Good that you test in a consistent way. However, even if it's in TARWAPM, this is not a definitive method for determining FTP.


"Believe" is not good enough, until you try it you simply don't know! This is fundamental, taking mFTP as 95% of 20 minute MMP is simply a rule of thumb, and does not apply to every one.


First off, you would do best to take your question to the WKO4 group on Facebook. However a couple of comments here:
– FTP is defined as the highest average power a rider can maintain in a quasi-steady-state without fatiguing. It is not necessarily the highest average power that a rider can maintain for about 1 h, just that the two values are often similar. WKO4 is calculating this via the PD model based on your MMP data. The fact that the TTE is only 33:57 (more or less predicted time you can hold mFTP) indicates that you have an early drop off in the PD model. This might in turn be linked to you not making efforts longer than 20 minutes in testing/training. Indeed, if you have no real maximal efforts around 40–60 minutes long, this will necessarily "drag down" your PD curve at these time durations. Your 5 minute PB for the blow out effort may have also kicked the initial part of the PD curve up, resulting in the right hand side (and mFTP) going down a little, e.g. corresponding to anaerobic capacity making more contribution to 20 minute power than previously modeled. To make any more comments we really need a screenshot of the PD model and its fit to the MMP curve.
– In any case, your mFTP and sFTP are within 5% of each other, in keeping with the accuracy suggested by Andy Coggan.

How much data have you put into WKO4? a single ride file would be not enough. I'd guess that a weeks worth of data with no racing or maximal efforts would make it hard for WKO to compute an accurate power duration curve.

I trialed wko4 for a small bit and once I had it trained on my training+race data it was pretty accurate on both mFTP and TTE.

 

I know. If I could afford "definitive," then I would go to a lab and get a test done. I'm doing the best with what I have and because I have historical data to compare it to.


I took 95% of tested 20 minute power. This came out to 316 watts. I did not take 95% of modeled power at 20 minutes on the PD curve. Isn't the point of modelling so that we don't necessarily have to go out and do an optimal 60 minute test to get an "absolute" FTP?


1. I don't have Facebook.
2. 308 for ~30 minutes is way different than ~316 for an hour (which I'm 95% sure I can do). So, if I am interpreting your comment correctly, my mFTP is likely artificially low because it is modelling off of non-maximal efforts in the 40-60 minute range?
3. I don't have much anaerobic contribution at 12.5 FRC. Or is that not the right association with the 5 minute power PB? Either way, I don't think I could do more than 316 for an hour even if I omitted a 5 minute blowout effort.
4. I'll try and post a screenshot of my PD curve later today.
5. Good to know they're close enough in agreement. I just want to understand the relationship to modeled power and tested power a bit better.

To the above poster, I have nearly 3 years of data loaded into WKO4.

Thanks for the help!

Regarding testing for FTP, yes I had understood how you had calculated sFTP. What I meant was that taking FTP as 95% of a 20 minute test has (AFAIK) never been recommended by Andy Coggan himself, and who invented the concept. AFAIK it is one of Hunter Allen's contributions to TARWAPM. IMHO it is not a good FTP test. For instance, I have no anaerobic ability but going off my own data still sometimes my 20 minute MMP is 10% higher than my 50 minute MMP (which I do occasionally specifically test for to help confirm sFTP); perhaps this is from some "biasing" from my naturally high VO2max.

What I meant about the 5 minute blow out effort is that, because it was a PB, it will have raised the 5 minute part of your MMP curve, which will have a certain impact on your PD curve – perhaps even pushing the right hand side down/lowering sFTP.

RE #2, being 95% sure you can do something is in no way similar to actually proving you can do something (writing from experience here). My way of understanding is that this is also why the PD curve is modeled on actual performance data, rather than being predictive in any respect.

Almost certainly.

Incorrect.

1. (316-308)/316 x 100% = 2.5%.

2. TTE isn't estimated with the same precision as other model parameters.

3. FTP is not, and never has been, defined as 1.0000000............. h power.


Alternatively, 95% of 20 min power slightly (see point #1 above) overestimates your FTP (which it does half of the time).


That would be helpful (especially if you use my chart which gives all of the fit statistics...something that WKO4 is still the only program to provide).

Can you inspect the fit statistics and determine the type of performance data that would create a better fit and improve its description of the rider?

That is, if the fit is poor in the 5min area can you go out and do maximal efforts in the 3-8min range and improve the P-D curve quality?

Yes. There are multiple charts you can use, but I find it easiest to just plot the residuals and use those to identify the lowest-hanging fruit.

The quality of the MMP data is far more important than any fit statistics. It is not unusual to find that an MMP curve of 10,000 points only has 20 truly maximal efforts on it.

The data to look at is:
- how much of that MMP data is from truly maximal efforts
- how many discrete efforts is that MMP data comprised from

Or put simply - fitting to MMP data is inherently unreliable.

Mark

You two........ :-)

It's like a whirlwind that randomly touches down in various forums and discussions. Wattage to Timetrialing to Slowtwitch and perhaps other places that I am unaware.
Anyway I do learn some things and I just hope I am learning the right thing since you two do not agree much.

