of a 5 hour stage in the final week of a GT. You can't. Then tell me they're 7 SD's from the mean. They're not in terms of pure VO2max.

You're forgetting, like most do, this is about POPULATION. If we narrowed this even further among pro's, it'd been an even more of a stickler.

Anyway, I've said enough. Lemond's VO2max is UPPER UBER ELITE LIMIT. Only Bjorne Daehlie has one recorded higher and his sport uses both upper and lower body power. If Lemond was a cross country skier, his would have been as high. If you tested enough people, you'd find Lemond and Hinault fall just outside the 3rd SD. Most cycling pros fall just on or over 2 SD's. But if you want to kid yourself, they're 7 times better...by all means, laugh away.

FWIW
Polar has been studying fitness for a long time.

Their own index is a proxy for VO2 max

They obtained an average of 43 for the 20-29 y. males group
with a s.d of 7

If that is the case and the distribution were normal, then somebody with a VO2 max of 93 would be at (93-43)/7 = 7 standard deviations above the mean.

However it has to be taken with a grain of salt as they have not collated VO2 for millions of people.

Still, it shows that the very best really are far far beyond 3 s.d.
HOW TO INTERPRET POLAR OWNINDEX®
OwnIndex is equivalent to the maximal aerobic power, VO2max, in ml/min/kg. This indicates how many millilitres of oxygen your body is able to transport and use per each kilogram of your body weight in one minute. The maximal aerobic power, as any other fitness test result, is most meaningful when used in comparing individual values and changes. Norms, rather national, can be used to compare the fitness results to the average values of those with the same age and gender. Below an example of normal values presented as a mean (standard deviation) according to the age group (Fletcher et al. 1995).

VO2max ,ml/min/kg
Age, years Men Women
20-29 43(7) 36(7)
30-39 42(7) 34(6)
40-49 40(7) 32(6)
50-59 36(7) 29(5)
60-69 33(7) 27(5)
Individual OwnIndex can be compared to the population norms as follows: One standard deviation around the mean (half SD up and half down) represents "average fitness". E.g. for a 33-year-old woman any index between 31-37 (34-3 and 34+3) represents "average fitness" compared to other women of the same age. Values less than 31 are below the average and those higher than 37 are above the average.

For international use the fitness classification by Shvartz & Reibold (1990) presented in Table 1 is recommended.

MEN / MAXIMAL OXYGEN UPTAKE (VO2max, ml/kg/min)

AGE 1 2 3 4 5 6 7
20-24 <32 32-37 38-43 44-50 51-56 57-62 >62
25-29 <31 31-35 36-42 43-48 49-53 54-59 >59
30-34 <29 29-34 35-40 41-45 46-51 52-56 >56
35-39 <28 28-32 33-38 39-43 44-48 49-54 >54
40-44 <26 26-31 32-35 36-41 42-46 47-51 >51
45-49 <25 25-29 30-34 35-39 40-43 44-48 >48
50-54 <24 24-27 28-32 33-36 37-41 42-46 >46
55-59 <22 22-26 27-30 31-34 35-39 40-43 >43
60-65 <21 21-24 25-28 29-32 33-36 37-40 >40
WOMEN / MAXIMAL OXYGEN UPTAKE (VO2max, ml/kg/min)

AGE 1 2 3 4 5 6 7
20-24 <27 27-31 32-36 37-41 42-46 47-51 >51
25-29 <26 26-30 31-35 36-40 41-44 45-49 >49
30-34 <25 25-29 30-33 34-37 38-42 43-46 >46
35-39 <24 24-27 28-31 32-35 36-40 41-44 >44
40-44 <22 22-25 26-29 30-33 34-37 38-41 >41
45-49 <21 21-23 24-27 28-31 32-35 36-38 >38
50-54 <19 19-22 23-25 26-29 30-32 33-36 >36
55-59 <18 18-20 21-23 24-27 28-30 31-33 >33
60-65 <16 16-18 19-21 22-24 25-27 28-30 >30
In this classification, class 1 corresponds to "very poor", class 2 "poor", class 3 "fair", class 4 "average", class 5 "good", class 6 "very good" and class 7 "excellent" cardiovascular fitness compared to individuals of the same gender and age. In a population, 11 % of the people belong to classes 1-2 and 6-7, 22% in classes 3 and 5 and 34% in class 4. This corresponds to "gaussian distribution", because the classification has been developed in representative samples of individuals from different countries.

