Regarding the power analysis of the Vuelta, would it be possible that the model was affected by the steepness of the climbs as well as their short lenght like you mentionned? the extremely slow speed we have seen could mean that very little energy is spent "fighting the wind" and more is spent going up, hence the apparent very high power being put out by riders. Thanks.
the power estimate model gets better the steeper the climb as drafting and CdA become less dominant sources of error. on the other hand the short length increases the error of timing and elevation change though to a lesser degree.
i don’t have enough data to know how well the DpVAM model handles very short climbs.
With the climbs done and Contador nearly sure of victory, its time to give some context to his winning effort.
Looking at the power duration plot its becomes evident just how well Contador controlled the race. On the stages were he lost time notice the minimal spread in power outputs of the GC contenders. Contrast that now to stage 16 and 20. On both of these stages the spread in the power outputs was large and Contador came out on top.
The pVAM for this race was is a bit difficult to interpret. The clear trend of highly positive values on short climbs and negative values on short climbs makes me wonder a bit if the model is simply not handling the short distances well.
To try to solve this question, consider Froomes 2013 TDF and 2014 Vuelta data.
From the power duration plot above, it looks like the short climbs from the Vuelta actually fall very nicely along the same trend from Froome’s 2013 TDF data. The 2 slow longer climbs from the Vuelta appear to be the outliers of the data set.
Taken together it is a reasonable generalization to say that aside from the 2 outliers, 2014 Contador and Froome were on par with 2014 Nibali and 2013 Froome. Like these performances, the 2014 Vuelta performances from Contador and Froome would likely have been frairly competitive with the 2002-2007 period of known doping.
Adding stage 14 and 15 into the analysis the power numbers reinforce the impression of a cautious race. Typically, the deeper that riders go trying to push the limits of their ability the more likely we are to see some erratic performances. With 5 finishing climbs completed, the estimates have a distinct lack of of this erraticness. Instead, the power estimates are falling incredibly neatly on a power duration curve with a large amount of clustering any given climb. This is not entirely surprising considering the definitive climb of the race (Ancares) does not come until stage 20. Waiting for stage 20 however, will surely leave some wondering what if.
As for the pVAM, the longer climbs according to the model have not been particularly impressive. Again, the short climbs are in a faded color as the model validity may be fairly question at these very short durations.
It seems CP models attempt to derive the CP value by extrapolating from a series of MMP values. The 'slope' of these MMP values can effect the derived CP. It is also suggested that some cyclists have a low CP but high W' or visa-versa. Could it be that CP value is actually 'constant' but the duration for which efforts can be sustained at the CP level is the variable? IE We should have a "CP Time"?
Yes there is clearly a Critical Time at play when it comes to fitting the Critical Power model. This issue is one of the major reasons for using the Veloclinic Plot.
From the power duration perspective, a more robust definition of Critical Power might be a threshold phenomena above which an identifiable W’ exists for a relatively broad range of power ie little penalty for intermittency within the Super Critical range.
For the explanation of the plot see: http://veloclinic.com/veloclinic-plot-w-cp-subtraction-plot/
For more on intermittency see: http://veloclinic.com/rethinking-intermittent-modelling/
With 3 finishing climbs done and analyzed the power duration curve is starting to take shape.
Overall, the performances have been more clustered than spread with the exception of Froome and Valverde falling off the pace on stage 9. Estimate wise, stage 11, the longest of the three was also the most pedestrian but road surface and tactics may have also been a larger than usual factor.
In terms of the pVAM model, only stage 11 is long enough to fall within the range that the model was intended for. I did leave the others stages in but I faded them out a bit as a reminder to not get to confident about the model result there.
as pointed out before the WKO4 model appears to be made up of the stringing together of 4 models that based on the distribution of their residuals function like a pmax model, a CP1-10 model, a CP3-30 model, and something that slopes down to cross the measured curve at the 45-60 minute range. the overlay of the CP models and the WkO4 model is previously illustrated in this post http://veloclinic.tumblr.com/post/72305824574/for-educational-purposes-only-post-contains-only
Just a quick one note here as this Vuelta so far is just following the curve. For the top riders on the day, the performances normalize out in the 6.2-6.3 naW/kg. Valverde and Froome were off the pace a touch at around 6.1 W/kg.
