You’re just trying to stir up contentious posts here, aren’t you?
This reminded me of one of my favorite quotes in literature:
“Beware of the man who works hard to learn something, learns it, and finds himself no wiser than before … He is full of murderous resentment of people who are ignorant without having come by their ignorance the hard way.”
— Vonnegut
I think it also depends on the race format, e. g. crit or circuit races can be very peaky while hill climb TTs are steadier. I tend to produce much higher normalized power on these spikier races.
I do agree that race files are quite useful, especially when you go all-in.
And an ever-growing number of institutions in developing countries* that expect their faculty to publish, with little to no emphasis on the quality of the work.
*Although Spain isn’t a developing country, and a lot of recent #sportsscienceatitsfinest papers seem to have come from there the last few years.
Honestly, what a bad methods section. No info on the participants, and a test protocol that isn’t fully explained. (“A test protocol that includes a 20 min test”). Plus a research question that nobody cares about, and gives pretty unsurprising results. How did that get accepted?
Hence the hashtag.
(The mantra I am attempting to drum into my students is, “Ask good questions, attack with rigor”.)
Putting the SIGH in science. It’s frankly embarrassing to the journal (and editors and reviewers) that they accepted it. I imagine the authors actually originally included more detail and had to make cuts for “space,” but yeesh, at least put it in the supplementary materials.
And, as you well know, the ever-growing number of “journals” that will publish anything (often for a fee) while espousing their quick turnaround times.
My wife keeps us on the edge of bankruptcy by visiting plant stores. I try never to go. 
I’m trying to imagine what pumpkin-flavoured soap smells like. I can’t remember any pumpkin I ever ate having a particularly strong smell.
So we can quote it on this forum and have fun?! Because some people believe 60-min power is both your FTP and its .95 of your 20-min power?! I’m going with the first one!
News flash: your 60-min power as % of 20-min power depends on “stuff” and varies!
I don’t know the exact meaning of 95% Credibility Interval, but if I guess at the meaning, out of 120 participants, the 60-min power varied from .84 to .99 when you throw out the outliers. Is that right?
I’ve seen mine vary from about .90 to .94 and I’m in the recreationally trained (2.8 to 3.0W/kg) category.
Another news flash: your 60-min power and willingness to suffer is trainable!
Basically a measure of uncertainty of the estimated parameter. from the variability of data observed in the study. So when the dataset is large and there is little variability between the participants you get a more precise estimate, the other way around it may be more uncertain.
Obviously the originality of the study is low. the value being that they have data on relatively many participants.
The news value is that less trained cyclists are worse at holding sustained power compared to trained ones?
Also, predicting the 60min by 20min is definitely possible but it’s not a perfect prediction (model R^2).
Anyways, not much new for a lot of work I’m sure.
The news value, IMHO, is that cycling media has portrayed FTP as 60-min power, and that 95% of 20-min power will predict FTP / 60-min power. And this is a reference that helps dispel that portrayal.
Also IMHO, there are a lot of people that aren’t comfortable with fuzzy math, and can’t wrap their heads around things like power at FTP can vary from 30-70 minutes. Partly because of the media portrayal, and for those that get a little farther, partly because 100TSS is riding at your FTP for an hour.
Just my opinion.
And thanks I just sat down at my computer and looked up credibility interval… I’ve got some rusty probability and statistics chops from studying and applying information theory (circa 1980s and early 90s). Bayesian statistics, check.
Well it’s a reasonable estimate, as shown by the data.
Also consider that the untrained people don’t have as much experience riding extended periods up to an hour as the trained did. (TTE/stamina?) which is what we try to predict here.
Would have been fairer for that comparison to do a 6 months training intervention where everybody goes through base-build-speciality for TT and at the end you compare.
Otherwise it’s quite possible to predict 60min effort by a lower multiplier eg 0.90-0.925 for untrained ones. However, for training, reasonable that they can still pull off their training efforts at e.g. 8-15min intervals using the 0.95 multiplier (again TTE).
The amount of data TR has vs the number of athletes in this study is not even comparable. In the absence of the many needed well done studies (probably not going to happen), I’d trust a TR AI model.
I do love how TR pioneered applying AI to performance estimation, and I still pay attention to that number. However, TR dug themselves a bit of a hole by relating the estimated FTP number to an arbitrary ramp test. Alternatively, why not create an AI estimated power curve and base training plans on that? If anyone can do it, TR can. With all of their data, I’d bet that an AI model would have a lower error range than day-to-day testing variability.
And yet, it does, and with some degree of precision at that. It’s only when you dive into the details/insist on high (enough?) precision that things fall apart.
FWIW, I’m not a fan of the “big data” argument. Whatever you gain in terms of power by increasing the N, you give up (and then some) by the sloppiness inherent in the data. Give me a well-controlled/executed “proof-of-concept” experiment over a crappily-executed multi center clinical trial any day.
Just in case anyone hasn’t seen it, this is great:
Thank you for making my day. That was outta sight lol 