 

[IndianCyclist]

The arguments are going back and forth about 3SD. For quality, the target is 6 sigma though rarely achieved in practice.
I would like to know the data exactly. What is the average VO2 max of the population of 6 billion and what is the standard deviation?
I see from the net that the average is ~40 and 90% of the people are less than 50

 

[Le breton]

a bit of input from doc on veloclinic or @veloclinic on twit
http://veloclinic.tumblr.com/post/52412273102/altitude-6-watts-onna-boar-continued

Hello mew^3

The hemoglobin saturation vs altitude given in that veloclinic article and the other site it refers you to is not an accurate representation of VO2 max drop with altitude (or FTP of your "average" cyclist as a function of altitude).

 

[Merckx index]

One SD is about 16% above and 16% below the mean in a normal distribution. Two SDs is about 2.2%, 3 SDS about 0.13%, Seven SDs is about 0.0000000001 %, or 1 in ten billion. So it’s possible by chance that one or a very small number of people on earth could be seven SDs above the mean in some parameter. Six SDs is about one in ten million, or roughly fifty on earth. So would definitely expect some athletes that are 5-6 SDs above the mean in certain physiological parameters. I don't know what the mean or SD is for V02, but I would guess someone like Lemond is close to 6 SDs above it, even more.

Most purely physiological characteristics are distributed normally. But many qualities aren’t, particularly anything social like wealth, which someone mentioned. hat is generally distributed on a power law or scale free distribution, where the vast majority have a relatively small amount, and there is a long tail of very wealthy people. Wealth is distributed this way for the same reason internet links are—because the more wealth you have, the easier it is to acquire some more. The more links an internet site has, the more other sites want to link to it to increase their own traffic. This property, along with growth in the parameter, in this case number of sites, are the main two requirements for a scale-free distribution to form.

 

[Le breton]

Most purely physiological characteristics are distributed normally. But many qualities aren’t, particularly anything social like wealth, which someone mentioned. hat is generally distributed on a power law or scale free distribution, where the vast majority have a relatively small amount, and there is a long tail of very wealthy people. Wealth is distributed this way for the same reason internet links are—because the more wealth you have, the easier it is to acquire some more.

Wealth distr. more likely log-normal distribution really as I was hinting, akin to power law. Wealth is distributed that way just because people always want to take a %age profit on whatever it is they are selling, not a fee commensurate with their contribution.

Simple case : you sell a house, the agency fee will be proportional to the value of the house, not calculated according to their effort.

A variable might be modeled as log-normal if it can be thought of as the multiplicative product of many independent random variables each of which is positive. (wiki)

Since VO2 max seems to me (as a non physiologist) to fit the bill for being a multiplicative product of (many) several independent random variables, it might well be better described as a log-normal distribution.

 

[86TDFWinner]

LeMond I guess.


Lemond Greg
Voeckler Thomas
Hinault Bernard
Fignon Laurent
Evans Cadel
Moreau Christophe
Virenque Richard
Wiggins Bradley
Froome Chris
Valverde Alessandro
Jalabert Laurent
Schleck Andy
Vinokourov Alexandre
Landis Floyd
Basso Ivan
Armstrong Lance
Contador Alberto
Indurain Miguel
Ullrich Jan
Riis Bjarne
Pantani Marco

Does this mean Hinault and Evans

 

[86TDFWinner]

Lemond Greg

Maybe now this will FINALLY put to bed this "Lemond doped too" nonsense all the butthurt Wonderboy fans keep spewing?

Hinault Bernard

I still dont believe this. Just mo though

Indurain Miguel

WOW! No surprise there.

 

[Merckx index]

Many studies have shown that wealth distribution in the U.S. and many other countries does in fact follow a power law. See Barabasi’s work, for example. Again, the key factor is social: people interact, or form networks, with each other, and those who are most heavily linked are the most sought after to be linked to. That’s why monopolies are always a threat in business. A company that is highly successful will attract more and more other businesses, those that supply its raw materials, for example, and those that market and distribute its product, and in the extreme it will take over the entire market.

Wrt V02 max, one relevant factor that might mess up a normal distribution is that it can be raised by practice or training. Not everyone does this, of course, and the people with the highest natural or genetic V02s are most likely to train, and train hard, because they have the potential to reap benefits from it. So in this sense one could say there is a social component to V02 max that would cause its distribution to deviate from normal and Gaussian and perhaps take on some of the characteristics of scale-free. But most of anyone’s V02 is still determined genetically, and genetic characteristics are generally normally distributed.

