Setting W' In GC 3.1

1,825 views
Skip to first unread message

JMcG4

unread,
Aug 28, 2014, 3:51:06 PM8/28/14
to golden-che...@googlegroups.com
In setting up my Athlete details, the W' default setting is 20000. Do you leave this, or if not, how do I come up with the W' value for me ?

Neil Pugh

unread,
Aug 28, 2014, 6:13:35 PM8/28/14
to golden-che...@googlegroups.com
From what I've learned so far I'd say use the CP / W' estimator (in one of the drop down menus) by inputting your 3 min and 20 min max efforts from the curve assuming those efforts are exhaustive. I'm no expert though so I could be wrong.

Hamish Ferguson

unread,
Aug 28, 2014, 7:23:54 PM8/28/14
to golden-che...@googlegroups.com
Look on the CP chart and in the box it will tell you W'. Then just enter it in the preferences. Hopefully the 40,000J limit will be fixed soon. 

Hamish

JMcG4

unread,
Aug 28, 2014, 9:06:41 PM8/28/14
to golden-che...@googlegroups.com
cheers, got it

Pete from AUS

unread,
Aug 28, 2014, 9:24:20 PM8/28/14
to golden-che...@googlegroups.com
My approach has been:
1. I'm more confirdent to have a CP 'fair value' so lock that in first
2. Try various W' such that my hardest ever rides don't go negative W'Bal and in fact realistically I would not expect to see below about 5Kj

Mark Liversedge

unread,
Aug 29, 2014, 5:28:33 AM8/29/14
to golden-che...@googlegroups.com
As I posted separately, you can get round the 40,000 limit by entering in kJ; i.e. if you want 55,000 just enter 55 and it will be scaled.

Mark


On Friday, 29 August 2014 00:23:54 UTC+1, Hamish Ferguson wrote:

Julian

unread,
Feb 22, 2015, 3:25:16 PM2/22/15
to golden-che...@googlegroups.com
Having some trouble with W' and hoping someone would take a look and offer some feedback.

The attached pic is of a workout I did yesterday.  CP is set correctly based off a 20min test to exhaustion.  W' is set based on the CP model chart.

Is is correct that W' should not be depleted unless riding over CP?  That seems odd that a 15min interval at 94% would have no effect on W'  At lowest W' is about 4,100.  So is my W' set correctly at 15,000?

Thanks in advance.
-J
gc.png

Andrea Sabba

unread,
Feb 22, 2015, 6:55:10 PM2/22/15
to golden-che...@googlegroups.com
Hi,
am I wrong or I've to continuosly set my correct values of CP and W', accordingly to my present fitness level? For me it's a mistake calculate these metric using my best historical data.

What do you think about it?

Andrea S.

Mark Liversedge

unread,
Feb 23, 2015, 3:36:02 AM2/23/15
to golden-che...@googlegroups.com
You are absolutely right.

CP and W' can vary over the course of a season, indeed, if you focus purely on aerobic work it is typical to see CP increase whilst W' reduces; this is a phenomena that happens to domestic, tour and cat x riders.

The CP history plot will show this, but o drill down it is better to look at he CP chart for short date ranges.
I look at my CP and model on the diary view since its easy to look at it month by month.

Mark

Andrea Sabba

unread,
Feb 24, 2015, 6:03:12 PM2/24/15
to golden-che...@googlegroups.com
Thanks Mark.
Another question, I made I workout of neuromuscolar power, see attached image. My CP and W' value, calculated using the last month data series, were CP=215 W'=13,3 kJ.

During the workout, the lowest W' value was -8,9 kJ, the question is: should I set my W' value to 13,3+8,9=22,2 kJ ?


Thanks in advance

Andrea

GC.jpg

Bryan Vasek

unread,
Mar 15, 2015, 2:35:38 PM3/15/15
to golden-che...@googlegroups.com
Andrea, Mark,

I have a similar question regarding a negative W'Bal after hard efforts.  I assume this means my W' input is incorrect although is accurate based on the 3 and 20 min efforts input in the estimator.  Assuming i understand W' correctly, you couldn't physically go below 0 without bonking completely.  

thanks

Nigel Van de Velde

unread,
Mar 21, 2015, 1:02:30 PM3/21/15
to golden-che...@googlegroups.com
Hi,

Back in December I started doing some 20minute FTP tests and analyzing them in GoldenCheetah. Today I did my 4th test and I noticed no further increase in FTP. Instead there was a big increase in W' (from 17600 J to 22400 J). This was somewhat unexpected based on the training I did last month (Z4 efforts, low amount of Z5+).

Could this 25% increase in W' be realistic, or would it be better to lower W' and increase my FTP value a bit? The FTP values I get from the extended model however are quite consistant with the 95% rule, so I am hesitant to change that value... 

Any help or thoughts and opinions are appreciated.

Regards,

Nigel

Nigel Van de Velde

unread,
Mar 21, 2015, 5:51:35 PM3/21/15
to golden-che...@googlegroups.com
Edit: I just fiddled with the numbers a little bit and I get them more "logical" when I set the critical power as 0,95 * 20min Peak Power and adjust the W' so I get to zero at the end of my FTP test. I think I'll stick to that method from now on.

Could it be that the extended model for the CP chart should only be used once you have enough data over the whole power range? As said in previous post I didn't do any short Z6+ sprints last month. Maybe that's why W' was modelled too high. 

