Power Data Estimates for the climbing stages

[Tonton]

Re: Re:

So on average the fastest riders in this tour were 100 seconds (7%) slower than Pantani et al.
Ranging from 42 to 160 seconds slower.

Why is 410W suspicious?

It's not the 410 W per se that Vayer deems suspicious, but 410/70 = 5.86 Watts/kg on the LAST climb of a heavy mountain stage. 70 kg = weight of std cyclist.
This for a 30-40 min. effort.
See here
http://sportsscientists.com/wp-content/uploads/2014/07/2013-vs-2014-Tour-power-outputs.png
Cheers
PS : Ventoux should have been excluded from the average I think (wind)

I like his calculation method as described/detailed in "Not Normal", which I keep in that special folder with the Pinot studies . Some disagree with the accuracy, but it's proven to set a realistic bar, and from year to year compares apples to apples.

If only 4 riders were flagged from '10-'15 and 9 this year, I wonder: has something changed in '16? It seems like the bikes are being scanned seriously at the Tour, so I would discard mechanical doping. Is there a new product or cocktail in town? It would explain how some get sick, some over-perform 9as in respond well), and some look weak: are they on last year's stuff and it's not enough? Or on (more or less) nothing and they will be out of cycling in two years? Is a new era of doping beginning before our eyes?

 

[Tonton]

Addendum: maybe others figured out Sky's recipe, and the recipe proves not to fit everyone?

 

[Alex Simmons/RST]

Re:

Alex, if you had to guess, who has the better aero coefficient with full TT equipment, Froome or Dumoulin?

On current/recent form Dumoulin has better W/m^2 (he was a minute faster in stage 13 TT), but parsing out how much of that is down to CdA alone is hard to say. The eye makes for a poor wind tunnel.

 

[Alex Simmons/RST]

Re: Re:

So on average the fastest riders in this tour were 100 seconds (7%) slower than Pantani et al.
Ranging from 42 to 160 seconds slower.

Why is 410W suspicious?

It's not the 410 W per se that Vayer deems suspicious, but 410/70 = 5.86 Watts/kg on the LAST climb of a heavy mountain stage. 70 kg = weight of std cyclist.
This for a 30-40 min. effort.
See here
http://sportsscientists.com/wp-content/uploads/2014/07/2013-vs-2014-Tour-power-outputs.png
Cheers
PS : Ventoux should have been excluded from the average I think (wind)

I like his calculation method as described/detailed in "Not Normal", which I keep in that special folder with the Pinot studies . Some disagree with the accuracy, but it's proven to set a realistic bar, and from year to year compares apples to apples.

Vayer sometimes has trouble with basic maths. I struggle to take anything he says as anything more than publicity seeking click bait. We all know doping is and has been a big problem. A monkey can throw a dart at a board full of rider's names and have as good a chance of picking dopers as Vayer.

But looking at the table, all the top riders this year were substantially slower (on average 7%) than the Pantani et al crowd. Which tells us that (i) they were slower and (ii) absolutely nothing wrt their doping status.

The minimal time gaps on mountain stages tells us that the top rider's W/kg are all very close. It was the time trials that made the difference and is where the best aero prepared GC contender did the vast bulk of the damage to his opponents.

 

[Tonton]

I don't know of a method to separate the wheatdopers from the chaffclean wrt ITT .

I studied Vayer quite a bit, and after "Not Normal", he got on my nerves, try to remain relevant. Yet, he leaked the Froome watts on Ventoux, so I tend to perceive him as part of the solution vs. part of the problem.

Good point: grouped, watts can go up, vs. a solo attack when you take all the wind and so forth.

Having said that, Nibali was horrible until his resurrection and I had visions of Landis. Aru has been bad all year, pretty much. Quintana didn't look like Quintana at this TdF. was it the real Quintana and other riders were Pantani , yet Valverde was Valverde (give him a tylenol, he'll respond well ). TJVG was atrocious, Porte did well. Pinot left his guts on the road, Seb R. got a top-15. On and on and on...

I have a tough time figuring this out. Then I see values that seem quite high, which triggers a doping alert.

If it makes sense.

 

[Alex Simmons/RST]

This was the 3rd slowest TdF since 1998. Only 2000 (just) and 2007 were slower, on average.
Not that it means much given year to year variability.

