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

[Alex Simmons/RST]

But if his math is mistaken and some of his assumptions misplaced, how does that invalidate Vayer's essential thesis that statistical analysis can be applied to athletic performance to reveal that which remains opaque to conventional drug testing: continued cheating in sport?

I didn't say a statistical analysis approach is not useful, but rather make the points:

(i) Why do analysis on possible/probable flawed calculated numbers, when just doing it on actual climbing times data removes the vast bulk of the errors/assumption? The Top 200 ADH times from that Finnish forum post is much more insightful if you ask me (and even with that there is debate over some of the times)

(ii) How does a statistical analysis of climbing times (or W/kg estimates thereof) possibly provide any insight into the vast bulk of riders who are not contesting the climb, but rather just finishing the stage?

And furthermore, how do his presumed errors indicate that his sole motive is to make money?

It doesn't, publishing does though, although you're right in that the authors rarely make much from such exercises.

Those who rock the boat are usually confined to the lower decks.

Well I can't subscribe to your entire theory as when there's a choice between a conspiracy and a screw-up, 99% of the time it'll be a screw-up. But you have no argument from me on this latter point. Have seen it first hand in a number of countries and sports.

Statistical analysis is the least assailable method of finding patterns of cheating in sport.

Statistical analysis on say baseball/cricket players or basketballers or footballers or track and field athletes is not the same, because for the most part, all players in these sports are seeking to perform at a high standard all the time (save cheating for match fixing).

In cycling though, only those for whom a best performance finish atop a mountain matters are going to put in best effort on a race deciding climb.

Such an approach therefore totally ignores the majority and hence as a diagnostic for tracking down dopers, is not all that helpful. We are already onto the leading contenders, so what extra insight is it adding?

Where the numbers are wrong, plug in more accurate numbers for an improved model and better results When used in conjunction with a serious testing regime, Vayer's method (also long endorsed by Greg Lemond) could help lead to a truly healthy sport.

I'm arguing that one should simply look at the numbers we know to be correct (i.e. the actual climbing times) or at least most likely to be the least inaccurate, and not bring in unnecessary error by making W/kg calculations.

Secondly, while the maths could be improved, the data required to improve the precision of estimates does not exist. The biggest flaw in presenting these W/kg values is not presenting them with error bars due to the unknowns and assumptions made, because when you do that, you then realise the futility of the exercise.

It seems to me that this or that error in math is a straw man here, and that only those who have a vested interest in maintaining the status quo could object to Vayer on that basis.

So if I like my maths and data to be correct, I support the status quo? or doping? What nonsense.

If you want to prove something, or change people's understanding, it helps to have your maths right. Why ruin your argument with sloppy maths, especially when the maths is not even required to make your point? i.e. just use climbing times. They are more than sufficient for the task.

I would argue that by introducing a large level of uncertainty by not having tightly known input assumptions and maths, it does more damage to "the cause" by providing wiggle room for those that seek it, and who could blame them for walking though that particular door?

 

[Ferminal]

(ii) How does a statistical analysis of climbing times (or W/kg estimates thereof) possibly provide any insight into the vast bulk of riders who are not contesting the climb, but rather just finishing the stage?

I agree with most of your points Alex, but what do you mean here? What is the benefit of knowing how "fast" the 35th placed rider went on every climb for the past 20 years? The benefit to us that is. It is hard enough looking at the top few riders and comparing them across the years, if you start looking further down it gets messier and messier because you no longer have the assumption of close to maximum effort. If you go even further down to the autobus the more accurate variable is actually the pure time gap or average speed given that on multiple mountain stages they are often dropped before the final, and the degree of effort varies.

I mean the individual rider might like to know they cracked 5.0 W/kg on a certain climb but they already have that data, I don't see how it benefits our "pub chat".

 

[Polyarmour]

I haven't read the report but it seems there is a great deal of uncertainty around the figures presented by Vayer. Rider weights are guessed as are bike weights, Cr's assumed, Cda's assumed, wind ignored, drafting ignored, start and finish points likely inconsistent, even course data would be variable? Yet despite plethora of inaccurate input data we have riders being labelled as suspicious on the strength of one ride that strayed on the wrong side of an imaginary line on a graph? Have I missed something here?

Pantani's figures are mutant though no argument

 

[Alex Simmons/RST]

I agree with most of your points Alex, but what do you mean here? What is the benefit of knowing how "fast" the 35th placed rider went on every climb for the past 20 years? The benefit to us that is.

There isn't any, which I thought was my point, but let me reaffirm that.

This approach to using climbing times (or rubbery W/kg estimates) as some form of modern day doping radar is flawed for a few reasons, but mostly because the only blips we can ever see on such a narrow range radar are those few we already know to be at the top. Everyone else is out of range this particular doper scope.

I mean the individual rider might like to know they cracked 5.0 W/kg on a certain climb but they already have that data, I don't see how it benefits our "pub chat".