Mark Liversedge

unread,
Mar 22, 2015, 5:35:15 AM3/22/15
to golden-che...@googlegroups.com
On Saturday, 21 March 2015 21:51:35 UTC, Nigel Van de Velde wrote:
Edit: I just fiddled with the numbers a little bit and I get them more "logical" when I set the critical power as 0,95 * 20min Peak Power and adjust the W' so I get to zero at the end of my FTP test. I think I'll stick to that method from now on.

Could it be that the extended model for the CP chart should only be used once you have enough data over the whole power range? As said in previous post I didn't do any short Z6+ sprints last month. Maybe that's why W' was modelled too high. 

That's an interesting question. Only you, as the user, will know if the data being used is representative.

e.g. you may have never conducted a test to exhaustion but have a large number of rides. Alternatively you may only have a couple of rides and they were both tests to exhaustion. 

One way to sort this would be to add a check box for rides to label them as TTE, then we could only include those in the model. But need to think more on that since a lot of folk never test !

Mark

mshalett

unread,
Mar 22, 2015, 4:07:12 PM3/22/15
to golden-che...@googlegroups.com
Funny, I guess Saturdays are the days to play with GC!  I fiddled with my CP/W' power numbers following a particularly hard intervals long threshold ride yesterday.  After lots of iterations, I settled on the W' as the number to change as it is hard to justify changing FTP if you have a legit test (which I do).  My 3' & 20' model does not generate a realistic W' so that is why I was fooling with it.  I am 65 and my CP is currently 155 last test -  using the model, W' comes out to 3 Kj which is far too low.  I ended up at 152 W, 13250kJ which was giving me a balance of .1 on W' at the end of the second high Z4 interval.  I began to think that this was not giving me a correct stress chart and I wanted to get the W' to zero at the end of my 25k TT.  I never reached that point because I ran out of time.  

However, just now, when I change the Athlete Prefs>Power>Add, the Stress chart is not updating like it did yesterday.   I am on the 2/13/15 3.2 build but I confess, I have no idea what I am doing and it is quite apparent that many features are not working properly.  I suspect that I have commingled my Athlete files somehow.

Is there a guide for installing and updating builds that I could refer to?

Monte

Pete from AUS

unread,
Mar 22, 2015, 6:33:11 PM3/22/15
to golden-che...@googlegroups.com
Monte a close & reopen of GC will usually trigger the refresh if nothing else does.

I presume you're getting a low W' because your MMP chart is relatively flat?

Mark Liversedge

unread,
Mar 23, 2015, 3:03:40 AM3/23/15
to golden-che...@googlegroups.com
Are you running 3.1 or the 3.2 dev build.
3.2 should refresh on demand, if not I want to fix it !

Mark

mshalett

unread,
Mar 23, 2015, 6:25:07 PM3/23/15
to golden-che...@googlegroups.com
Pete and Mark,

Thanks for the reply, running 3.2 dev build.  Found out one reason why it was acting up was that I did not have Athlete>Prefs>Power data input for the periods prior to 3/17/15, so all stress numbers W' were set to zero.  Input some pref's and it worked again.  However, this does bring up my confusion in how to properly set up the dev build.  I created a Monte New (new athlete) and still have lots of very obvious errors throughout GC especially with Compare panes in Summary where my weight is 300+ etc.  Swim workouts from the 910 are all coming in as zero's on everything.  Run workouts import ok on the map but paces are not there.  It is apparent to me this is a case of GIGO.

mshalett

unread,
Mar 23, 2015, 6:44:13 PM3/23/15
to golden-che...@googlegroups.com
What is the MMP chart, you mean Critical Mean Max?  By the way, I have a 25k TT assessment tomorrow.

Pete from AUS

unread,
Mar 23, 2015, 9:52:40 PM3/23/15
to golden-che...@googlegroups.com
Yes that chart.

Good luck with the 25km!

Pete

mshalett

unread,
Mar 24, 2015, 12:53:17 AM3/24/15
to golden-che...@googlegroups.com
... time trailing is one of my favorite things... NOT, but it is a challenge I must admit :  )


I thought that I would attach a couple of crash reports but somehow I must not have saved them, I was pretty sure I did, but using searchlight and cannot find them and it is too late to spend the time, but GC crashed multiple time when changing units from metric to imperial or vice versa when pressing the save button while the stress chart was open.

On Monday, March 23, 2015 at 8:52:40 PM UTC-5, Pete from AUS wrote

mshalett

unread,
Mar 25, 2015, 12:59:07 AM3/25/15
to golden-che...@googlegroups.com
I am with Rob. I was all excited about this when I read the Veloclinic curve stuff but my question is the same as below from Mar 23, is the Feb 13 release the 3.2 dev release?  I don't seem to have the Veloclinic curve or I don't have the 3.2 after all. 

And I had a really decent ride in the wind but it was thankfully cool.  Nothing like 47 minutes all out to get your attention :  )

Thanks again Mark for the great work and to all of the other contributors, thanks too!

Monte



Rob Manning

I'm not seeing how to add this?  
 
I've got the Feb 13 development release, is that not the 3.2 dev release?
 
On Friday, March 20, 2015 at 8:25:07 AM UTC-4, Patrick wrote:
...more 

Karl Billeter

unread,
Mar 25, 2015, 2:02:25 AM3/25/15
to mshalett, golden-che...@googlegroups.com
On Tue, Mar 24, 2015 at 09:59:07PM -0700, mshalett wrote:
> I am with Rob. I was all excited about this when I read the Veloclinic
> curve stuff but my question is the same as below from Mar 23, is the Feb 13
> release the 3.2 dev release? I don't seem to have the Veloclinic curve or
> I don't have the 3.2 after all.

