W'bal in "Train mode's" performance chart vs W'bal in activitiy's "Ride chart"

[Karl-Axel Zander]

Hello!

Got a bit confused when comparing the W'bal curve during a training session (image 1) in GC compared to how it looks like in the activitiy's Ride analysis chart afterwards (image 2).

image 1

image 2

The shape of the W'bal curve is naturally the same, but they're not same at all when looking at the values. Feels like something gets very wrong for me with W'bal in Train mode, assuming they both correctly uses the same (differential) formula, it's like CP used in Train mode are of a higher value than I got set. From earlier empirical testing I know I wouldn't even be close to drain W'bal down to 0 on a fresh all-out short (< 5min) interval in Train mode, even though in the analysis afterawards it would be "correctly" somwhere around 0.

Setup: Using differential-formula in Options. Tau left at default 300s (just FYI, if I didn't missunderstood this is only used in the Integral formula). Pretty sure CP and W' is accurate set from recent proper all-in short- & long efforts in the same environment, using the estimator in GC

There is probably an obvious explanation, but I don't see it!

Thanks in advance for input! Cheers

/Karl-Axel

[Julian]

There's a bug in the computation of W' in Train mode.  I asked this a few weeks ago and Mark answered.

[Karl-Axel Zander]

oh! Yeah indeed I saw that thread and even posted in it few days ago, but I kind of got snowed in on details and forgot what the main topic was about. :-) Great that it already was noticed!

Question to you Julian: When you talked about the terms "recorded" W'bal and "predicted" W'bal - is recorded what I see in Train mode, and predicted that I see in the post analysis? Got confused when you posted an image of those two curves in a single chart, how did you do that?

[Julian]

"Predicted" is what's on the workout editor window as you create a workout or the workout window if you import a workout that you're about to "play".  The "recorded" is what you see in when you're riding the workout (it's the light red line as opposed to the "predicted" which is a dark red line) or in your post-workout analysis.  That image I posted was a grab of the chart before I cleared it for a new workout.  In the upper left corner of the chart, when you mouse over the top there's a menu where you can save a png of the chart...

On Thursday, February 16, 2017 at 2:39:23 PM UTC-6, Karl-Axel Zander wrote:

oh! Yeah indeed I saw that thread and even posted in it few days ago, but I kind of got snowed in on details and forgot what the main topic was about. :-) Great that it already was noticed!

Question to you Julian: When you talked about the terms "recorded" W'bal and "predicted" W'bal - is recorded what I see in Train mode, and predicted that I see in the post analysis? Got confused when you posted an image of those two curves in a single chart, how did you do that?

Den torsdag 16 februari 2017 kl. 17:29:53 UTC+1 skrev Julian:

There's a bug in the computation of W' in Train mode.  I asked this a few weeks ago and Mark answered.

On Thursday, February 16, 2017 at 10:00:36 AM UTC-6, Karl-Axel Zander wrote:

Hello!

Got a bit confused when comparing the W'bal curve during a training session (image 1) in GC compared to how it looks like in the activitiy's Ride analysis chart afterwards (image 2).

image 1

image 2

Correction

[Mark Liversedge]

Typically you will deplete W' in ~3 mins when performing an all out effort.

Are you saying that the plot afterwards in Analysis does show W'bal going to 0, whilst in train view it didn't?

Mark

[Karl-Axel Zander]

Ah ok! I never used the workout editor, that's why I missed that. Thanks!

[Karl-Axel Zander]

I'm soon about to do a new short interval test, will post a comparision like the one in thread start then. Haven't taken any screenshots from Train mode for those earlier observations I was referencing to

[Karl-Axel Zander]

Proper all out 5min interval - Wbal difference between train mode's performance chart, and after analysis Ride chart (CP/W' a bit too high set for me here most likely)

[Mark Liversedge]

On Thursday, 23 February 2017 20:35:57 UTC, Karl-Axel Zander wrote:

Proper all out 5min interval - Wbal difference between train mode's performance chart, and after analysis Ride chart (CP/W' a bit too high set for me here most likely)

I'll check the code, do you use the "integral" or "differential" version in preferences?

Mark 

[Karl-Axel Zander]

Differential! Thanks

[Karl-Axel Zander]

@Mark Have you had the time & desire to take a look at it? Very looking forward for having Wbal working in train mode for observation & experimentation :-)

Best regards

Den torsdag 16 februari 2017 kl. 17:00:36 UTC+1 skrev Karl-Axel Zander:

[Pontus]

FYI, If you have a Garmin device you can get live W' there as well, from a connect iq app. Nice as it can also be used outside (provided you have a power meter on your bike).

[Karl-Axel Zander]

Great tip, unfortunately no Connect IQ support for those still hanging around with old Garmin devices :( (e.g yet popular Edge 500 & 510)

[Guido Nottebrok]

Same problem here.
GC 3.4 on Xubuntu 16.04
"differential" version in preferences

light red line: W' calculated by GC when creating the interval
dark red line: W' calculated by GC during the intervals

The max and min values are at the same time but the amplitude of the W' during the interval is much smaller

[Guido Nottebrok]

Is this already fixed in the current dev built?

[Marcelo Heinig]

Using build 3970 v3.5-dev1903 and the error is still there...  

[jrans petrosian]

I have the same problem, anybody knows how to fix this?

[Ale Martinez]

Something to consider is realtime W'bal while training is computed using an optimization (http://markliversedge.blogspot.com/2014/10/wbal-optimisation-by-mathematician.html) which depends on W'bal Tau in Athlete > Model preferences and it is independent of differential/integral setting used for offline W'bal computation.