If you feed the WKO4 model a valid MMP60 data point and it spits out that TTE is 45min @ mFTP of 234w then your model (and determination method for TTE) is broken. The fact that the mFTP and MMP are equal is just luck and you know that is a very very very deceptive argument. If you know as much as you claim to do then you should be ashamed for saying that.

You might want to review the annotated graph below showing how poor the WKO4 model does at fitting the data. Why is there a 12w gap at the 60m when the MMP data from 40-60m is so flat? Model error. Sure if you feed it more data it will get better, but if the model can't get the 40-60minute domain correct after feeding it a good 60MMP data point then it has issues.

Almost certainly a combination.

(1) I have not conscientiously attempted this duration over the last 3 mths.
(2) I am weak over this duration (compared to myself at other durations and my peers).
(3) I think my physiology that my strongest durations are 1-2.5 minutes combined with <30sec being my weakest makes it hard for models to fit even when I have datasets over multiple durations (like the last couple of years).

You must hate Critical Power as well. And power at MLSS. Neither have a specific time associated.

It's just a physiological reality.

I looked at my last decent season. The dW/dt on my MMP curve was about 0.5W/min between 40 and 60 minutes.

That means a difference of just 5W creates an uncertainty in duration of 10 minutes.

Yet on any given day my power for a long TT or intervals can readily be more variable than 5W.

Natural biological variability.

Anyone who claims to be able to nail down an MMP duration with great precision is full of shit.

---

http://www.cyclecoach.com
http://www.aerocoach.com.au

Of course the goal here is to model maximal performance.

That means the athlete needs to be prepared for maximal performance. Whether that's nutrition, hydration, residual fatigue, mental alertness, motivation, readiness for environmental conditions. illness, injury etc.

As a coach I'm sure you know better than anyone that prep for competition isn't just about a taper. And that performance is impacted by lots of factors. And yes, at durations out at 45m-1hr the drop off is minimal in well trained adult athletes.

A maximal performance model has to assume you're gonna get these things right.

This is very interesting.

So the only change in your MMP data was at-and-around the 3 minute mark? And this affects the slope of curve enough out at 30+ mins and beyond that it results in a 7W decrease in your predicted FTP?

Yes that is the interesting point about the hyperbolic nature of power and duration as expressed in the CP model:

P(t) = CP + (W' / t)

At 3 mins W' still has a big influence on your maximal power. Once we get out past 12 minutes we're mostly aerobic. So the new increased value for P(180) suggested a higher value for W' which necessarily means that the contribution from CP is less.

Or in simpler terms - if you can put out big watts for short durations you've probably got more fast-twitch which means fewer slow twitch muscles than previously thought. I'm over simplifying /massively/, but hopefully you get the point*.

With the extended CP model we also consider fatigue at longer durations, but this is fraught with difficulty since hardly anyone rides maximally beyond 1hr, so the model fit tends to dip off early reflecting the paucity of maximal data at longer durations. For elite marathon runners maximal data goes out to 2hrs+ and is generally flat to this point.

Mark

* a study was done to look at the relationship between W' and thigh muscle size it found a meaningful correlation: https://www.ncbi.nlm.nih.gov/pubmed/12111284

My role as a coach is about helping people attain goals, more often than not when things are not perfect.

Despite best intentions, it's very rare that everything is perfect, indeed it's a fragile mindset that requires it to be that way. e.g. for hour records, I try to prepare the athlete to break it on an ordinary day, not just on their best day, while at the same time trying to help ensure they do indeed have one of their best days.

But even if everything is perfect, natural biological variability will still mean no model is perfectly predictive and the TTE for any given power will still have quite a range.

As I explained earlier to Trev, that 5W range on my MMP represents approximately 1.6% of my power at around FTP/CP level yet results in a 17% range in MMP.

Put another way, for me to be able to nail down an FTP/CP MMP to the nearest second as Trev would wish, it requires I be able to nail down an MMP value to the nearest 8 milliwatts. Every. Single. Day.

I think we can all agree that's a ridiculous and a physiologically implausible expectation.

But apart from being ridiculous, it's unnecessary, since the purpose of such models is to be useful by providing insight into one's underlying characteristics, actionable intelligence if you like, and not to be perfectly predictive of MMP.

If a model needs to rely on perfect inputs in order to be useful, then by and large it's not a particularly useful model since perfection is rare. All one needs to understand is what inputs are required to make a model useful for the intended purpose and over what domain the model is valid.

---

http://www.cyclecoach.com
http://www.aerocoach.com.au

10W actually.

What Mark said....

I knew the predicted W' was low and, much as I would like to believe that I could sustain the previously calculated CP currently, I know that I can't (give me a few more months, though!).

I also knew I had not really gone for it from 1-3 minutes durations. I think the curve will change a bit more (not as much as that) as I get back to what I know I can do for 1.5-2.5 minutes.

On this I agree. Although it side-steps the issue at hand: WKO4 uses inappropriate data for model fitting.


Inputs for fitting a maximal power duration model are easy: maximal efforts.

Domain of validity for WKO4 model is unknown until it is has been fitted to appropriate data. If it is the Peronnet model* then it's reasonable to assume it covers 1s-1hr well.

