power as dopeometer

[acoggan]

I thought that the attached pic would be worth sharing with those who believe that you can estimate power based on climbing speeds with sufficient accuracy and precision to discern who is/isn't doping. I snagged it directly from Ross Tucker's twitterfeed, and while I don't know the source of the data or the precise model being used, he and others have frequently cited the CyclingPowerLabs ('CPL model') approach as being the one they feel is most reliable (which is not only ethically a bit questionable but also rather ironic, since that website merely implements our peer-reviewed model, i.e., they should be citing our paper). I'll leave it to others to draw their own conclusions, but all I can is, I feel sorry for anyone labeled a doper based on such questionable data...

 

[the sceptic]

performance is never proof of doping, unless speeds are down then it proves cycling is clean. But when the speeds go up again it means everyone has moved on to advanced training plans instead of doping.

 

[acoggan]

performance is never proof of doping

I think you should have stopped right there.

 

[IzzyStradlin]

OK Dr. Cog, I've taken a lot of good advice from your books and forum posts and whatnot over the past several years, but you and Hunter have gone totally off the rails over the past year or so.

Power modeling is not detecting dopers, nobody is claiming it is. It is simply providing a basis to compare climbing times, and therefore power outputs over the years. Ie, trying to compare racers who are riding the same climb in different races.

You know this, Hunter knows this. Claiming that people are trying to detect doping solely through these models is a strawman argument to disparage the work of some bright guys who don't happen to be certified coaches or physiologists.

One question remains, how do clean riders currently ride as fast as doped riders? Faster than historically clean(ish?) riders. Not a little faster, a lot. We have seen climbing times slow down due to improved drug testing, but now we've seen them surge again....how? Why?

 

[goggalor]

Please include a figure caption in your next draft.

 

[Ferminal]

Why are you talking about estimates when you don't believe even true data can be a "dopeometer" ?

 

[acoggan]

Power modeling is not detecting dopers, nobody is claiming it is.

Actually, many take it as such, including Tucker himself, who just a couple of years ago proposed that anti-doping authorities use the data for that very purpose.

 

[Netserk]

Actually, many take it as such, including Tucker himself, who just a couple of years ago proposed that anti-doping authorities use the data for that very purpose.

To sanction or target testing?

 

[acoggan]

Why are you talking about estimates when you don't believe even true data can be a "dopeometer" ?

Because there are two aspects to the issue:

1) the accuracy of estimated power output; and

2) what we truly know and/or don't know about the limits of human performance.

Because the two are really separate, somebody could agree w/ me that estimated power is too inaccurate, but disagree w/ me about whether you positively identify doping based on directly-measured values.

In any case, the graph I shared really only speaks to the 1st issue listed above.

 

[acoggan]

To sanction or target testing?

And as I pointed out several years ago, when Tucker 1st advanced the same argument: aren't stage winners, yellow jersey wearers, record setters, etc., already subject to the maximum scrutiny that the system can and/or does bring to bear? That is, what incremental benefit in terms of 'targeting' does estimating power output provide? I don't see any...

 

[Ferminal]

Well if you're talking about your random image, I'm amazed that the good doctor did so well simply by pulling numbers out of his orange juice. At least make the comparison to more thoughtful methods.

 

[Netserk]

And as I pointed out several years ago, when Tucker 1st advanced the same argument: aren't stage winners, yellow jersey wearers, record setters, etc., already subject to the maximum scrutiny that the system can and/or does bring to bear? That is, what incremental benefit in terms of 'targeting' does estimating power output provide? I don't see any...

Perhaps to target those that have a sudden change in performance, not necessarily those with the strongest performances?

 

[acoggan]

Please include a figure caption in your next draft.

1. Not my figure (which, BTW, is reasonably well-labeled already...the only thing really missing from the face of the figure is identification of the regression line and the line-of-identity).

2. Everything I know about the figure is already in my post, i.e., I couldn't write a better caption.

 

[acoggan]

Perhaps to target those that have a sudden change in performance, not necessarily those with the strongest performances?

Times and/or placements don't tell you enough already?

 

[acoggan]

Well if you're talking about your random image, I'm amazed that the good doctor did so well simply by pulling numbers out of his orange juice. At least make the comparison to more thoughtful methods.

As I explained, I don't know if those are Ferrari's estimates, or based on our model. At one point, though, the same correlation coefficient was cited on the CyclingPowerLabs website, which makes me think the estimated powers are actually derived from our model. For some reason, though, that webpage seems to have disappeared...

 

[Ferminal]

As I explained, I don't know if those are Ferrari's estimates, or based on our model. At one point, though, the same correlation coefficient was cited on the CyclingPowerLabs website, which makes me think the estimated powers are actually derived from our model. For some reason, though, that webpage seems to have disappeared...

Maybe you should go to the source of the image and read the bit where it says Ferrari calc.

 

[LaFlorecita]

performance is never proof of doping, unless speeds are down then it proves cycling is clean. But when the speeds go up again it means everyone has moved on to advanced training plans instead of doping.

Duh

 

[acoggan]

Maybe you should go to the source of the image and read the bit where it says Ferrari calc.

I was just off doing that (before I had only looked at the graph, not the original tweet), and you're right, he does state that it is based on Ferrari's method.

Still, there isn't that much difference between the various approaches*, so I stand by my original conclusion.

*EDIT: See the table in this post:

http://www.sportsscientists.com/2013/07/on-performance-analysis-common-sense.html

(Also, I find it rather interesting that both Tucker and Puchowicz seems to be backing away from their original stances, now trying to "sell" what they are doing as simply "performance analysis" - see http://veloclinic.tumblr.com/post/56209956033/after-thoughts.)

 

[Ferminal]

I was just off doing that (before I had only looked at the graph, not the original tweet), and you're right, he does state that it is based on Ferrari's method.

Still, there isn't that much difference between the various approaches, so I stand by my original conclusion.

It would be good to back that up with some data.

 

[acoggan]

It would be good to back that up with some data.

See the edit to my last post.

 

[spetsa]

I just realized that 3 weeks may not have been long enough

enough already

 

[Dr. Maserati]

Originally Posted by IzzyStradlin
Power modeling is not detecting dopers, nobody is claiming it is

Actually, many take it as such, including Tucker himself, who just a couple of years ago proposed that anti-doping authorities use the data for that very purpose.

I did a quick search and any interview Tucker gave he is exceedingly careful to show that analysis of performance alone is limited.

So, where and when did he "propose to use data for that very purpose"?

 

[acoggan]

*EDIT: See the table in this post:

http://www.sportsscientists.com/2013/07/on-performance-analysis-common-sense.html

BTW, I just followed this link from the above:

http://www.fillarifoorumi.fi/forum/...km-h-VAM-W-W-kg-etc-%29&p=2066034#post2066034

and have to say (having now seen the actual files from some of the prominent performances modeled) that in some cases even using our approach gives an answer that is wrong by ~10% (due to GIGO, not any limitation of the model).

 

[acoggan]

I did a quick search and any interview Tucker gave he is exceedingly careful to show that analysis of performance alone is limited.

So, where and when did he "propose to use data for that very purpose"?

Go read The Science of Sport blog from a couple of years ago.

 

[Dr. Maserati]

Go read The Science of Sport blog from a couple of years ago.

No, I wont.

But I will happily read what you link to and since you know where it is you can find it.