Of course, any physiological characteristic like V02 is a result of numerous other physiological factors, such as multiple genes, and the multiple metabolic pathways their products act on. But the requirement of multiple factors does not prevent a characteristic from distributing normally. Height of males is distributed very normally, and that is definitely the product of multiple factors. These multiple factors may interact with one another in very complex ways, including networks that have scale-free properties, but the emergent characteristic is still normal. E.g., scale free properties have been described in the brain of humans and other mammals, both in the way neurons are connected, and in the activity that occurs in these networks. But IQ,which can be understood as emergent from these networks, is still distributed in a near normal way. Again, as a general rule, innate qualities are more or less normally distributed, whereas those that involve interaction with others take on more complex patterns such as scale-free.

In fact, even the presence of social components may not obscure a normal distribution. Height and weight have social components, in that they are enhanced by certain forms of nutrition. People who are raised in an environment of good nutrition are more likely to raise their children in like manner, so effects of nutrition should modify the normal distribution. Despite these effects though the normal distribution can be visualized even in the presence of those factors.

 

[LaFlorecita]

Maybe now this will FINALLY put to bed this "Lemond doped too" nonsense all the butthurt Wonderboy fans keep spewing?

No butthurt Wonderboy fan but I do think he doped no offense

 

[86TDFWinner]

No butthurt Wonderboy fan but I do think he doped no offense

Oh dear god!

 

[vrusimov]

I thought it was in French, or did I miss an English language version?

Oh it's in English...and I have to say it is quite a read. I stayed up all night doing exactly that and I was vacillating between laughter and absolute disgust. After reading Indurain's numbers, there can be no doubt that he was enhanced in 94 and 95...there is a lot of red ink on those years, with stupendous performances at Hautacam 94 [470 W normalized over 35 min], Les Deux Alpes in the 94 Giro [480 W for over 20 min], Guzet Niege in 95 Tour [470 W for close to 17 min] and Cauterets, a short effort but at a whopping 560W for about 11 minutes. Again, these are all normalized to a 70kg rider, so his real power outputs for the above data were firmly above 500W... that's 616W, or 7.7w/kg at Cauterets!!

Lots of mutant performances for Big Mig...It seems Padilla did his job well...

 

[The Hitch]

No butthurt Wonderboy fan but I do think he doped no offense

Lemond said he would ban riders for straight up blood doping /epo positive, but thinks its a little more complicated with riders who test positiive for small numbers of mickey mouse drugs (like for example i dont know, clenbuterol)

Do you reconsider your opinion?

 

[Le breton]

..........
E.g., scale free properties have been described in the brain of humans and other mammals, both in the way neurons are connected, and in the activity that occurs in these networks. But IQ,which can be understood as emergent from these networks, is still distributed in a near normal way. .

I disagree at times mildly, at times strongly with what you said, but what I left out is absolutely outrageous.

IQ, a totally artificial construct, would be distributed in a near normal way.

Is it possible to be more absurd?

But luckily this out of topic, a good excuse for me to get out of what might become an unpleasant discussion.

PS what I mean by IQ, a totally artificial construct, would be distributed in a near normal way.
is, it might not be clear, that IQ is a totally inappropriate measure of people's intelligence. Therefore imagining that because testers force the IQ distribution to be gaussian, jumping to the conclusion that people's intelligence is "normally distributed" would be totally absurd.
Maybe in fact you agree with that.

 

[Alex Simmons/RST]

That would be an interesting exercise, very time consuming though as you need to a lot of adjusting and "bootstrapping".

I use my own models, but it can be done at first pass using Aerolab in Golden Cheetah software, but requires some knowledge about appropriate inputs and an experienced eye with such data to know when it does/doesn't make sense.

Sounds like you might have accumulated enough data to write a book of your own on that topic, but you probably would need your clients to agree to it.

Nah, not even close. If I ever were to write a book (I won't since I have a Sydney mortgage to keep up and other more important things to do than pub chats like this), it wouldn't be about this.

 

[Alex Simmons/RST]

http://www.alternativeditions.com/not-normal/

Ah, thanks.

 

[IndianCyclist]

FWIW
Polar has been studying fitness for a long time.