It doesn't.

 

[Fearless Greg Lemond]

Statistical analysis is the least assailable method of finding patterns of cheating in sport. Where the numbers are wrong, plug in more accurate numbers for an improved model and better results When used in conjunction with a serious testing regime, Vayer's method (also long endorsed by Greg Lemond) could help lead to a truly healthy sport.

It seems to me that this or that error in math is a straw man here, and that only those who have a vested interest in maintaining the status quo could object to Vayer on that basis.

Good post Maxiton.

Statistics can be very usefull to analyze the past and present.
What is believable and what is unbelievable. Patterns.

Why are Indurains' performances in 1991 till 1993 ranked as believable and afterwards not? Was it because he upped the game after he saw someone like Ugromov come very close to him in the Giro? Or Zenon Jaskula? How come he destroyed everyone on the mountains after Giro 1994 whenever he wanted? How come he cracked in 1996 and subsequently hung his bike in the garage?

Or was it because he took so much time in the timetrials he didnt need to explode on the climbs? Given his and Bugno's demonstration of power on l'Alpe d'Huez 1991 I think its the last.

 

[Maxiton]

I didn't say a statistical analysis approach is not useful, but rather make the points:

(i) Why do analysis on possible/probable flawed calculated numbers, when just doing it on actual climbing times data removes the vast bulk of the errors/assumption? The Top 200 ADH times from that Finnish forum post is much more insightful if you ask me (and even with that there is debate over some of the times)

Precisely, to the bold. When your analysis depends exclusively on the exactitude of numbers, you will never be able to prove anything definitively. The doper will always be left with sufficient doubt, sufficient wiggle room, to claim you are all wet.

That's the beauty of what Vayer is doing. Statistical analysis normalizes the variables and models phenomena to look for patterns. The numbers don't need to be 100% accurate in order to create a model that will indicate these patterns, they just need to be as close as possible. Of course, better numbers mean more accurate patterns, so the modeling is dynamic.

(ii) How does a statistical analysis of climbing times (or W/kg estimates thereof) possibly provide any insight into the vast bulk of riders who are not contesting the climb, but rather just finishing the stage?

First, leaders don't ride up the cols alone (recall the Blue Train, or for that matter all the other trains). Also, while it is mainly the leaders we are concerned with here vis a vis cheating, when a leader tests positive - or, potentially, is shown by statistical analysis to be cheating - it invalidates the race for his entire team. So what's happening back in the pack is less germane to this analysis. Catch the leader, in other words, and maybe some lieutenants, and you have effectively caught the team (even if it only means that they are shamed and booted from the competition).

It doesn't, publishing does though, although you're right in that the authors rarely make much from such exercises.

Well I can't subscribe to your entire theory as when there's a choice between a conspiracy and a screw-up, 99% of the time it'll be a screw-up. But you have no argument from me on this latter point. Have seen it first hand in a number of countries and sports.

How about in the case of the UCI, where we have both conspiracy and screw up? Why choose?

With normal people and normal organizations, you're right, screw ups will out. SNAFU. But when we are talking people and organizations that are corrupt, even criminal, conspiracy is the order of the day. Corruption, by definition, requires conspiracy.

Now you might think the UCI is basically a wholesome operation, I don't know, a bit incompetent perhaps, led maybe by a big, dumb lummox but with its heart in the right place. But in any event I look at the UCI and see not just incompetence but years and years of pernicious corruption. And thus conspiracy.

Statistical analysis on say baseball/cricket players or basketballers or footballers or track and field athletes is not the same, because for the most part, all players in these sports are seeking to perform at a high standard all the time (save cheating for match fixing).

In cycling though, only those for whom a best performance finish atop a mountain matters are going to put in best effort on a race deciding climb.

Such an approach therefore totally ignores the majority and hence as a diagnostic for tracking down dopers, is not all that helpful. We are already onto the leading contenders, so what extra insight is it adding?

With current methods we are not really onto anything. With some exceptions (notable for their rarity) it's a few small fry who are caught while an entire team is allowed to dominate a GT, sometimes for years. With Vayer's method those leading riders and their teams would be called out (if they're cheating).

I'm arguing that one should simply look at the numbers we know to be correct (i.e. the actual climbing times) or at least most likely to be the least inaccurate, and not bring in unnecessary error by making W/kg calculations.

Secondly, while the maths could be improved, the data required to improve the precision of estimates does not exist. The biggest flaw in presenting these W/kg values is not presenting them with error bars due to the unknowns and assumptions made, because when you do that, you then realise the futility of the exercise.

See above. By the way, as I understand it, statistical analysis is exactly what the NSA is applying to all that data they are collecting from us all.

 

[Alex Simmons/RST]

First, leaders don't ride up the cols alone (recall the Blue Train, or for that matter all the other trains). Also, while it is mainly the leaders we are concerned with here vis a vis cheating, when a leader tests positive - or, potentially, is shown by statistical analysis to be cheating - it invalidates the race for his entire team. So what's happening back in the pack is less germane to this analysis. Catch the leader, in other words, and maybe some lieutenants, and you have effectively caught the team (even if it only means that they are shamed and booted from the competition).

