Tim Cusick discusses the differences between the CP model and WKO’s PD curve in one of his videos and there are also a few articles online, but the main take-homes for me on this point are:
The premise of Skiba’s model was for it to be predictive. The WKO model is not intended to be predictive.
Coggan has ensured that while the model itself is “black box”, the outputs of that model are not. So it can be validated scientifically. This is discussed online as well in different places.
Anyway, why is it “rightly so” that the CP model is only accurate between 7 and 30 minutes? Isn’t the very definition of CP a power that one can sustain “indefinitely”
This is you. Not the model. The model is simply plotting a number based on the best efforts you put in. they either are your best efforts or they are not.
CP is not a construct like FTP where your coach can prescribe workout intensities based on zones. There’s probably a coach out there that uses critical power to prescribe workouts somehow, but I just use it to give me insights into the changes in my fitness.
anyone playing with smart segments - created in WKO5 and not using lap button? Wondering if an update has come out and that fixed a few things. There was a lot of opportunity to improve smart segments, I suspect that will take much longer to do.
I think WKO is supposed to be predictive, too. I mean, otherwise what’s the point of the software?
I can tell you your one-hour power; But first, I need you to go ride your bike really hard for one hour, come back, then tell me the results. Thank you–that will be $130.00.
Is this what they are saying? That you can scientifically validate an equation by looking only at the output of the equation?
It is “rightly so,” because the limitation is what makes the model meaningful. Theoretically, time at CP is indefinite because the relationship between your ability to produce power and the amount of time you can sustain that power is an asymptote. In real life, however, all athletes fatigue before infinity. Therefore, exhaustion at your CP will begin to set in at about 20 or 30 minutes of exercise. Of course, you can keep pedaling from there, but a whole lot of subjective variables enter the picture that the model does not (and cannot) account for. You can find tons of peer reviewed papers on how different variables affect work at CP, but CP itself does not try to account for those variables–what you ate for breakfast, how you slept, etc.
And that’s basically where I have a question: If you can move the “FTP” needle by including one and three second sprints in the model (I’m looking at you WKO), then isn’t something wrong with the model? Those efforts rely on completely different energy systems that are seemingly unrelated. I mean, I’d like to see the justification for how the model “accounts” for the output, but as you said, the model is a secret for marketing reasons. This seems to me to be really bad science.
On a lighter note: I am thrilled to report that according to WKO my mFTP is 5 watts higher today because I sprinted for three yellow lights during my last aerobic base ride. On the downside, according to WKO, those three efforts changed me from a “time trial phenotype” to a “sprinter phenotype”. This is serious problem; I have a 40k TT coming up and am concerned that I will under-perform if I’m a sprinter phenotype. Therefore, I better start stopping for those yellow lights so I can win my upcoming TT.
Yes - this is how science works. You make a prediction with a theory / model, and then design an experiment that tests the prediction. If the experimental data matches the prediction, then the theory / model is valid for that set of conditions. You keep doing this, and you get a greater confidence in the theory / model.
For an example of this, see Einstein’s Theory of Gravity Passes Toughest Test to Date. Scientists keep testing Einstein’s Theory of General Relativity looking for cases where it fails, and thus that there is unexplained phenomena that calls for a new and more complete theory of gravity.
This is actually one of the big problems with String Theory at the moment: it hasn’t produced any testable predictions, so it is purely theoretical construct.
Kind of. But it’s more like if Einstein wanted to sell his predictions for $130.00, so concealed the model as intellectual property, thus ensuring that no one can test whether it’s sound.
OK, disclosure request: are you one of the people that argued with Dr Coggan himself the topic of whether/why “moving the “FTP needle by including one and three second sprints in the model” de-validates the model, over on Time Trialling forums?
If so, I have nothing to add over and above his very detailed responses - which I believe to be correct.
If not, I would direct you to that thread (it picks up from around page 2). Lotsa science and maths and stuff
It appears when Coggan really gets into the nitty gritty, the detractors back off
I posted a question at the bottom of page 4 RE: the difference between the WKO4 PD model and the WKO5 PD model, which wasn’t answered. But, I found the answer here:
This is nothing new. You know what something is supposed to do, but you don’t know how it does it. Using known test cases or inputs known to generate specific output, you can validate the black box.
i really wish WKO would get off facebook as their platform for forum questions and onto a system like this (what TR is using). god damn impossible to search anything, especially as a new user when there is a 99% chance that questions have already been asked/answered. Facebook 1) is a shitty company, 2) is garbage with your personal data, and 3) isn’t conducive to a forum platform.