CP chart -> More... -> All Chart Settings -> Basic -> Data series -> Veloclinic Plot


Karl

Rob Manning

unread,
Mar 25, 2015, 5:11:00 AM3/25/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Correct.  I was surprised to see it under the Critical Mean Maximals option, not under the W' options.  I guess it truly is a CMM, but seems like it may be better associated with W'?

Mark, is there any way to get that Ward Smith equation in there to verify the correct CP placement?  Visually drawing an asymptote and trying to fit in that ward-smith isn't an easy task.

-Rob

Mark Liversedge

unread,
Mar 25, 2015, 8:30:45 AM3/25/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Good idea. We have a framework to be able to add models shouldn't be too hard.

Mark

mshalett

unread,
Mar 25, 2015, 4:49:24 PM3/25/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Thanks!  got it!  161 W and 4000 KJ agrees with the model CP & W' Estimator as well.  It is very sensitive at least for me, one watt one way or the other really changes the curve!.  I don't know what the dashed line represents, I guess W'.  It would be good if the Y axis scale could be adjusted as once you play with the slider, the curves go off scale.  The slider is very sensitive on this Macbook Air as well, prob a trackpad pref?

Good work guys! 

Monte
Screen Shot 2015-03-25 at 3.44.24 PM.png

mike veloclinic

unread,
Mar 27, 2015, 2:06:34 AM3/27/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Mark,

I agree that the Ward Smith model would be a nice addition since it plays very nicely with the VC (W' plot). Just make sure the appropriate version gets in unlike what was used in the WKO4 miniseries. The correct version:

P(t) = W'/t * ( 1- exp(-t/( W'/(wPmax) ) ) ) + CP

note that wPmax = Pmax - CP

and substitution is possible as 

tau = W'/wPmax

so alt versions that are also correct but have increased correlation of variables are:

P(t) = W'/t * (1-exp(-t/tau)) + CP

or

P(t) = wPmax*tau t*(1-exp(-t/tau)) + CP

For the fit of quality points the search ranges: 1-60 seconds for Pmax, 60-600 for W', and 600-3600 for CP seems to be. 

As an aside if tau is relatively fixed then wPmax will decrease as W'bal drops. The extension of that is Skiba's model could potentially be augmented by the Ward Smith model so that you not only see W'bal but also wPmaxbal ie see how your sprint potential falls as you burn up your W'bal. My next self experimentation project will be to see how well this holds up. 

m

Mark Liversedge

unread,
Mar 27, 2015, 6:37:57 AM3/27/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Will have a go this weekend, should be pretty easy to add.

Any others whilst I'm at it ?

Mark

Mark Liversedge

unread,
Apr 6, 2015, 7:08:13 AM4/6/15
to golden-che...@googlegroups.com, msha...@mindspring.com
On Friday, 27 March 2015 06:06:34 UTC, mike veloclinic wrote:
Mark,

I agree that the Ward Smith model would be a nice addition since it plays very nicely with the VC (W' plot). Just make sure the appropriate version gets in unlike what was used in the WKO4 miniseries. The correct version:

P(t) = W'/t * ( 1- exp(-t/( W'/(wPmax) ) ) ) + CP

note that wPmax = Pmax - CP

Just added this to the models, will be in the next development release.

Mark 

Nathan Townsend

unread,
Apr 7, 2015, 8:19:28 AM4/7/15
to golden-che...@googlegroups.com, msha...@mindspring.com
In multiple posts in this thread I'm seeing the same basic mistake.... using some method or another which is a suggested "test" of FTP and using that value as CP in the W'bal model, then tweaking the value of W' to make the W'bal model "fit" properly.  This is wrong due to real physiological mechanisms and not because of some mathematical anomaly of the model.   In virtually every case where an estimate of FTP is used in place of CP, it looks as though W'bal is being underestimated and thus going negative prior to the point of fatigue.   Getting CP set correctly is FAR more important than tweaking W'.  

Do not use the 95% fudge factor. Do not do a 40km TT. Do not estimate CP using any method whatsoever which is a recommended "test" of FTP. I know it might be hard for lay people to understand why, but there is simply far too much scientific evidence which invalidates the use of an "FTP like" variable for the purposes of representing a true physiological threshold.  Making sure critical power is set correctly underpins the W'bal model.  

I do not recommend the 3/20 test to estimate CP + W' because there are not enough points on the curve and so there is a larger risk of error in estimating CP when a linear regression is used in this case. eg: if two cyclists have a 20MMP of 300 but one of them has a 3MMP of 450 and the other is 400W, then the guy with the higher 3MMP will end up with a lower CP by about 10W and it will increase W' by a large amount also.  This might be a possible explanation, but exactly the same result could occur by having a higher CP in combination with a marginally higher W'.  If you start to add in more points, then each individual data point has less leverage and thus less chance of inducing an error.  The recommendation is to conduct 3-5 tests in the range 3-15min and my strongest advice is to stick to these recommendations.  You could even fit several models to this data and then take an average CP + W' value. See Bull et al (2000) as an example.

It's not that hard to do people.  You can even do three test durations in a single workout with a 30min recovery period in between!  As a doublecheck, go and do a 20-30 min TT.  If you are relatively out of shape, choose closer to 20min.  If you're in good shape then choose closer to 30min. This value should coincide closely with your CP estimate. 