Sure, catch the leaders. But we don't need stats analysis to know who the leaders are so they can be targeted for appropriate doping detection activity. Heck we don't even need to know the climb times for that, just the results sheet.

For the rest, they don't cause a blip on the radar, so it adds nothing.

And what happens back in the pack is very important. Ever had a rider miss out on opportunity due to a doper?

Fir
How about in the case of the UCI, where we have both conspiracy and screw up? Why choose?

I'm not.

 

[Maxiton]

Good post Maxiton.

Statistics can be very usefull to analyze the past and present.
What is believable and what is unbelievable. Patterns.

Why are Indurains' performances in 1991 till 1993 ranked as believable and afterwards not? Was it because he upped the game after he saw someone like Ugromov come very close to him in the Giro? Or Zenon Jaskula? How come he destroyed everyone on the mountains after Giro 1994 whenever he wanted? How come he cracked in 1996 and subsequently hung his bike in the garage?

Or was it because he took so much time in the timetrials he didnt need to explode on the climbs? Given his and Bugno's demonstration of power on l'Alpe d'Huez 1991 I think its the last.

Thanks. Maybe Indurain experimented from year to year with the amount of EPO he took. When he cracked, it could have been caused by guilt or fear catching up with him. Maybe he pumped so much EPO into his body he feared for his health. Maybe he wanted to stop while he was ahead. He's a very religious man, so I hear. Maybe his priest told him to stop. Maybe he'd just had enough. Who knows? Good for him that he never looked back.

 

[Alex Simmons/RST]

With Vayer's method those leading riders and their teams would be called out (if they're cheating).

So how does Vayer's (or anyone's) statistical analysis of climb data method tell us if a domestique / team sprinter / non-GC candidate or non-climber is potentially doping?

A: It can't. It can only tell us about the leading GC riders. We already know about them.

See above. By the way, as I understand it, statistical analysis is exactly what the NSA is applying to all that data they are collecting from us all.

NSA is a red herring.

 

[Maxiton]

Sure, catch the leaders. But we don't need stats analysis to know who the leaders are so they can be targeted for appropriate doping detection activity. Heck we don't even need to know the climb times for that, just the results sheet.

While Vayer's method could (and should) be used to target riders for closer controls, what he and Lemond are suggesting is to use it as a control. In other words, if the statistical analysis (or a power meter, presumably) shows you to be cheating, you should be sanctioned accordingly.

So how does Vayer's (or anyone's) statistical analysis of climb data method tell us if a domestique / team sprinter / non-GC candidate or non-climber is potentially doping?

A: It can't. It can only tell us about the leading GC riders. We already know about them.

See above. We might know about them, but it would be nice to catch the ones who are cheating. Could also presumably be applied to sprinters. As for the purely water carriers (any of those left?), I suppose we'll have to rely on conventional testing.

NSA is a red herring.

??

 

[Fearless Greg Lemond]

So how does Vayer's (or anyone's) statistical analysis of climb data method tell us if a domestique / team sprinter / non-GC candidate or non-climber is potentially doping?

A: It can't. It can only tell us about the leading GC riders. We already know about them.

Thats for you to find out? Look up the statistics on riders performances, put them into context and draw your conclusion, isnt that what scientists are supposed to do?

When the front end rides an x percentage faster, what does that do with the time limit for the sprinters?

edit:
why were for instance Bouwmans and Boyer crap after 1992?

 

[Alex Simmons/RST]

While Vayer's method could (and should) be used to target riders for closer controls, what he and Lemond are suggesting is to use it as a control. In other words, if the statistical analysis (or a power meter, presumably) shows you to be cheating, you should be sanctioned accordingly.

See above. We might know about them, but it would be nice to catch the ones who are cheating. Could also presumably be applied to sprinters. As for the purely water carriers (any of those left?), I suppose we'll have to rely on conventional testing.

A valid trend analysis of riders' or any individual's performance needs to be based on the premise that each data point is a measure of their maximal performance capability. Else the data is not relevant.

So, what proportion of riders are exhibiting their maximum performance potential on each climb?

??

You brought the NSA example into it, not me.

I said the example of the NSA is a red herring. How is that confusing?

 

[Alex Simmons/RST]

Thats for you to find out? Look up the statistics on riders performances, put them into context and draw your conclusion, isnt that what scientists are supposed to do?

i. I'm not the one suggesting this is a valid method for doping control, so I'm not sure why I'd be the one to demonstrate how or why it is.

ii. I'm not a scientist.

When the front end rides an x percentage faster, what does that do with the time limit for the sprinters?

It changes. So?

why were for instance Bouwmans and Boyer crap after 1992?

I suppose you'd better ask them.