59 +\- 17 min? Thought it was +\- 13 min?

Is 17 a typo?


"on average (as you've already read), about as long as it takes someone to complete a ~40 km TT, but with some variability between individuals.

Based on the WKO4 model? 59 +/- 13 min, as I've told you previously.

Note the close agreement between the two above...funny how that works out, huh? "
http://forum.slowtwitch.com/...s;so=DESC;t=mpfgsljm

nope, I just rechecked - the facebook said 17.

Mumm - ' approximately 60 minutes ' would seem to be xpanding - that's now 42 minutes to 76 minutes.

So is FTP approx 60 minutes, or 59 min +\- 13 min or 59min +\- 17 min, or 30 min to 60 min, or 30 min to 70 min, or for a long time?

What is FTP this week?

I was posting from memory. The SD is in fact 13 min, which makes the 95% confidence interval 34 to 84 min.

Yes, the word "approximately" is there for a reason.

Yes, it is now possible to estimate individual differences in TTE (albeit with less precision than the other model parameters), which could not be done before.

Yes, it is still true that you can estimate FTP from the power you can sustain during a longer (e.g., ~40 km, or ~1 h if you prefer) TT.

As a corollary, FTP still represents a power that can, on average​, be maintained for ~1 h.

Yes, I meant what I said, and I said what I meant.

No.

In addition to comparing mFTP to 95% of 20 min power for n = almost 200, I validated it against ~1 h TT power in a subset of a couple dozen individuals for whom I had such data. (Note that I excluded the TT file from the MMP dataset, to avoid any sort of auto-correlation). I have also presented data comparing mFTP to the NIRS breakpoint in a handful of subjects.

Finally, I have also validated Pmax and FRC against other accepted measures that reflect the same underlying physiological characteristics.

Now remind me again the efforts you have made to validate the approach for modeling the entire power-duration relationship that you have implemented in GoldenCheetah?

Once again you are confused. No one has ever claimed that mean maximal power data *accounts* for biological and technological variability. Rather, the issue is that your approach of fitting only the extremes of extremes ignores the fact that it does.

As for appealing to my own authority, it's not a logical fallacy, but simply a statement of fact: I have demonstrable expertise in the matters at hand, and you do not. People should keep that fact in mind when evaluating the accuracy of your claims.

You claim 1 hour power is not FTP, yet you use it to validate it?
You claim your model is exceptionally well validated, but nobody can verify this. You have significant monetary conflicts of interest, so I feel your claims cannot be accepted as such just because you say so.
Just an example: your opinion on running power meters has changed over the years, and in the mean time you have received money from a manufacturer. Your status and expertise are not enough to make these claims.

I'm probably being nieve but it would be nice to be able to engage in discussions in a thread for once (purely out of general interest) without being made to feel you slept with someone's else's sister at the end of it.

24 data points is not more than all studies combined.

Mark

What *is* your justification for using data that is so unreliable ?

So are you saying 34 minutes to 84 minutes is approximately an hour?

The fact that it is reliable enough that you can extract accurate and precise parameter estimates. However, that is only true if you fit all of the MMP data using OLS - if you fit just the extremes of the extremes the way you do, the parameter estimates are biased.

The OP and so many others over the last 4 years have posted data to show this is untrue.

Mark

Heh. No scientific study has ever used more that about a dozen points per individual. So, I have got them all beat there as well.*

*Note that that is true even if you assume, as you repeatedly and mistakenly do, that just because you set a personal best at N seconds, the power you generated for N-1 seconds can't also be the most you can produce.

Regardless, remind us again of how many efforts and individuals you used to validate the extended CP model in GoldenCheetah before you decided to implement it?

Indeed, while you are at it, maybe you can explain why you chose to implement it at all, since you believe that MMP data are too unreliable?

"If you stand for nothing, what will you fall for?"

You mean the OP whose mFTP was within 2.5% of another accepted estimate of FTP, i.e. 95% of 20 min power? That OP?