At only about 13 minutes of climb time, Altos Cumbres Verdes is a bit short to use the pVAM model. But it is still somewhat illustrative to take a look at how the performance estimates compare to the 2008-2013 reference range.
The raw nW/kg estimates came in between 6.4-6.5 nW/kg. Since this was a relatively low altitude climb the power comes down to 6.3-6.4 naW/kg when normalized for altitude for sake of the historical comparison.
From this first climb it does look like the favorites are on good (especially considering the heat) but no necessarily shocking form. We may just get the Froome Contador battle hoped at the TDF as both Froome and Contador likely to ride into form as the race progresses. Look for Quintana to bounce back on stage 9 as altitude will be a bit more of a factor.
The Vuelta Climbs by the numbers (as usual thanks to @ammattipyoraily and note that some of the numbers may be adjusted prior to the climb as he hunts down the best available climb data):
Across the board the Vuelta climbs are steep, mostly in the 7-10% range (note that for the climbs with a flat finish at the top this portion of the climb has been excluded for the analysis). The outlier climb is Stage 14 La Camperona at 15%.
The climbs are also relatively short mostly in the range of 4-800 vertical meters of climbing. The longest climb on tap is Stage 20 Puerto Ancares with over 1100 meters of climb.
Altitude ranges from the a 500 meter finish on Stage 18 Monte Castrove to just under 2000 meter on Stage 9 Valdelinares.
In general short steep climbs should diminish the effects of drafting and tactics so flat out efforts will be expected on most climbs.
Although the pVAM/DpVAM metrics are probably not entirely suited for the sub 20 minute climbs I used it nonetheless to get the basic gist of times and power outputs that can be expected.
In the blue is the VAM predicted from the 2008-2013 post-biopassport data set and in red is the prediction from the 2002-2007 doped data set. According the model, VAM is positively related to the natural log of gradient and negatively related to altitude and length of the climb. With the extreme range of climbs the model gives some fairly extreme predictions.
From the pVAM prediction its possible to project ascent times for the climbs. As you can see most of the climbs are below 20 minutes so take these predictions with grain of salt.
Lastly, with predicted times in hand the normalized power output expected can be calculated. Note the huge range of predictions here as short low altitude climbs will likely see big numbers well over 6 nW/kg. The longer or higher altitude climbs on Stages 9 and 20 will likely be in the 5.8-5.9 nW/kg range.
Re: MCV. You missed the point here. Dehydration (from alcohol) might cause changes in MCV (hyperosmolarity of plasma when fluid is lost, thus smaller red cells). EPO influence on MCV not discussed in the case.
right but follow along here for a bit
the heart of the case is the probability of doping given the ABP values estimated with Bayesian statistics.
for simplicity and because the actual ABP software is a bit of a black box consider walking through the argument in terms of Bayes theorem.
the probability of doping given the ABP results equals the prior probability of doping times the probability of the ABP results given doping divided by the probability of the ABP results.
P(doping/ABP) = P(doping)*P(ABP/doping) / P(ABP)
UKAD basically argued that the probability of doping was astronomically high because the probability of the results were astronomically low
the issue is that if you argue that MCV is good enough to rule out dehydration (which its not) then you can argue that its certainly good enough to rule out high dose EPO (which again it is not) AND since the results can not be explained by micro dose EPO then EPO can be ruled out all together.
that last part is a MAJOR issue for UKAD had JTL’s side been savvy enough to pick up on it.
since the entire argument revolves around a Bayesian estimate of probability as soon as UKAD introduced evidence of a normal MCV and also insisted on no dehydration then this evidence must be used to update the Bayesian estimate. (I understand this is not how the legal system works as you seem to get to coast through if your opponent is weak but it is how stats/science/reality works)
so with this evidence (if its good enough for UKAD then its good enough for this blog for the sake of discussion) we can populate the equation with some numbers to see how ruling out EPO changes the probability.