Finally, once you've got you CP + W' estimates, then you should lead with tweaking the CP value first in order to derive W'bal = 0 at the point of complete task failure.

 

Todd

unread,
Apr 7, 2015, 1:01:38 PM4/7/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Exactly. This is why I have abandoned the notion of FTP -- it is just one point on my MMP and does not tell me much about anything else. Whereas the CP/W' model can explain my MMP from two to twenty minutes.

When I am preparing for or into race season, I have enough quality hard efforts in the 2 to 20 minute range that I can just use a month of data along the the GC CP 2-parameter fit. This provides a first guess at CP and W'. If I can't drive W' to near zero at specific time duration, then I know at least one of the numbers is too big. My own experience is that as I get fitter, I start to get negative W' values. I agree that the first place to look is CP, especially since it can be sensitive to the time intervals used in the curve fit. But I also consider the duration at which I drove W` to be negative. If W' is negative on the long end for a long time (long here is say 6 to 12 minutes), then it has to be the CP value. It is easy to see why -- a change in CP of 10 W over 10 minutes is the same a changing W' by 6000 kJ. From my experience,  a change is CP of 10 W is to be expected but a change of W' of 6000 kJ is not!

The "model" column on the GC summary page is getting better and better. At some point it might make sense to "hardwire" that into the "ride" tab as the default. I think the model will provide most users with a better first guess than they will be able to produce for themselves. 

Pete from AUS

unread,
Apr 7, 2015, 8:28:08 PM4/7/15
to golden-che...@googlegroups.com, msha...@mindspring.com
But in GC you need an FTP value to get your TSS / PMC / Time-in-zones etc to "work".
Whereas if you get CP/W' wrong the only consequence is W'Bal plot is off.
IE Until GC separates CP & FTP it seems more practical to use FTP as 'CP' and 'fudge' the W'
Pete

Nathan Townsend

unread,
Apr 8, 2015, 3:39:04 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
On Wednesday, 8 April 2015 03:28:08 UTC+3, Pete from AUS wrote:
But in GC you need an FTP value to get your TSS / PMC / Time-in-zones etc to "work".
Whereas if you get CP/W' wrong the only consequence is W'Bal plot is off.
IE Until GC separates CP & FTP it seems more practical to use FTP as 'CP' and 'fudge' the W'
Pete



You only need to use FTP for TSS/PMC if you want to compare your own personal values to those of others.  I know of several coaches who use training peaks but simply use an actual measured value of 20MMP instead of FTP for use in their PMC.  Works perfectly.... because a 20min TT is just as good a benchmark of aerobic performance as a 40km TT.  Thus you can use CP for both the W'bal model and TSS/PMC.  All it will do is make you look as though you train a little easier than your doppleganger who uses FTP, but for personal use it is all relative because TSS has arbitrary units.  So it doesn't matter if your TSS values are slightly lower than your doppleganger, as long as the procedure that YOU use to calculate TSS remains consistent over time for YOU. 

The cliche "the best measure of performance is performance itself" is a sentiment I agree with 100%, so why then do so many people insist on applying an arbitrary correction factor to an actual performance measure?  It makes no sense. If you multiply 20MMP by 95% you just broke that rule and you no longer have an actual measure of performance over approx 40km or 60min.  You have an estimate which contains error and in at least one recent study, was shown to be out by several percent. In Thomas et al 2014, average power over 40km (65min) was 91% of average power over 20km (31min).  So it might even be slightly lower than 91% if we took 20MMP.  That is the AVERAGE though which means some of the subjects might be smack on 95% but others could have been as low as 85%.   

In fact I would speculate that within an individual the percent difference between 20MMP and 60MMP could change according to training status.  There is evidence to suggest that a 60min max effort is a far more complex integrative task than a 20min effort.  20min is closer to a task which mostly just comes down to raw aerobic potential, whereas 60min not only includes aerobic potential, but also factors related to central fatigue, temperature regulation, hydration, high energy phosphate depletion, substrate utilisation and possibly ventilatory mechanics.

Thus these tasks measure slightly different things so the idea that everyone can blanketly apply the same correction factor and expect the same outcome all of the time no matter what your current training status, is total bogus.  I've seen AC's excel figure showing a near perfect correlation coefficient of r = 0.99 between FTP and 95% 20MMP, but we all know that he is a cherry picker who will ignore evidence to the contrary to support his beliefs.  His excel figures are not worth the keyboard they were typed on.


Nathan Townsend

unread,
Apr 8, 2015, 4:08:15 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com


On Wednesday, 8 April 2015 03:28:08 UTC+3, Pete from AUS wrote:
But in GC you need an FTP value to get your TSS / PMC / Time-in-zones etc to "work".
Whereas if you get CP/W' wrong the only consequence is W'Bal plot is off.
IE Until GC separates CP & FTP it seems more practical to use FTP as 'CP' and 'fudge' the W'
Pete



OK I just reminded myself of the TSS formula....

TSS = (sec x NP x IF)/(FTP x 3600) x 100

The value of 3600 is the number of seconds in one hour.  So even though Coggan has started to confuse everyone by stating that FTP is not defined as 60MMP and never has been (even though it is written in his book TARWAPM), we can still see that the TSS formula normalises the FTP value (which by definition has units of joules/second) to 60 minutes.  Therefore, if you do a 60MMP test, by default your TSS must be 100.  It is inherent in this forumula that FTP = 60MMP.