 

[bikenrrd]

Moved this here from the "Random French Guys" thread as it seems more relevant here.

One thing I've never understood about Vayer's method (and something that leads me to believe it's ill-formed) is that he computes a single figure for wattage that he considers suspicious - i.e. 410W, as above.
Now, I understand that he computes a W/kg ratio and multiplies it by a standard weight to attain the watts, no matter how heavy a rider.
This is flawed in another way, but I really want to know why he doesn't use the temporal aspect of the data. All of the climbs are timed, so why not express the figure in W/kg/s i.e. watts per kilogram per second, aka Joules per kilogram?

Anyone who's used Strava or Training Peaks or worked with a coach knows the power distribution graph which gives you the estimated power output over a certain time period. A climb that lasts for 20 minutes can be climbed at a higher power than one that lasts 30 minutes, and so the time should be taken into account.

Before I get flamed - I think his method (for calculating power, not identifying dopers) has merit but I'd like to see more scientific rigour attached to it. Openly publishing the method, for a start, and an exploration of the sensitivity of the method to the values chosen for the parameters (rolling resistance being one).

 

[bikenrrd]

Re:

This was the 3rd slowest TdF since 1998. Only 2000 (just) and 2007 were slower, on average.
Not that it means much given year to year variability.

I've wondered whether it's possible to take into account amount of climbing when comparing Tour speeds across the years? Average gradient or some such could be used, but you'd have to retain the average speed as well.

 

[biker jk]

Re: Re:

This was the 3rd slowest TdF since 1998. Only 2000 (just) and 2007 were slower, on average.
Not that it means much given year to year variability.

I've wondered whether it's possible to take into account amount of climbing when comparing Tour speeds across the years? Average gradient or some such could be used, but you'd have to retain the average speed as well.

Don't forget taking into account tailwinds in the fast and/or hard years.

 

[vedrafjord]

Re:

Moved this here from the "Random French Guys" thread as it seems more relevant here.

One thing I've never understood about Vayer's method (and something that leads me to believe it's ill-formed) is that he computes a single figure for wattage that he considers suspicious - i.e. 410W, as above.
Now, I understand that he computes a W/kg ratio and multiplies it by a standard weight to attain the watts, no matter how heavy a rider.
This is flawed in another way, but I really want to know why he doesn't use the temporal aspect of the data. All of the climbs are timed, so why not express the figure in W/kg/s i.e. watts per kilogram per second, aka Joules per kilogram?

Anyone who's used Strava or Training Peaks or worked with a coach knows the power distribution graph which gives you the estimated power output over a certain time period. A climb that lasts for 20 minutes can be climbed at a higher power than one that lasts 30 minutes, and so the time should be taken into account.

Before I get flamed - I think his method (for calculating power, not identifying dopers) has merit but I'd like to see more scientific rigour attached to it. Openly publishing the method, for a start, and an exploration of the sensitivity of the method to the values chosen for the parameters (rolling resistance being one).

Yep, one good thing about veloclinic is he always includes the time element, e.g. this graph comparing Froome 2013 to Nibali 2014:

 

[Alex Simmons/RST]

For reference:

 

[carolina]

This year there were several flat stages where the peloton was very slow. Is it that common? I can't remember many editions where that happened.

 

[Le breton]

Re: Re:

....
This is flawed in another way, but I really want to know why he doesn't use the temporal aspect of the data.
....
Yep, one good thing about veloclinic is he always includes the time element, e.g. this graph comparing Froome 2013 to Nibali 2014:

Just because he does not show such graphs does not mean that Portoleau (the engineer behind Vayer) isn't aware of that factor as you can see in their 2015 magazine (21 something ?? can't remember the title).

 

[Netserk]

I'm not sure, but I think it's not just standardized for weight, but length of effort as well.

 

[vedrafjord]

Anyone have a link for Vayer's complete method? And are there any other estimates of individual climbs for this year's Tour?

(Time permitting) I'd like to put together something looking at trends in the GTs of the last 10+ years. In my day job I help make loss estimates (both initial and claims paid out, and after selling bits of risk on to third parties) for hurricanes, earthquakes, medical malpractice, terror, all kinds of hazards at all levels of grain from county level to worldwide, across several years etc. So we have some nifty tools to make nice visualisations of different levels of aggregation.