UKAD calculated P(doping/ABP) at 99.9999%
a conservative estimate of the prior probability of doping p(doping) might be 15%
plugging these values in
.999999 = .15*P(ABP/doping) / P(ABP)
6.6666 = P(ABP/doping) / P(ABP)
in terms of oxygen vector doping EPO likely accounts for the majority of doping and we’ll use 65% here as a reasonable example
taking EPO out of the equation means that your prior probability of doping is no longer 15% but now 5.25%
or solving for the probability of doping given the values
P(doping/ABP) = .0525 * 6.6666
P(doping/ABP) = .350
probability of JTL doping = 35%
JTL not guilty
the values were more likely a lab or physiological anomaly
now this is a bit of an oversimplification but had JTL’s side picked up on the above then this case would likely not have played out as such a clear slam dunk decision
on the other hand
if UKAD had said fine there could be dehydration they could have then gone on to point out that the MCV must have come down from an elevated value making high dose EPO all the more likely given the extremely low retic and a still high for even a dehydrated state in season Hgb concentration
for those wondering what MCV has to do with EPO
when EPO stimulates red cell production the new cells tend to be larger in size so that the mean corpuscular volume MCV tends to get elevated. after coming off of EPO the MCV drifts back down as the cells age. MCV has in the past been proposed as an indirect marker for EPO use. MCV can also be affected by blood storage conditions so could potentially have some utility in picking up transfusions as well.
One of the points that the UKAD side used was that a normal MCV (mean corpuscular volume) was evidence against the presence of severe dehydration.
The issue is that taking enough EPO to get your Hgb to 17.9 should also have increased the MCV (young red blood cells are larger and shrink as they age and to still have a Hgb of 17.9 at a point in time when the MCV has normalized would imply that the peak Hgb would have been high enough to even impress Bjarne Riis). So it would be expected that had the mechanism been EPO use then a potentially normal MCV could be expected in a dehydrated state as the MCV was reduced from a drug elevated level.
Alternatively, to make the conclusion they did regarding dehydration, they would need to take the stance that the method was more likely blood transfusion than high dose EPO and that the amount of transfusion was several units. Remember that part of the reason high dose EPO raises Hgb concentration is due to a diuretic effect that does not occur with transfusion. So to get a 90s era Hgb would take more relative doping with blood transfusion than EPO.
Overall, dehydration seems plausible on top of a recent cycle of high dose EPO.
“In fact, as much as 80 percent of healthcare data is unstructured, according to a recent Institute for Health Technology Transformation (iHT2) report. Matters are made worse when you couple in the fact that 50 percent or more of a patient’s health information typically not captured and available for view in an electronic format.”—medicine fights tooth and nail to stay in the dark ages it seems
Hi veloclinic. I was looking at my CP curve in Golden Cheetah this morning and I noticed that the widths of my power zones were not what I'd expect them to be. Specifically, zone 7 is quite wide, which is what you'd expect given the scale of the time axis. But then zone 6 is narrower than zone 5. So I wondered, does the relative width of my power zones tell me anything about how to tailor my training? Thanks!
it might just be the log scale causing some visual gremlins
on the other hand
after we are done with performance modelling study
we will be rewriting some of the power zone concepts
in terms of taking an applied aproach
ie what can actually be determined from field data
versus the current aproach
of trying to shoe horn the data into a priori physiological guruistic constructs
Is it such a bad thing that riders with "dodgy" passports (but no case opened against them) appear to be having difficulty getting a contract? If the UCI are gonna leave such a gaping hole in the passport that Armstrong and Horner drove their figures through, is it so bad that the teams get to make their own minds up looking at prospective riders' values?
this would be fine if teams had altruism to the sport as a priority
“We are the ones making the money and carrying the liability,” Patterson said. “The others don’t make any money. Nobody wants to watch them on TV. I don’t accept the argument that you have to have total socialism.”—
“I’ll reveal something important. Last year there was a rider, quite a famous rider, who we were considering for the team but we didn’t sign him because our experts and our doctors looked at his Biological Passport data and thought there was something wrong with it, with irregularities in there. But one of the other major teams in the peloton did sign him and the UCI was okay with it and now this rider is racing.”—
JV says the same thing happened with Thomas Dekker who was later banned based on positive EPO tests rather than his obviously doped biopassport.
Hopefully these examples are the exception rather than the rule. However, the opacity of the current system prevents anyone from actually knowing.
“Both, paradoxically, and also frustratingly because I know people want a definitive answer.”—Ross Tucker, highlighting that the biggest problem with rational analyses aren’t the methods, but the audience. (via cyclocosm)