Thus if you use 20MMP instead of FTP, then you can simply alter the duration in the formula from 3600 to 1200 which will normalise TSS to a 20MMP.  Now you will have a close comparison for TSS to your doppleganger who uses FTP x 3600.




Mark Liversedge

unread,
Apr 8, 2015, 4:31:36 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Hi Pete,

The devs have discussed adding CP and FTP and came to the conclusion that for the Coggan metrics using CP was good enough. Since zones are setup independently it only really affects the PMC and even then, as Nathan has described in detail, its just a minor scaling issue.

It would confuse too many people if we allowed you to set both and I reckon most folks would set them to the same value anyway !

Mark

Rolf Kjøller

unread,
Apr 8, 2015, 4:42:47 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Nathan, thank you for your testing guidelines.

There's one thing you write, that I have a hard time understanding, can you help:

> As a doublecheck, go and do a 20-30 min TT.  If you are relatively out of shape, choose closer to 20min.  If you're in good shape then choose closer to 30min. This value should coincide closely with your CP estimate.

Why is that? If I ride all-out, shouldn't my average power then be above CP, since P(t) = CP + W'/t?

/Rolf

Nathan Townsend

unread,
Apr 8, 2015, 5:21:50 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
On Wednesday, 8 April 2015 11:42:47 UTC+3, Rolf Kjøller wrote:
Nathan, thank you for your testing guidelines.

There's one thing you write, that I have a hard time understanding, can you help:

> As a doublecheck, go and do a 20-30 min TT.  If you are relatively out of shape, choose closer to 20min.  If you're in good shape then choose closer to 30min. This value should coincide closely with your CP estimate.

Why is that? If I ride all-out, shouldn't my average power then be above CP, since P(t) = CP + W'/t?

/Rolf

You do not insert the longer TT value (eg: 30MMP) into the model though.  You've highlighted precisely why you should NOT use a "long" TT in the above model which is the 2p nonlinear model.  No matter how long you ride for, CP will ALWAYS be slightly lower than the longest TT when using the 2p nonlinear model.  If you ride for 5hrs then CP will be slightly below your 300MMP but that is clearly a flawed approach to estimating CP.

If you ride as hard as you can for about 15min, and then model the results using P(t) = CP + W'/t, then the asymptote value here (ie: CP) will be pretty close to what you can actually do for about 25-30min.

Make sense?

Rolf Kjøller

unread,
Apr 8, 2015, 8:17:22 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Hi Nathan

> If you ride as hard as you can for about 15min, and then model the results using P(t) = CP + W'/t, then the asymptote value here (ie: CP) will be pretty close to what you can actually do for about 25-30min.,

> Make sense?

Yes and no. Empirically / by experience, you may very well be right, but mathematically no, given a) P(t) = CP + W'/t and b) valid estimates of CP and W'.  

But, and this is where my limited knowledge of physiology kicks in, perhaps the discrepancy boils down to differences between the assumption af what CP represents in the classical two-parameter-model and what seems to be a more current / advanced physiological understanding of CP that you are working with?

In any case, it's something I've been trying to reconcile / unite conceptually for a while; FTP, MMP60 and the classical two-parameter model.

In my naive world, I would like the following to be true: FTP ~= MMP60 (or MMP55) ~= CP + W' / 60 (or 55) minutes :)

Does that make sense?

/Rolf

Rolf Kjøller

unread,
Apr 8, 2015, 10:10:15 AM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com

Nathan, just to elaborate a little:

My intuition is that there has been a shift in the understanding of CP from back when the classical 2-parameter model was formulated, so that the CP you are working with is a somewhat different CP-definition from the "classical one" and hence that the test procedure you describe captures a different physiological state than the one "classically" captured with the 3 and 20 minute tests. That would for me explain why CP can be above FTP, which really is incompatible with P(t) = CP + W'/t for any W' > 0... at least within the timespan where the model is valid.

And more about definitions - because, if the CP you are working with ("modern CP") is sustainable for a significantly shorter timespan than the quasi-steady-state that Coggan is referring to can be sustained, if it perhaps is only sustainable for 15 minutes, then the FTP-model and the "current CP-model" probably don't converge... because, a two-parameter model based on the "modern CP" is not able to predict performance for longer timespans than 15 minutes ... because, after 15 minutes the VO2 Slow component effect kicks in and lowers one's "effective CP" ... so even though P(t) = CP + W' / t still is true, a continually decreasing CP means that the two-parameter model overestimates P(t).

Is this completely off the wall? And if not, then how do I predict MMP60 from CP and W'?

:) Rolf

Nathan Townsend

unread,
Apr 8, 2015, 12:06:28 PM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com


On Wednesday, 8 April 2015 15:17:22 UTC+3, Rolf Kjøller wrote:
Hi Nathan

> If you ride as hard as you can for about 15min, and then model the results using P(t) = CP + W'/t, then the asymptote value here (ie: CP) will be pretty close to what you can actually do for about 25-30min.,

> Make sense?

Yes and no. Empirically / by experience, you may very well be right, but mathematically no, given a) P(t) = CP + W'/t and b) valid estimates of CP and W'.  

But, and this is where my limited knowledge of physiology kicks in, perhaps the discrepancy boils down to differences between the assumption af what CP represents in the classical two-parameter-model and what seems to be a more current / advanced physiological understanding of CP that you are working with?

In any case, it's something I've been trying to reconcile / unite conceptually for a while; FTP, MMP60 and the classical two-parameter model.