There are lots of interesting questions like fast vs slow years, Tour vs Giro vs Vuelta, individual outliers, how much does performance on the last mountain go down when there are multiple climbs before it, etc.

vetooo/ammattipyoraily has estimated w/kg on their Twitter a/c for the Tour going back to 1999, the Giro back to (I think) 2005, and a few Vueltas. There's no length/time information in those tables though. They also post in a Finnish-language forum e.g.

http://www.fillarifoorumi.fi/forum/showthread.php?38129-Ammattilaispy%F6r%E4ilij%F6iden-nousutietoja-(aika-km-h-VAM-W-W-kg-etc-)/page16&s=87d3929b950b5e586223af5b1965b1c1

but it'll be a slog pulling out times.

 

[adamfo]

Re: Re:

Highest watts 2016 Tdf : Bardet 444 W on Le Bettex

I saw this graph today in Vayer's article in Le Monde. Since 2010, only Wiggo, Froome, Quintana, Nibal had been above the 410W-suspicious mark, and this year there were nine suspicious riders - the ones in yellow on the graph. Vayer made it very clear that he believes Bardet doped.

Most of Vayer's articles are behind a paywall on Le monde. Given just how few people post in the comments section there I can't help thinking nobody reads his stuff or cares enough to post.
Many his articles are based on little more than conceit. If he starts saying some of the stuff he trots out about Froome, in connection French hero Bardet, my guess is the paper will quietly drop him.

 

[Alex Simmons/RST]

Updated chart, the 1998 value was revised down due to removing 149km from total distance since stage 17 was abandoned. Average speed is still a little higher (40.14km/h) but not much over what the official site reports (39.983km/h) but can't seem to work out why. It equates to about 14.4km missing from total distance or about 21 and a half minutes in total duration being unaccounted for. Prologue that year was 5.6 km and I've included that and time bonuses for stage wins by GC winner wouldn't account anything like that much.

edited to update chart using TdF online data archive reported average speeds

 

[Alex Simmons/RST]

and another way to view the data:

with residuals:

Further comments here:
http://alex-cycle.blogspot.com.au/2016/07/tdf-speed-trends-1947-2016-take-2.html

 

[Alex Simmons/RST]

And one more for fun, this time average speed v average stage distance:

 

[bikenrrd]

Thanks Alex, those are really interesting graphs and show that average speed is inversely correlated with average stage distance.

 

[Catwhoorg]

Very nice visualizations Alex

 

[Le breton]

Re:

Thanks Alex, those are really interesting graphs and show that average speed is inversely correlated with average stage distance.

Only if you use the word "inversely" in a loose sense, not the mathematical sense.

 

[Alex Simmons/RST]

Re: Re:

Thanks Alex, those are really interesting graphs and show that average speed is inversely correlated with average stage distance.

Only if you use the word "inversely" in a loose sense, not the mathematical sense.

There is an inverse or negative correlation which is quite apparent from the negative slope of the trend line. If however data is confined to a smaller time period (e.g. a decade), then the relationship is far less definitive.

 

[Le breton]

Re: Re:

Thanks Alex, those are really interesting graphs and show that average speed is inversely correlated with average stage distance.

Only if you use the word "inversely" in a loose sense, not the mathematical sense.

There is an inverse or negative correlation which is quite apparent from the negative slope of the trend line. If however data is confined to a smaller time period (e.g. a decade), then the relationship is far less definitive.

I agree.

What I meant was that an inverse relationship in this case would mean that the average speed would be proportional to 1/distance.

However you plotted speed against distance and drew a straight line across the data, not a parabola.

Alternatively you could have plotted speed vs 1/d.

 

[Alex Simmons/RST]

Re: Re:

Thanks Alex, those are really interesting graphs and show that average speed is inversely correlated with average stage distance.

Only if you use the word "inversely" in a loose sense, not the mathematical sense.

There is an inverse or negative correlation which is quite apparent from the negative slope of the trend line. If however data is confined to a smaller time period (e.g. a decade), then the relationship is far less definitive.

I agree.

What I meant was that an inverse relationship in this case would mean that the average speed would be proportional to 1/distance.

However you plotted speed against distance and drew a straight line across the data, not a parabola.

Alternatively you could have plotted speed vs 1/d.

OK, this is speed v 1/distance.

Sure looks like an inverse relationship to me.