Mark

* its components MAP(t) and AWC(t) have been shown to be *identical*

Knowing your FTP is useless if you don't know how long you can maintain it. A model which tells you that you can maintain FTP for 45 minutes when you can maintain it for over 60 minutes is not just useless but misleading.

There is no physiological measure or indicator that can nail down a TTE value at those longer durations with a high degree of certainty. To suggest otherwise is what would be misleading.

You clearly do not understand some basics about modelling human physiology. But more importantly, you most clearly do not understand the value of such models and their parameter estimates.

Here's a hint: it's not about predicting TTE with a high degree of certainty. If you think it is, then you are a fool.

---

http://www.cyclecoach.com
http://www.aerocoach.com.au

AFAIK, the WKO4 model has no MAP (or VO2max like) component/parameter, so I'm unclear as to how they could be deemed identical.

---

http://www.cyclecoach.com
http://www.aerocoach.com.au

If only you would shut up and stop misinforming people...alas, despite your repeated claims that you are going to do so, here you are in this thread, once again making patently false claims.

#3 isn't a significant factor.

Ok (I am no expert on modeling, it occurred to me that may be relevant but I have no problem to have been corrected).

1. How you collect/select data and how you model it are two different things.

2. While all sorts of data selection/filtering approaches are theoretically possible, you have to keep in mind A) bias (which is why an "envelope fit" is demonstrably inappropriate), and B) computational speed (which can be an issue even when fitting a model to just a couple hundred points using OLS).

That's not luck, that's by intent/design.

To wit: both 60 min* MMP and mFTP provide estimates of maximal physiological steady-state. As such, you would expect close agreement between them, and that is what you find.

*~40 km power, actually


You don't understand how OLS works, do you?

That's okay, clearly not many people here are.

Indeed, I would have to say that has been my biggest revelation in all of this, i.e., the fact that VO2max doesn't leave a clearly discernable "fingerprint" on the exercise intensity-duration relationship.* Given the importance ascribed to VO2max in setting the upper limit of aerobic energy production and hence determining endurance exercise performance, you would think it would.

*Which is not to say that you can't predict VO2max from the same, and with considerable accuracy... just that you need to know what you are doing.

Not being an expert, I would appreciate an explanation of why the WKO4 model doesn't seem to reflect what I have achieved on multiple occasions in the last 3mths from 5 to 60min-ish durations.

What sort of data is needed for the model to reflect what I am putting out? (ie can it be quantified by x number of efforts over y to z durations?). Does missing full on efforts between ~1-2.5 min affect it all?

I can post an updated curve since my 3ish min effort yesterday if needed (but won't get a chance for 24hrs or so).

(no criticism intended, just trying to understand).

"The map is not the territory." - A. Korzybski

Except that isn't the logic being employed in using OLS to fit a model to mean MAXIMAL power data ( a.k.a., "the extremes of the extremes").

Define what you mean by "doesn't reflect". Are you saying that the parameter estimates are too high or too low? If so, why do you believe that to be true, and how far off are they? (In a relative sense, i.e., as a percentage.)

(Regardless of your answers to the above questions, the answers spit out by WKO4 are the best you can expect based on your data, because the WKO4 model is the most accurate, precise, physiologically-correct, and statistically-valid model that has been developed. If it still isn't good enough for your purposes, all I can really say is "good luck.")

ETA: Although all points exert some influence on every parameter, some exert more on a particular parameter than others. It is possible to calculate the leverage that each exerts, but the answer varies somewhat depending on the exact shape of your curve. Thus, the most practical/pragmatic advise I can give is 1) look for durations at which your actual mean maximal power falls significantly below the fitted curve, then attempt to push those points up, and 2) use the fit statistics (i.e., parameter CVs and, to a lesser extent, SSE) and the guidelines I have provided to evaluate whether the fit is "good enough."

As bright as you are I would think you could count to ten:)

My mFTP has been reported as 244-246W for the last few months. In than time period, I have had the following powers on rides:

Aug 22: 1hr AP 252W, NP 255W (20min AP 262W)
Sep 17: 20min AP 265W, NP 265W
Sep 23: 20min AP 265W
Oct 8: 20min AP 258W
Oct 12: NP 271W
Oct 21: 1hr AP 248W, AP 249W (20min AP 255W)
Oct 26: NP 257W
Nov 9: NP 263W
Nov 10: 30min AP 256W, NP 258W

So there are numerous occasions where I have performed above the estimate of my mFTP. The 20min efforts in brackets after the 1hr efforts are to show the relatively small drop off so to add more meaning to the sub 1hr efforts shown. Where NP is noted without AP, it is because that was reflected by eg 30 sec to 1 min efforts with longer recovery (so AP over the duration is less relevant).

Aren't my examples above contrary to that? (please explain if this is not the case).

Please understand that I am a fan of WKO4 and find it very useful. I am in no way criticising you or the product.
I just don't understand what the model is showing based on these results.

May answer the question. (I have not yet tried my hardest at >1min to about 2.5min).

I will post the curve tomorrow.