The CP concept was originally conceived as a model derived from a series of exhaustive, constant-load, exercise bouts (see Hill, DW. The critical power concept. A review. Sports Med 16: 237–254, 1993.). Skiba was the impulse-response model, aka PMC, based on heart rate.
Critical power and functional threshold power are two completely different thing. One is based on a linear fit of two points along your ride duration vs total kj expended and the other is a based on a physiological phenomenon, MLSS. Just because they sometimes align for some doesn’t mean they are the same thing.
dFRC is akin to W’ which has been part of Golden Cheetah for a number of years. Same with smart segments and Stamina (aka, Endurance Index). Training using PDC is nothing new but iLevel makes it easy. Only difference between the two is that WKO gives your the answer (based on your performance history) and you need to find it in Golden Cheetah (model selection and fit, although the generic modeling perimeters are close enough for “average person”).
I would gladly give up on Golden Cheetah If WKO can work as fast. Or, allow ride characteristics be edited in an acsii file format outside of the program (e.g. GC’s json). WKO is just too slow (and painful).
No. But this is interesting. They are arguing the exact point. Though, in working through Coggins’ “very detailed responses,” I’m not seeing him address how it’s sound to connect 1 second power to 60 minute power. He mentions that it reduces heteroscedacity, but that doesn’t have much to do with answering the question of how it’s a physiologically sound practice to connect super short efforts to relatively long ones.
He provided charts showing that there is a relationship between, albeit a weak one, between 1s power and mFTP. He then showed that the model in WKO4 is not incredibly sensitive to extreme changes at Pmax on its projection at mFTP.
The model isn’t trying to rebuild the things that are happening inside your body, it’s trying to take the data that is available and generate a descriptive model of what you can do as a cyclist.
I’m no scientist, but its seems pretty obvious to a layman that someone who over the course of the last 90 days worth of ride data put out a 1 sec 1500w effort probably has a higher FTP than someone who’s 1 sec max over the past 90 days of data was only 1000w. Actually, even if its the same rider and we’re looking at different 90 day periods, a change in ftp would be reasonable to assume. Totally ignoring that data would if its available would be silly.
thanks for the links, I’ll give a read sometime this week.
Models are fun… what I find humorous is that after a month of no riding and then one week with 6.5 hours of mostly aerobic endurance work outside/inside.
This morning WKO4 has my modeled FTP at 156W. On Sunday I averaged 148W for 2 hours and based on HR/feeling it was a solid aerobic endurance ride. FWIW I have my FTP set at 222W based on experience.
While my modeled FTP has dropped off a cliff, the good news is I have a 64 minute TTE yippee!
I think the things you “can do as a cyclist” are governed by “the things happening inside your body.”
In this case, I’m questioning whether the model tells me “what I can do as a cyclist,” because “the thing happening inside my body” when I generate one second power isn’t relevant to mFTP (the highest power I can maintain in a quasi-steady-state without fatiguing).
It’s relevant as an input to modeling your Pmax, which also plays a role in determining your anaerobic contribution to an effort, which has cascading effects as it relates to mFTP in the context of how WKO models what it has seen from you in order to describe your profile as a cyclist so it can generate guidance that correspond to your energy systems.
e.g. at a baseline if my anaerobic contribution for a given duration changes, the balance between FRC and mFTP shifts. Obviously changes over different durations will have different magnitudes of effects.
If Pmax goes from 1000-1500 and changes FRC, it’s logical that mFTP would change as a result of the curve shift to accommodate the new data.
Also, the PD Curve with error calculations shows you the range and the confidence variable, so you can see this info for yourself for how the answer changed.
Agreed - but if you’re looking at a data set rather than literally “inside your body” , the only way you can make predictions is by working with that data. You are correct that a full range of test efforts (plus some lab tests) would be ideal, but all mFTP is trying to do is make a prediction ABSENT that stuff. Its a tool, not some new sort of science.
You’re over thinking it. The fact is, if you go out and knock out 1500w for 1 second this afternoon and last month your max was 1200, there is a huge chance that your FTP is a little bigger now than it was last month. That kind of jump can only mean one of two things - a) you are a fitter cyclist now or b) you were sandbagging last month. Either way, the model is on to you. My understanding of mFTP is that is how its using those shorter power numbers and nothing really more than that.
yippee!