Their own index is a proxy for VO2 max

They obtained an average of 43 for the 20-29 y. males group
with a s.d of 7

If that is the case and the distribution were normal, then somebody with a VO2 max of 93 would be at (93-43)/7 = 7 standard deviations above the mean.

However it has to be taken with a grain of salt as they have not collated VO2 for millions of people.

Still, it shows that the very best really are far far beyond 3 s.d.
HOW TO INTERPRET POLAR OWNINDEX®
OwnIndex is equivalent to the maximal aerobic power, VO2max, in ml/min/kg. This indicates how many millilitres of oxygen your body is able to transport and use per each kilogram of your body weight in one minute. The maximal aerobic power, as any other fitness test result, is most meaningful when used in comparing individual values and changes. Norms, rather national, can be used to compare the fitness results to the average values of those with the same age and gender. Below an example of normal values presented as a mean (standard deviation) according to the age group (Fletcher et al. 1995).

VO2max ,ml/min/kg
Age, years Men Women
20-29 43(7) 36(7)
30-39 42(7) 34(6)
40-49 40(7) 32(6)
50-59 36(7) 29(5)
60-69 33(7) 27(5)
Individual OwnIndex can be compared to the population norms as follows: One standard deviation around the mean (half SD up and half down) represents "average fitness". E.g. for a 33-year-old woman any index between 31-37 (34-3 and 34+3) represents "average fitness" compared to other women of the same age. Values less than 31 are below the average and those higher than 37 are above the average.

For international use the fitness classification by Shvartz & Reibold (1990) presented in Table 1 is recommended.

MEN / MAXIMAL OXYGEN UPTAKE (VO2max, ml/kg/min)

AGE 1 2 3 4 5 6 7
20-24 <32 32-37 38-43 44-50 51-56 57-62 >62
25-29 <31 31-35 36-42 43-48 49-53 54-59 >59
30-34 <29 29-34 35-40 41-45 46-51 52-56 >56
35-39 <28 28-32 33-38 39-43 44-48 49-54 >54
40-44 <26 26-31 32-35 36-41 42-46 47-51 >51
45-49 <25 25-29 30-34 35-39 40-43 44-48 >48
50-54 <24 24-27 28-32 33-36 37-41 42-46 >46
55-59 <22 22-26 27-30 31-34 35-39 40-43 >43
60-65 <21 21-24 25-28 29-32 33-36 37-40 >40
WOMEN / MAXIMAL OXYGEN UPTAKE (VO2max, ml/kg/min)

AGE 1 2 3 4 5 6 7
20-24 <27 27-31 32-36 37-41 42-46 47-51 >51
25-29 <26 26-30 31-35 36-40 41-44 45-49 >49
30-34 <25 25-29 30-33 34-37 38-42 43-46 >46
35-39 <24 24-27 28-31 32-35 36-40 41-44 >44
40-44 <22 22-25 26-29 30-33 34-37 38-41 >41
45-49 <21 21-23 24-27 28-31 32-35 36-38 >38
50-54 <19 19-22 23-25 26-29 30-32 33-36 >36
55-59 <18 18-20 21-23 24-27 28-30 31-33 >33
60-65 <16 16-18 19-21 22-24 25-27 28-30 >30
In this classification, class 1 corresponds to "very poor", class 2 "poor", class 3 "fair", class 4 "average", class 5 "good", class 6 "very good" and class 7 "excellent" cardiovascular fitness compared to individuals of the same gender and age. In a population, 11 % of the people belong to classes 1-2 and 6-7, 22% in classes 3 and 5 and 34% in class 4. This corresponds to "gaussian distribution", because the classification has been developed in representative samples of individuals from different countries.

Very good post. Thanks for the info. Obviously the VO2 max cannot be represented completely by a normal distribution but most likely a log normal or a weibull distribution as a minimum VO2 max is required to survive and there appears to be a cap at high level of ~100.

 

[Galic Ho]

Lemond said he would ban riders for straight up blood doping /epo positive, but thinks its a little more complicated with riders who test positiive for small numbers of mickey mouse drugs (like for example i dont know, clenbuterol)

Do you reconsider your opinion?

If the legal rules allowed and the UCI weren't dodgy as all hell, AC would have been done for blood doping. Ashenden had AC's expert pretty much admit that was the only explanation for Contador's parameters. The clenbuterol part was a joke.

 

[Fortyninefourteen]

No, I thought the riders were in order of ridiculousness. I dunno.

Anybody else beside me notice no mention of Sastre???