In my naive world, I would like the following to be true: FTP ~= MMP60 (or MMP55) ~= CP + W' / 60 (or 55) minutes :)

Does that make sense?

/Rolf


Not entierly sure what you're getting at here but am I correct in thinking you are referring to one of the fundamental flaws in the CP model?.... P(t) = CP + W'/t is the equation for a hyperbola with the horizontal asymptote = CP, but we all know that you cannot exercise forever at CP.  This just doesn't happen.  In reality.... and de Lucas tested it in their study, trained cyclists on average sustained CP for 23 +/- 7.5min.  

Really what the CP model (and by extension the W'bal model) is aiming to describe is the development of peripheral muscle fatigue.  In the range 2-15min or so (ie: power above CP) peripheral fatigue tends to dominate over central fatigue, but as we start to extend the duration out towards 30min, central fatigue takes on a greater role.  Secondly, the rate that different fibre types develop fatigue is variable also.  The CP model works well over short durations because type IIx fibres fatigue quickly, but type IIa and type I fibres take much longer to fatigue (which induces inefficency ie: they start to reduce their power output for a given VO2).  Long enough that we've got that additional central effect starting to make a greater impact.

CP represents a true boundary between whether or not you recruit type IIx fibers.  60MMP does not represent this boundary appropriately because there is an additive effect of central fatigue which thus limits central motor drive to the muscle and there is also fatigue induced metabolic inefficiency occuring within the muscle.

In fact I believe that during long TTs such as 40km, the typical power profile essentially reflects these effects. ie: CP gradually declines during this type of activity, and thus so does the power (eg: Brandle and Dennis hr record pacing profile). An unresolved question though is what happens during intermittent exercise lasting >30min.  Does the central component start to dominate in the same manner as it does for TT exercise or do the frequent recovery periods allow for some level of CNS recovery to occur also?  Same question goes for fatigue induced muscle inefficiency.  Does this recover somewhat during rest periods say throughout a 1-3hr ride peppered by hill reps and attacks etc?  


Nathan Townsend

unread,
Apr 8, 2015, 12:41:03 PM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com


On Wednesday, 8 April 2015 17:10:15 UTC+3, Rolf Kjøller wrote:

Nathan, just to elaborate a little:

My intuition is that there has been a shift in the understanding of CP from back when the classical 2-parameter model was formulated, so that the CP you are working with is a somewhat different CP-definition from the "classical one" and hence that the test procedure you describe captures a different physiological state than the one "classically" captured with the 3 and 20 minute tests. That would for me explain why CP can be above FTP, which really is incompatible with P(t) = CP + W'/t for any W' > 0... at least within the timespan where the model is valid.

And more about definitions - because, if the CP you are working with ("modern CP") is sustainable for a significantly shorter timespan than the quasi-steady-state that Coggan is referring to can be sustained, if it perhaps is only sustainable for 15 minutes, then the FTP-model and the "current CP-model" probably don't converge... because, a two-parameter model based on the "modern CP" is not able to predict performance for longer timespans than 15 minutes ... because, after 15 minutes the VO2 Slow component effect kicks in and lowers one's "effective CP" ... so even though P(t) = CP + W' / t still is true, a continually decreasing CP means that the two-parameter model overestimates P(t).

Is this completely off the wall? And if not, then how do I predict MMP60 from CP and W'?

:) Rolf


This is right on the ball.  The CP model is no good at predicting sustained performance for any duration at, or below CP, because by definition, CP is the asymptote. The quasi steady state that Coggan talks about can be sustained for longer because it isn't a maximal quasi steady state. A typical power profile for a 40km TT is to see a gradual decline in the power output. Therefore, the power in the first 20-30min is higher than the average power.  But if you commence a 40km TT right on your previously measured average 40km TT power, for the first 20-30min it will feel quite easy because you are actually below threshold ie: CP.  Coggan's belief seems to be that FTP (aka 60MMP) is equivalent to a true maximal physiological steady state, but this is virtually impossible according to known human physiology.... because fatigue induced inefficiency and central fatigue (amongst other things) gradually develop within 30min.

Maybe in world class TT like Dennis, Cancellara, Martin, Wiggins etc these guys can hold CP for close to an hour, and this would make sense because they should have unusually high percentages of type I fibres which take longer to fatigue. 


As stated above though, it remains to be tested how well the W'bal model holds true during longer duration intermittent activity.  Phil's paper from 2014 shows that it was pretty good over an hour or so, but my own personal experience is that I find it increasingly difficult to drive W'bal down to zero after a long ride eg: 2-3hrs.  This is what we would expect if there is greater central fatigue and other factors such as substate depletion occurring. 


Rolf Kjøller

unread,
Apr 8, 2015, 2:25:40 PM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Thank you, Nathan, this is really interesting.

Just to be clear:

> The CP model is no good at predicting sustained performance for any duration at, or below CP, because by definition, CP is the asymptote.

So, a) if I'm somewhat average and able to sustain CP for 25 minutes, then the two-parameter model should be OK at estimating my "power potential" for durations up to those 25 minutes (CP plus ekstra anaerobic work "fueled" by W'), but for durations longer than the 25 minutes the model will gradually overestimate more and more because of central fatigue, VO2sc and other effects not modelled?

And b), as for the relationship between CP and FTP, there really isn't any, because they are based on two different ways of evaluating performance potential. FTP can be said, and this is meant kindly, to be a performance-based gold standard (we have hour records for a reason, there is a raw appeal in going all-out for an hour, cracking 40 k's for an hour for the first time on your road bike just feels really good :), whereas CP is a measure of the power one is producing in a physiologically well-defined maximal-quasi-steady-state. Both are true, both are usefull, but there's no immediate connection.

