Recorded W' not matching predicted W' in Train Mode

[Julian]

I ran this workout today on my kickr and everything was peachy on that front.  CP is set at 250 and W' is set at 14,000.

I do have a question- why is the recorded W' (the orange line) not matching the predicted W' (the red line)?

Ancillary question: at about 32mins I paused the workout (dropped my water bottle) and when I restarted it I noticed that the wattage or the communication between the computer and kickr wasn't as smooth and definitely more irregular.  It was almost as if when I un-paused GC the link to the kickr was weaker or somehow compromised... Is that a thing or am I just dreaming.

[richardcarsonmcbride]

In relation to your first question I also gave noticed this. I have tried searching for answer, but no avail.

My best guess us that in somway it is related to W' and tau

Richard

[Mark Liversedge]

Yes, if using the integral equation we predict using Tau in settings, but the computed value may differ.

Mark

[Julian]

So is this an error or a bug?

When I plan workouts I try to take my W' into consideration but if what I plan has no relation to what gets recorded then it seems plotting W' in the workout creator is kind of pointless, no?

Or maybe I'm just seeing/understanding it wrong and need to learn more about t in which case can anyone point me to more info...

[Mark Liversedge]

Which model do you have configured in preferences and what is tau set to for the athlete?

[richardcarsonmcbride]

Mark,

Apologies if this has been asked before but how is tau computed?

Richard

[Julian]

Hi Mark,

I have the defaults set:

W bal formula: differential
W bal tau: 300
W: 14,000
CP: 250
FTP: 250

Just returning from a long injury break so I haven't yet done any testing- I have to build a bit of fitness before I attempt a 20 min or a ramp test.

Is there a primer on how to set and manage w and tau and the formula?

[Mark Liversedge]

Hmmm. It must be a bug. I will need to check to code, because tau is not used when running the differential version.

Mark

On Tuesday, 17 January 2017 14:32:39 UTC, Julian wrote:

W bal formula: differential

[Mark Liversedge]

On Tuesday, 17 January 2017 07:10:20 UTC, richardcarsonmcbride wrote:

Apologies if this has been asked before but how is tau computed?

http://markliversedge.blogspot.dk/2014/07/wbal-its-implementation-and-optimisation.html

Jump to the bit that describes the maths.

Mark 

[Julian]

Thanks Mark, 

So Tau is measured in seconds and is the rate at which an athlete has recovered his/her W' to half.  GC sets the default to 300 (5mins) but of course like everything we will need to adjust it according to our own physiology.  The lower the number the faster one recovers and visa versa.  Tau is kind of like an adjuster for age, as we get older our recovery rate gets slower...

And the differential method doesn't need Tau as it has an accounting of recovery built in.  

Ok.  That makes a lot of sense now.  

There is so much science in GC and us dummies have full access to all the knobs and dials!  

[Potuz]

Ooops, my apologies Mark I emailed you and not the list. Here it goes.
So I'll take this thread to continue asking about tau. I read the linked page
a few times. And now I have opened the file WPrime.cpp to confirm some of my
confusions.

Firstly, the formula that is indicated in the page as well as in
WPrime.cpp:215-217 is

tau = 316 + e^(...)

which clearly is bigger than 316, yet the default in GC is 300? how can this formula be
empirically tested and the software defaults to an unattainable value for the formula?

In the webpage however it's mentioned that highly trained athletes won't see
any difference with the differential formula which results in tau of 60-100,
this is far from what the formula of adding a positive number to 316 would
predict.


So my question is: am I missing some unit conversion or something or simply the default
value is not compatible with the computed value when there's integral data in the ride?

Second question/comment: after reading a little bit about the implementation
of W' I think I take back my statement (and it seems also Coggan's statement)
that the model is overdetermined. I think one should be able to determine
independently an optimal W' and tau from ride data as follows.

From what I see in the code, having good long enough intervals to exhaustion
and with no points below CP during the whole interval should be good enough to
determine W'. This value thus obtained is completely independent of tau since
there is no point in those intervals where one is recovering. Having at hand a
good determination of W' and of CP one should be able to determine tau easily
by regressing from a Microburst session to exhaustion. Namely, on a good
15"-30" on 15"-30" off for 10'-20' one would typically recover according to
tau on the off intervals, but not enough to recover fully. Having a good
estimate on W' and knowing that this was a session to exhaustion should leave
tau as the only free parameter to adjust such that the final W' of the session
is 0. This shouldn't be hard to implement per ride, and then one can do a
minimal square regression or so on a number of points.

I'm sorry to continue asking about this,don't want to be a pain in the @#$ but
I find the concept of W' a very interesting concept just that clearly in my
setup (the default from GC, tau at 300 and W' computed from the model) either
W' is overestimated or tau is below what it being computed wrongly cause I
experience Microburst intervals that I crawl back home with my thrashed legs
just to find out that the model predicted that I still had over 6KJ of W' to
expend.

Best,

P.

>
> --
> _______________________________________________
> Golden-Cheetah-Users mailing list
> golden-che...@googlegroups.com
> [1]http://groups.google.com/group/golden-cheetah-users?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "golden-cheetah-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [2]golden-cheetah-u...@googlegroups.com.
> For more options, visit [3]https://groups.google.com/d/optout.
>
>References
>
> Visible links
> 1. http://groups.google.com/group/golden-cheetah-users?hl=en
> 2. mailto:golden-cheetah-u...@googlegroups.com
> 3. https://groups.google.com/d/optout

[Karl-Axel Zander]

Without being into the mathematics on W', but if I understand this idea correct by Potuz about estimating an current individual Tau, it sounds potential meaningful to me to add such a third test-protocol session after the e.g monthly recovery week.

First session a short interval, second session a long interval (with adequate rest between, a day or 2 to make sure your fresh) to set CP and W'. Then the third session when you do short intervals at fixed power until complete exhaustion, conclude W' bal must be 0 at the point you couldn't go any further, and then the software would present/modify Tau to fit the athlete at the current time.

Of course it would be challange to the not very dedicated athelte to do such a testing protocol montly over 4-5 days in the same environment, and staying fresh for all sessions i.e not screwing up with performance affecting parameters like recovery, food & sleep routines etc

Maybe a step closer to make (as I understand) unproven W' science, somewhat useable in practice if used correctly?

/Karl-Axel

>   email to [2]golden-cheetah-users+unsub...@googlegroups.com.

>   For more options, visit [3]https://groups.google.com/d/optout.
>
>References
>
>   Visible links
>   1. http://groups.google.com/group/golden-cheetah-users?hl=en

>   2. mailto:golden-cheetah-users+unsub...@googlegroups.com
>   3. https://groups.google.com/d/optout