Obviously, an infinite number of curves can pass through a single point.
However, it sounds like ICU actually uses two points: Pmax plus whatever it cherry picks. That individualizes things somewhat, but it is still a bit generic.
Too bad more programs don’t tell you the certainty of their predictions. I think that would be quite eye-opening for many people. For example, what good is eFTP if it is only accurate to +/- 30% (to just pick a number)?
@bbarrera has it exactly right. The curves are averages of many real curves (separate for male and female). So the one in the middle is the 50% percentile, then there is one for the 51st and so on. So for a given gender, power and duration there is only one closest curve. Intervals.icu interpolates between the the two closest curves actually but you get the idea.
Thats a good idea! I might even be able to adjust the minimum estimation threshold automatically and effectively give these athletes a lower eFTP and higher W’.
So are these reference curves all the same shape?
No. Similar but not the same.
Read this:
Here are a family of curves based on GoldenCheetah OpenData project:
As you would expect, the primary difference in shape of curves is below 200 seconds (~3 minutes). And Mark Liversedge found the following:
the average (male) athlete; a Critical Power of 261w, a W’ of 15.5kj and a Pmax of 1100w.
of course thats the average from users participating in the GC OpenData project.
Well lets see how I map onto those curves… In March I did some testing and plucking out a medium length max effort (March 10) I have 326W at 240 seconds (4-min), which puts me almost on the red 50% curve (see spreadsheet highlighted cell):
And PD predictions based on user data and that single 4-min max effort on March 10.
Based on RPE and HR my FTP was set at 234W at the time.
About 2 weeks later (March 22) I did a 255W 20-min max effort.
About 3 weeks later (March 31) I did a 243W for 52-min near max effort.
Here is WKO5 after the March 31st long (almost TTE) effort:
File that in the n=1 and an imperfect snapshot of (rising) fitness over 3 week timeframe.
Looks like the WKO4 standards.
So what’s the logic of assuming that just because you match one curve at a particular point, your curve shape must be similar? Isn’t the best estimate of the population from which you are drawn the average of all cyclists pooled together?
shocker. Get enough data and the curves look similar.
Just eyeballing, the curves tend to be linear in the (aerobically dominated) area above ~240 seconds. My fitness was rising from March 10 thru March 31, and I didn’t do enough testing around March 10th. So I’d say at least for March 2020, using the methodology of taking a single 3-5-min max effort and using that to estimate FTP gave me a result that was a little on the high side. If I’m not mistaken CP tends to be above FTP (correct me if wrong), so not surprising.
Why do you always ask questions without offering an opinion? Here is my opinion - sometimes empirical data and averages can be useful as a guideline 
Why don’t you try it and share how well the methodology works or doesn’t work? I’m just an average Joe cyclist without much experience, and well into the masters 55+ age group.
Yea, that would be cool too - I have not looked into how to separate W’ yet. In my case I’m around 70% on longer durations, low 80s at 5 mom and 95% for 1 min. So I think the variance in curves can go well beyond 3 mins, and I’d guess my W’ might be higher as when I start doing hard efforts, it regularly goes negative with the 20kJ default. I am certain my true 40-60 min from is not too low - might still be a few watts high, TBD soon.
For what it’s worth I buried myself yesterday and got a 20min PR that at least finally reached my eFTP using the 300 sec duration. No chance I could have held it for 40-60 mins though!
One other issue that probably comes up in all this is that best efforts are often not evenly paced in these shorter periods. Even that 20 min PR had a lot of surging out of saddle - don’t think I can match it seated.
In any case, while nobody likes to hear it, I wish I had done my first year or two on TR at a much lower indoor value - bad ftp estimate using outdoor test made it way too hard and I didn’t know any better, started failing workouts in build and gave up on structure the second I could ride outside again. Ramp test or eFTP has that same risk of setting up feeling of failure so any clue to warn someone of their “anaerobic gifts” is probably worthwhile 
'cause I don’t have access to oodles and oodles of data?
However, that doesn’t mean that somebody can’t think logically about a situation, and picture in their mind’s eye what a likely answer would be.
Yeah I like thought experiments and use them a lot at my day job. Heck while out shopping this afternoon I looked at some June testing data and that spreadsheet to test some ideas. FWIW consistently asking questions comes across as handing out homework assignments which can put people on the defensive. Anyways I hope we are past infinite curves passing through a single point 
Not everything is on pubmed. If you are curious about some of the models/algorithms that are out there then wander thru GoldenCheetah source code… I spent ~5 minutes on github looking at GoldenCheetah source code and pretty quickly found inline documentation that led me to that Excel spreadsheet. It was at the top of PowerProfile.cpp in this directory: GoldenCheetah/src/Metrics at master · GoldenCheetah/GoldenCheetah · GitHub Hope that helps if you are curious, even if you don’t know anything about C++ there is some interesting stuff to glean from skimming the source code.
FWIW, the eFTP intervals.icu reports for me is usually no more than 3 watts off my tested FTP.
Me too.
Intervals.icu reported a 6W increase to my FTP after a zwift race.
I had the bright idea to replace my TR Ftp with the EFTP from intervals.
I proceeded to get destroyed by Palisade.
Be careful.