:) Rolf

Todd

unread,
Apr 8, 2015, 7:04:22 PM4/8/15
to golden-che...@googlegroups.com, msha...@mindspring.com
I compute TSS based on CP (in that I set my "FTP" to be my CP).

This shifts my TSS downward for a given effort since CP > FTP, but as Nathan points out -- TSS is a relative measure.

A bit more on my switch from FTP to CP .... there is published literature that estimates CP based on a power ramp test. When I first did that ramp test (on a Kickr I got at the end of last year) I found that it very closely matched the value that GC was computing. The ramp test is somewhat different than how we normally try set CP -- so it was another (somewhat independent) estimate of CP that lined up with existing estimates. Right now I think my best estimate of fitness is the Kickr ramp test that I do every three weeks -- the test is simple (just run my Hurts Ergo file), quick (takes about 15 minutes after warm up), blind (don't see power during the test) and  and open year around (inside!) so I expect to keep using it through out the year.

(also, my original post had a typo, 10 watts over 10 minutes is 6000 J of work (not kJ!))

Mark Liversedge

unread,
Apr 9, 2015, 1:40:58 AM4/9/15
to golden-che...@googlegroups.com, msha...@mindspring.com
On Thursday, 9 April 2015 00:04:22 UTC+1, Todd wrote:
I compute TSS based on CP (in that I set my "FTP" to be my CP).

This shifts my TSS downward for a given effort since CP > FTP, but as Nathan points out -- TSS is a relative measure.

A bit more on my switch from FTP to CP .... there is published literature that estimates CP based on a power ramp test. 

Do you have a reference for that ?
Would be useful to refer to it on the, to be written, CP testing article :)

Mark 

Nathan Townsend

unread,
Apr 9, 2015, 9:16:12 AM4/9/15
to golden-che...@googlegroups.com, msha...@mindspring.com
On Wednesday, 8 April 2015 21:25:40 UTC+3, Rolf Kjøller wrote:
Thank you, Nathan, this is really interesting.

Just to be clear:

> The CP model is no good at predicting sustained performance for any duration at, or below CP, because by definition, CP is the asymptote.

So, a) if I'm somewhat average and able to sustain CP for 25 minutes, then the two-parameter model should be OK at estimating my "power potential" for durations up to those 25 minutes (CP plus ekstra anaerobic work "fueled" by W'), but for durations longer than the 25 minutes the model will gradually overestimate more and more because of central fatigue, VO2sc and other effects not modelled?


Yes exactly. However, I think the practical application of the CP model though is best realised through the W'bal model extension, which by its very nature, it is suited to analysis and understanding of intermittent exercise ie: interval training and road racing (other than ITTs).


And b), as for the relationship between CP and FTP, there really isn't any, because they are based on two different ways of evaluating performance potential. FTP can be said, and this is meant kindly, to be a performance-based gold standard (we have hour records for a reason, there is a raw appeal in going all-out for an hour, cracking 40 k's for an hour for the first time on your road bike just feels really good :), whereas CP is a measure of the power one is producing in a physiologically well-defined maximal-quasi-steady-state. Both are true, both are usefull, but there's no immediate connection.

:) Rolf


Absolutely! The hour record is sacred in cycling history. There is no harm whatsoever in having a performance benchmark tied to a global standard unit of time.  3000 yrs ago the ancient egyptians developed the idea that a day should be divided into 24 equal parts, but when they came up with that method of time keeping I doubt they were thinking about corresponding 1/24th of a day to the duration that a human can maintain the maximal power output which does not induce a progressively rising level of peripheral muscle fatigue.  From a physiological perspective, the duration of one hour is a completely arbitrary figure.  Its just a random point on the P(t) curve which for most people lies somewhere between two established physiological thresholds. But the issue is that for over a decade one man with a massive ego and high profile online bully presence has been indoctrinating the lay cycling public that FTP is tightly linked to an underlying physiological threshold, whereas CP is not and is thus little more than a mathematical anomaly floating about in the wilderness of non-steady state metabolism.  He is wrong and there is an almightly mountain of scientific evidence which proves it.   

IMO the search for truth and scientific understanding regarding the world around us is a principle worth aspiring to for its own sake, and this is why many people would consider it important enough to distinguish between an arbitrary hour long performance test and figuring out exactly what power output corresponds to a particular underlying physiological event.  

Todd

unread,
Apr 9, 2015, 2:31:16 PM4/9/15
to golden-che...@googlegroups.com, msha...@mindspring.com

Here is the use of a 3-minute all out effort to estimate CP and W'. I am using a different method (ramp opposed for all-out) and still digging for that reference.

Vanhatalo, Anni, Jonathan H. Doust, and Mark Burnley. "Determination of critical power using a 3-min all-out cycling test." Medicine and science in sports and exercise 39.3 (2007): 548.