Bottom line: the model performs exactly as I have always stated, and as I have always intended. I would not have released it into the wild if it didn't meet my standards/requirements (just as I have performed lots of scientific studies that I have never wasted time trying to publish, because the results weren't sufficiently impactful).

The answer to your first question is no. However, ~1 h (note the '~') power does provide a valid estimate of FTP, and so provides a useful point of reference.

Your second statement is patently false: all one needs to test my claims is a copy of WKO4, and a bunch of data to feed into it. Knowing the precise structure of the model is neither necessary nor sufficient to validate it. Regardless, the question is why you or other critics haven't taken a stab at doing so? Don't understand how? Too lazy? Afraid that you will find out that I am right after all? Or maybe you would just prefer to sit on the sidelines and bitch, as opposed to getting up off your ass and trying to actually accomplish something? You tell me.

As for running power, I have previously stated that I underestimated its potential utility, especially for pacing purposes in longer events (where a number of people have reported considerable success). Also as I have said before, though, I believe that the primary benefit to the Stryd resides not in the power data, but in the other metrics it provides. (Ever notice how they moved from the combined heart rate/power meter chest strap back to a foot pod and started reporting leg spring stiffness only after they started paying me for my advice?)

Yes, the OP would be much better off just using 95% of 20 min power.
The model fit is so clearly skewed down by submaximal data.

The CP model outperforms yours here, and would be really interesting to see your 24 ~1hr TT datapoints applied to the CP model using only data 6 weeks prior to each TT (but not the actual TT) using data between 2 and 20 mins as I did for Morgan earlier in this thread.

Mark

And you claim that because...? (Note that the model provides much more than just an estimate of FTP, and unlike 95%of 20 min power, is equally accurate regardless of your resistance to fatigue during supra-FTP exercise).

As for the CP model, I am surprised you didn't just suggest that he calculate things using~1 and ~10 min efforts the way you used to do it. After all, if you want to flatter somebody by overestimating their true abilities, might as well go all the way, right?

I claim that because of the information provided by the OP regarding the *quality* of the MMP data and evaluating the actual data he provided.

But back to the OP. He will just have to perform longer efforts to find out what he is currently capable. Or switch to a different model.

Mark

IOW, you have performed an eyeball assessment based on your own preconceived notions, and reached a definitive conclusion despite lacking sufficient information to do so.

As for the OP, there is no evidence that the WKO4 model isn't working perfectly well for him. Again, in this context the point of modeling is not to predict performance, but 1) to extract valid estimates of parameters reflective of different aspects of physiology, and 2) provide a robust mathematical description of the exercise intensity-duration relationship that can be used for further calculations (e.g., adaptation scoring). Mere prediction isn't worth the intellectual and programming effort required to do things right.

I digitized the MMP data, and worked on it in R using multiple power duration models and examined the data. I posted the results earlier in the thread from a fit to the Classic CP model on the basis of the information provided by Morgan.

Mark

So what? That still doesn't demonstrate what you claim it does.

The CP model, for example, can be made to fit practically any data set, but will grossly overestimate at all durations much shorter or longer than you choose.

Other models may provide a better description over the entire range of durations, but nothing will provide a better fit (on average) than the WKO4 model.

Basically what your claiming is that you don't like the WKO4 model, therefore it must be wrong, and your evidence of that is that other, unvalidated models give different answers.

I'm not sure you understand my point, but frankly its not worth arguing.

If Morgan is still watching it would be worth performing a prolonged effort, lets say 45ins-1hr and see how close WKO4 is to the result.

Mark

Coming over Memorial day weekend at a stage race:
1) prologue (call it ~16-20 minute effort)*
2) road race up a mountain (there's the ~40-1hr effort), down the other side, through the valley of rollers
3) TT (22-35 minute effort)*
4) 4 corner industrial park crit.

That should provide some very solid and reliable data across a wide set of durations all at maximal intensity. The problem will be freshness for the effort (i.e., the TT will be after the previous 2 stages, the crit is the same day as the TT).

*the range is broad because I don't necessarily want to inform the competition.