Let's talk about W'

First, double blind experiments are not possible here. The person doing the test, certainly knows what they are doing. Second, even with a large “n”, if the subjects studied do not correspond to your characteristics, the study is not relevant to you. Third, a sufficient number of controlled experiments on one individual can be equivalent to a large “n”. The problem of course with the last statement is that you are not the same after every experiment.

This is why understanding human biological is hard.

1 Like

Yes, it is possible to be double blinded. You just need to take this and what you are comparing it to and then not tell people which scheme you are using. Then you need to analyse the results without knowing which scheme was used.


can’t really argue

i guess that’s the “art” part :slight_smile:

In the vast majority of scientific studies of exercise, the exercisers knew what they were doing, or could figure out what was being tested. It is almost impossible to design a protocol where that does not happen.

If you read my response carefully you will see I did not say that blind analysis was not possible. Nonetheless, you need both sides to be blind to be double-blind.

1 Like

But this isn’t a study of exercise. This is a study of a particular formula and whether or not it can predict better than your chosen alternative whether someone fails a workout.

That’s easy to blind. You know the workout. You use W’ and your chosen comparator to predict whether failure occurs or not. There’s no reason to tell the person doing the workout whether one or both of the methods predicted failure.

We are not discussing whether your proposal is a good test. It probably is.

We are discussing whether it is double-blind or not. What exact instructions are you going to give the participants so they do not know what your are testing? If they know you are testing for some way of maximizing power, it is not double blind.

The title of this thread is “let’s talk about W’”. The person above mentioned predicting failure during a workout. Where on earth have you got this “maximising power” thing?

Because I asked you what instructions you would give to the participant. Participants do not think in terms of energy expenditure exhaustion, but in terms of activity exertion measured in power.

Cool. So moving back to the W’ question - @Holger1980 I’m interested in this as well, and am at some point going to work this out either from Golden Cheetah telling me it’s ‘value’ based on the data I give it, or some other source to enable me to put a number in to GC.

We might not get a coaches comment I guess, with that not being the metrics used in SUF but let’s see.

I like it that you can see that being ‘emptied’ on NH and FF. That sounds like you’ve got that set well. Would be interested in knowing how you did that !!

Over to the coaches or anyone who knows this stuff.

I started this thread because I was really interested in the SUF coaches’ opinion on this.
To your question how I got a seemingly valid value in Golden Cheetah. Well I just put in my Full Frontal numbers in the CP estimator tool in the software itself. :wink:
So it’s somehow self fulfilling prophecy that the numbers match Full Frontal but it’s also pretty good in matching the perceived exertion in Nine Hammers. So that made me think about the whole approach in the first place.

1 Like

Have you ever put your FF numbers into Xert?

No. I don’t use Xert and have no account for it.

I’m thinking W’ based off of 4DP values would be a tad lower than W’ based off of CP, assuming of course that you are doing maximal efforts at each duration within the last 90 days. NM and MAP should be close, but a fatigued 20 minute 4DP would not match CP20, and likely would not match CP60 (although it might be close). And the same could be said for AC != CP1min.

I do think there are too many assumptions being made in many of these models for them to accurately determine future performance. There was a discussion on the forum about WKO which Sir Neal Henderson chimed in on, where he elaborated on the limitations of power duration modeling for this purpose. Bearing in mind that his job often entails this very need, predicting an athlete’s performance on an hour record attempt for example.

If the UN Commission on Human Rights ever found out about those last two intervals in Nine Hammers, Sufferlandria would probably be facing international sanctions for cruel and unusual punishment!

1 Like

Here’s a talk about W’ (not saying I agree or disagree with the content, just sharing the words)


Thanks for that. That explains everything nicely.
My CP and W’ are the same. As soon as I go above CP I die in seconds lol.

Seriously - thanks for sharing.

1 Like

The insight in to 3mjn testing is really interesting - especially after doing a ramp first.
In particular ‘how’ to do a 3min test.
I’d love to know how that 3min test works as if I go flat out I’ll literally not be able to turn the pedals after 30-60s

I used to do 3min “max” (not evenly paced) tests because they were one of the Wattbike things back in the day (supposedly a good estimate of MMP, another TLA). I stopped doing them when they broke my lungs. There are other, “better”, ways to get “useful” numbers for a W’ styley model imho… although they take more than 3 minutes and are also deeply unpleasant :grin: . Or, for an easy life, you could try to reverse engineer your “numbers” from regular sessions, XERT styley (although beware GIGO).

GIGO - spot on.
Without a consistent set of stuff going in all done ‘appropriately’ things end up a mess.

3’ did my lungs in when I tried it. Actually sounded odd afterwards. One for people with stronger/healthier/pick your term lungs than me.

End of the day it’s just hard to find a test that is consistent.

4DP still the winner for me for now - still
GIGO but easier to assess how GIGO it is each time.

1 Like