Determination of Critical Power Using a 3-min All-out Cycling Test

ANNI VANHATALO1, JONATHAN H. DOUST2, and MARK BURNLEY1

1Department of Sport and Exercise Science, University of Wales, Aberystwyth, UNITED KINGDOM; and 2Chelsea School Research Centre, University of Brighton, Eastbourne, UNITED KINGDOM

ABSTRACT

VANHATALO, A., J. H. DOUST, and M. BURNLEY. Determination of Critical Power Using a 3-min All-out Cycling Test. Med. Sci. Sports Exerc., Vol. 39, No. 3, pp. 548–555, 2007. Purpose: We tested the hypothesis that the power output attained at the end of a 3-min all-out cycling test would be equivalent to critical power. Methods: Ten habitually active subjects performed a ramp test, two 3-min all-out tests against a fixed resistance to establish the end-test power (EP) and the work done above the EP (WEP), and five constant–work rate tests to establish the critical power (CP) and the curvature constant parameter (W¶) using the work–time and 1/time models. Results: The power output in the 3-min trial declined to a steady level within 135 s. The EP was 287 T 55 W, which was not significantly different from, and highly correlated with, CP (287 T 56 W; P = 0.37, r = 0.99). The standard error for the estimation of CP using EP was approximately 6 W, and in 8 of 10 cases, EP agreed with CP to within 5 W. Similarly, the WEP derived from the 3-min test (15.0 T 4.7 kJ) was not significantly different from, and correlated with, W¶ (16.0 T 3.8 kJ; P = 0.35; r = 0.84). Conclusions: During a 3-min all-out cycling test, power output declined to a stable value in approximately the last 45 s, and this power output was not significantly different from the independently measured critical power. Key Words: EXERCISE INTENSITY DOMAINS, EXERCISE TESTING, ANAEROBIC CAPACITY, POWER–DURATION RELATIONSHIP, PEAK V ̇ O


I will add that in the paper cited above they estimate V02Max from a standard ramp test and from their 3-minute all out test. The mean VO2Max differed by 3% between the two approaches.
Message has been deleted

Todd

unread,
Apr 9, 2015, 2:42:19 PM4/9/15
to golden-che...@googlegroups.com, msha...@mindspring.com

I don't go out and do that many 60 minute intervals, but looking at my power files I do notice that my 60min MMP gets closer to my CP as the season progresses and I get fitter. So in that sense, the 60 min MMP is a useful number.

Tony

unread,
Apr 10, 2015, 11:34:41 AM4/10/15
to golden-che...@googlegroups.com
Sorry,where do I find your hurts ergo file?

Todd

unread,
Apr 10, 2015, 3:05:17 PM4/10/15
to golden-che...@googlegroups.com
Hurts Ergo is an iPhone app to control the Kickr (It is my favorite. Clean design, light weight, free)  http://www.hurtsdevelopment.com

Hurts Ergo is linked to the ErgDB. Download and "run" and ergo file from the data base (for free). http://www.73summits.com/ergdb/

The file I use for my ramp is here:  http://www.73summits.com/ergdb/workout.php?id=1176  (disregard the name of the file)

In Hurts Ergo you set a threshold power (called FTP in the app) on the front screen. The ergo file above starts the ramp at 50% FTP at minute 28 and ramps to 150% of FTP at minute 48. So the ramp rate (W/m) is FTP/20. So I set FTP to be 250 then I get a ramp rate of about 12.5 W/m starting at 125 W at minute 28.

Tony

unread,
Apr 10, 2015, 5:34:01 PM4/10/15
to golden-che...@googlegroups.com
Thanks. Yes I have Hurts Erg and had found a likely file assuming that I had found found the right Todd in the author list! I am still not quite sure how one derives FTP from your ramp test. Having set a likely FTP at the outset and performed the test, what happens then? Am I missing something?

Tony

unread,
Apr 12, 2015, 4:23:26 AM4/12/15
to golden-che...@googlegroups.com
Are you measuring Peak Power Output as described by Laursen and described as 'critical Power' by some or are you measuring critical power as described by Monod in 1965?

Brad

unread,
Jul 2, 2015, 8:09:25 AM7/2/15
to golden-che...@googlegroups.com, msha...@mindspring.com
Nathan you seem like you are very experienced, your posts make so much sense.

I started cycling and training with power last July and analysing my data with GC and i'll admit I have done the 20min FTP test and used the 95% fudge factor every month to track my progress and to base my training off it and have also used that number for the CP entry in GC but left the W' to 20kj since I never really knew what that was up until recently. I wish I had a more of an understanding of it all back then and performed more MMP testing over the 3-15min range so I could have more accurate CP entries over the last year instead of the 20min test x .95 as my CP entries. Would you recommend going back and changing my CP entries from the last year? If so, should I change them from the 20MMP x .95 to just the untouched 20MMP results?

Also tonight I tried what you have suggested and started testing my MMP from the 3-15min range, I've decided to test my 3MMP, 6MMP, 9MMP, 12MMP and 15MMP or would I still get a good enough CP result if I only use the 3 MMP points (3MMP, 9MMP & 15MMP) instead of 5. Tonight I did 6MMP first then 15MMP and then the 3MMP last with 30min recovery in between.  The 6MMP went really well obviously since I was fresh, then When I did the 15MMP my legs definitely felt a little fatigued, not much but they didn't feel as good as when I did the 6MMP  and half way through I really thought I wouldn't be able to hang in there for 15min, but kicked those negative thoughts to the curb quick smart and did manage to finish it and got a pretty good 15MMP considering. When I did the 3MMP I still believe I did pretty good but they felt really fatigued, I feel if I did the 15MMP and 3MMP tests separately when legs were fresh I'm sure I could have gone that little harder.  Do you think this would be the case? Or should you still be able to achieve a fairly close end result doing all 3 tests in one session with 30min recovery in between tests compared to if you did all 3 tests separately when you have fresh legs?
Reply all
Reply to author
Forward
0 new messages