# Using power curve to detect increases in FTP?

I use both TR and Zwift and I went to climb ADZ (Zwifts Alpe d’Huez) in under an hour and just about managed. It was an all out effort for me and I couldn’t have gone harder. I then looked at my power curve…

My question - surely my 60min PR is now the most accurate representation of my current FTP?

1 Like

Yes.

assuming you held steady power for the entire effort. For example if I went all out for ~60 minutes and the power curve for the Zwift workout looked like this:

My ftp is related to the first inflection point. Despite setting a 60 minute PR.

(Bad example, see Steve’s screenshot below)

So then you’d take 95% of the power number from 20min…at the first inflection? And set your training levels (Z2-4) with that number?

It probably is, but your FTP might be a little bit higher and time to exhaustion at FTP a little bit lower than an hour.

Key question to ask yourself: did you hit a state that felt like a line you could hold, but going just over would cause you to blow up?

That line is probably your FTP.

Post the power file and we can prob see it.

No. Poor example and misleading the way I wrote it.

1 Like

Visual example without axis so people don’t anchor:

2 Likes

Thanks a much better example than mine.

Here is an long all-out effort and the power curve:

And I set season PRs for durations >20 minutes.

Like Steve’s example above, on my ride’s power curve you can see power drop-off between 45 minutes and 1 hour. Using my mouse I can see it occurs at 52 minutes. The power at 52 minutes is my FTP.

Because your all-out effort was 52minutes long?! What if you had gone all-out for the entire 60min, would your PD curve be different and show an “inflection point” at 60min?

If your TTE is 60 mins, yes. If it is 52 mins, you’d still see the drop off at 52 mins because after that you’ll be dying and the power drops off.

Actually trying to go past your TTE with an FTP effort gives you the best power curve to base this on.

If you do the whole thing at too low a power, your TTE will increase a lot, which also gives you a hint where FTP is. (If you’d want to go crazy with this, you could do long steady efforts at different power levels, and plot them against TTE. There should be a pretty clear change in gradient at FTP.)

There is a danger of this thread heading down a rabbit-hole with everyone’s different views of what the definition of FTP is.

I have always felt that setting your FTP is an approximation and setting it to the nearest 10W is good enough to determine efforts in interval sessions.

To answer the original post, it would be useful to know what you had as your FTP before and how it compares to the 60 min effort. If you are talking about a 5W difference does it matter? If it is substantial, the question is how did you determine your FTP previously?

That’s precisely the thing I don’t understand. Why does power go down dramatically if you slightly exceed your “TTE (FTP)” (meaning the TTE at power equal to FTP)? If FTP has the characteristic that small changes in power around FTP correspond to big changes in TTE (above FTP you fatigue much faster, under FTP much slower) then small changes in TTE should correspond to very small changes in corresponding power. (The derivative of the function TTE(P) for P close to FTP is big so the derivative of the inverse function, whose graph is precisely the PD curve, is small.)

Haha, I’m sorry about annoying everybody again with my not-understanding FTP (or rather my real problem is “TTE at FTP”, because I think it is not a robust quantity to track). I should keep silent and just work with my old school 1h power.

Suggestion: go ride at your FTP until you can’t.

After a minute or so, try to ride sweetspot.

You’ll see.

On a pure math perspective, it’s understandable why you are coming from this angle, but in practice it’s not how it works, especially because the FTP inflection point inspection is best done on an aggregate power curve rather than one power file.

1 Like

Nope, start with SST and see how long you can go. I’ll last way longer (perhaps double). Small ΔP, huge Δt. In other words, the derivative to the PD curve stays small even “to the right of” FTP.

I was assuming this. PD curve only makes sense when it’s filled with lots of all-out efforts in a fresh state.

Edit: To add something more constructive. It would be interesting to see the PD curve of somebody who has done a decent amount of testing at powers slightly below FTP. I’d suspect the “inflection point” would be nonexistent.

I don’t understand what you are confused about then?

Why there is an inflection point in the PD curve at FTP (i.e. a small increase in time leads to way less power).

Ok, I’ll try to take a stab at this since there are few things that have been floating around.

In the context of one ride for any given maximal effort, by definition, going past the point of exhaustion should create a noticeable inflection point because you have reached exhaustion for that duration. E.G. a true maximal 5 minute effort should leave you totally empty and you aren’t going to keep riding to minute 6/7/8 at a slightly lower power. (hence why my graph posted above has a noticeable inflection point at 5 minutes since I specifically test that duration and don’t also test 4:45, 5:15, 5:30 etc)

The generated PD Curve across multiple rides specifically that WKO uses is a way of taking all of those maximal data points and curve fitting them to a known phenotype and extracting a set of parameters. The PD Curve FTP inflection point is partially a function of the scale of the lower axis, which is not linear (I don’t exactly know the timescale factor off hand)

So TLDR the inflection point is a useful artifact of the model and scaling that has evolved over time. I can’t explain to you the exact math reason since that’s not my strong suit and honestly Coggan made a lot of these types of decisions a long time ago and influenced others who then implemented power curve type products for their own products.

1 Like

Ok, we are on the same page. The inflection point is just a modelling artifact (*). This was my understanding so far (I’ve read a couple of papers that describe PD curve models similar to the one used in WKO, e.g. Puchowicz et al.). Given this I found it funny people were trying to interpret something physiological from this inflection point.

Edit: (*) in combination with a lack of data points for durations corresponding to powers under (suspected) FTP.

Isn’t it just a log scale? Usually they tale it off with something that is basically linear in log t.

I mean, there is something important physiologically going on from being above or below FTP.

But we’ve shaped the model to try and reflect the body and give us useful guidance.