My point above is a general point about MMP data, it applies to GC just as much as WKO4 !

Yes, Mark's frequently patently false statements and non-sequiturs are indeed tedious.

Yet despite your claim, you have chosen to implement multiple options for doing so within GoldenCheetah, without ever assessing the accuracy and precision of the resultant parameter estimates. Why is that?

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

I have said the same thing on many, many occasions. But for the umpteenth time:

1. If you want to estimate "threshold power" follow a proven, repeatable and reliable test.

2. If you really must use MMP data then at least filter it for maximal efforts.

3. If you can't filter then accept that the data is unreliable and this will affect the results.

Mark

1. Done and dusted.

2. The issue is that it is impossible to determine what is or isn't a maximal effort. Indeed, even the approach you use to generate mean maximal power curves in GoldenCheetah (i.e., ignoring any subsequent modeling of such data) rejects perfectly-valid results. (Of course, that is because although you understand coding, you have no background in either physiology or statistics, and thus frequently make grade-school errors.)

3. In the words of W. Edwards Deming: "In god we trust, everyone else must bring data", and I have already provided more data showing that the approach taken in WKO4 is accurate than in all peer-reviewed articles of modeling combined. You, OTOH, continue to claim the approach doesn't work, while providing absolutely no data to back up your claims. So, who do you think people should believe?

You certainly can see if it is the maximal for the data provided, that's just math.
But this is where problems begin with MMP, and why a test protocol is the way to go.

Mark

Yet you even reject some (almost all, with your "envelope fit") of those values - go figure.


PPP: Training is testing and testing is training.

Your problem is you think approximately 60 minutes equals 30 to 60 minutes. So given your inability to deal with simple numbers how can anyone trust your mathematical model?

Which is correct, old FTP ( approximately 60 minutes) or new FTP, ( somewhere between 30 and 60 minutes) ?

I thought 'old ftp' was something (approximately) like a 40km TT (which could be anything from 45 mins upwards). It's never been an hour (although this was often incorrectly stated). In other words, the old definition was a variable duration as well and not far different to the current description.

Go back and read Coggan & Allen's book. Read Alex Simmons Blog.

Read the chapter, What is Functional Threshold Power in the book - I have the Kindle edition - the hard copy I used to put my front wheel on doing turbos sessions got lost - so I can't give the page number.


"FTP is the highest power that a rider can maintain in a quasi-steady state without fatiguing for approximately 1 hour."
http://alex-cycle.blogspot.co.uk/...ven-deadly-sins.html

http://alex-cycle.blogspot.co.uk/...s-testing-ftp-2.html

"FTP - maximal quasi steady state average power one can sustain for about an hour). "


FTP is simply defined as follows:

"FTP is the highest power that a rider can maintain in a quasi-steady state without fatiguing for approximately 1 hour."

Back then and as you full well know,mthere was no mention of 30 to 60 minutes. There was mention of a 40k TT being a good way to test for FTP, but don't confuse testing for FTP with a definition of FTP.

If you know someone who can do a 40 k or 25 mile TT in 30 or even 40 minutes please post evidence.

See also in Coggan's book, the chapter Determining FTP,

No 4 - One Hour Time Trial.
" Since by definition the best measure of performance is performance itself, the most direct estimate of your FTP will be obtained by simply doing a one hour time trial."

No mention in the book of FTP being a power one can maintain for 30 to 60 minutes. The definition changed.

I edited some data spikes out that I found. That dropped the mFTP down to 304 as you can see, but upped the FRC. I hope that is the chart you were requesting. Thanks for everyone's help! Bummed it is turning into a bit of a flamewar.

Few thoughts from my side.

1. The MMP data are not that bad, certainly much better than if you have spent the last 90 days doing various isopower intervals on a hometrainer.

2. The cut off in your maximal efforts at 20 minutes is obvious in the MMP curve, which suddenly drops from being above the PD model to below the PD model out to about 50 minutes. Similarly you may not have hit your maximum MMPs at the 1 to 3 minute marks.

3. How much difference will it make to the PD model if you test or train to "fill in" these "gaps", and would it change the sFTP estimate much? AC will answer this better than me but IME not necessarily that much. This is because the PD model is also being determined by the MMP data at shorter times and out beyond 1 h, where the fit looks pretty good. OTOH, the PD model is for sure indicating you can "do better" at 30 to 40 minutes than you have done in actual performance.

4. Indeed, again IME, that's the beauty of AC's PD model; despite what one or two other thread contributors have written with fairly decent MMP input data it really is pretty robust – and often uncannily accurate in a predictive sense!

5. While the algorithm is proprietary and hidden, the fitted process, resultant PD curve and residuals are all open and visible. This really helps one evaluate the result and what it means, and with ilevels, VO2 estimation etc it keeps becoming a more powerful resource. Further, the model does not depend on any user input apart from raw data. To me, these factors give it a significant advantage over a certain recent alternative training software where a) fitting is hidden to create a "magic number" and b) user intervention is required to train this "magic number".