Achieved level of training performance.

GPLama

Shane, how does the evaluation of the performance of the training on E1040 compare with you?
Have you ever done the comparison I described?
Do you have any idea what I am "doing wrong"?

Sorry to refer to you, but you are passionate and knowledgeable about technical issues ;)

Greetings
Artur

Hi Arthur - I'll have to tap out on this one. It's something I haven't dived into for a few reasons - One being that I'm using two or three Edge devices multiple times a day for the work I do, and that really screws with any training data that it's trying to report on. The other being that my forward focus will always be with companies who support what I do... and in recent weeks I've been informed that is no longer the case with Garmin. I'm doing my best to remain passionate and knowledgeable with their products.. it's clear they rather I didn't. 

Happy 2024 anyhow! :) 

Hi Shane.

Thanks a lot for your prompt reply.
You made me worried about the coverage with the Garmin :(

I also wish you a Happy Year!

Greetings
Artur

What is the "execution score" shown in Garmin Connect?

I would also help understanding what's going on if you could share the details, possibly screenshots, of your workouts and your execution of those.

From my experience it's easy to get good execution scores for power-based indoor workouts with a smart trainer, it's more difficult for outdoor rides. A lot depends on the exact definition of the target. The narrower the target range the more difficult is to keep the metric in that range (in the "green" on the workout page). For power, 3 second power target will be less noisy and easier to maintain than spot (1s) power target. Same goes for dual-sided power meters vs single-sided power meters - the latter are more noisy as they multiply your measured single leg power by two. Furthermore, there can be more unusual targets, for example TrainerRoad outdoor workouts have a lap (interval) average target, so it's a bit difficult to maintain it in the beginning, but then it becomes trivial due to longer averaging period. Finally, I think there is something weird with how the activity outside of the workout (if you keep recording your ride after the workout or start recording before you start your workout) is treated - I believe this can lower your score, possibly due to a bug in the score computation algorithm.

So if your activity is pretty much your workout and your workout target is sane for your setup - so that you keep it in the "green" almost all the time - your workout execution score should be high. If it's not (or if it's different between your different devices as your are explaining) I agree that there is something fishy going on. But without further details it's difficult to say what this is.

At the outset - thank you for your commitment.

You are right about the difficulty of meeting training goals during outdoor activities.
However, I do my workouts on a trainer (Kickr v6), the other gauge is Assioma Duo.

Regarding the structure of the workout itself, it is created on "interval.icu" and synchronized with GarminConnect.
To answer your question about what is "workout execution", below is a screenshot because I don't know how to describe it. It's simply an evaluation of workout execution but before the summary and before the training effect.

Below is a sample screenshot of one training block showing the power range I should maintain. The average power from that block is also presented - it is within the target range.

The most annoying thing for me is that two devices (Edge 1040S and FR955) have the same workout imported, from the same platform (interval.icu), at almost the same time started the workout (not even one second difference) and despite this, after the workout they present a diametrically different level of performance.
"Edge1040S" oscillates between 21-23% and "FR955" - 97-99%.

This is what I can't explain to myself. Well, how? :)

greetings warmly
Artur

It seems that your target is pretty narrow, From my experience 7W between min and max is not much, unless your power measurements is smoothed out and controlled through the resistance feedback loop.

Assuming that:

- Your watch is connected to the trainer (possibly even controlling it) and records very smooth and spot-on (due to feedback loop) power.

- Your Edge is connected to an independent power source (pedals) and records much more noisy power measurements with some offset due to typical 1-2% power measurement error.

I wouldn't be surprised (with that narrow power target for spot power measurements) if the watch was happy with the execution while Edge was regularly in the red. Isn't this what you are observing when executing the workout? From your screenshot it looks like the average power was 145 with the upper limit being 146, so I'd expect it to be regularly above the target and possibly sometimes below. To my understanding if your power is not within the limits 78% of the time then your execution score will be 22%.

If you can, I would suggest increasing the target margins. It shouldn't matter for your trainer control, but would more likely make your independent power measurements to fall in the range despite some offset and some noise. Another or additional option could be switching to 3-seconds-average power target (rather than spot power target), but I am not exactly sure if this would make your trainer control smoother as well, or not.

Thanks again, Martin, for your participation in the discussion and your valuable insights.

Responding to your thoughts, the trainer is usually controlled by the "XERT EBC" app installed on a smartphone. It also receives power from "Kickr v6".
"Edge 1040S" receives power from "AssiomaDuo" (because I have driving dynamics in it.
"Forerunner 955" receives power from "Kickr v6".

Multiple record comparisons between the "Kickr v6" and the "AssiomaDuo" (so-called dual recording) performed on ZwiftPower are very, very even. Of course, with very short sprints the "AssiomaDuo" responds faster, and here there is a noticeable difference of ~1.5% compared to the "Kickr v6".
But I don't have a lot of such workouts, and for them I could understand the differences in performance ratings.

But, there is another but - there are times when I swap the power saved between the devices and on the "Edge1040S" I save the power from the "Kickr v6" and on the "Forerrunner 955" the power from the "AssiomaDuo".
I've done so many attempts, tests because I'm not comfortable with this "achieved training result" displayed on the "Edge". Such a change in the power source does not change anything in the "Training Score" :(

You consider that the target range of the training block is too narrow. For the example screenshot from the previous post, the block power is 143W but the margin up and down is ~2.5% each. Which gives a range of 139 - 146W.
For common sense, if the average of such a block falls within these ranges then the block has been done completely correctly - in my opinion.

I mentioned, I have workouts from "interval.icu" and synced with GarminConnect, in which there is no editing.

If I create a workout in "interval.icu" with the parameter "power=3s" then in "Edge1040S" the given block is displayed with "power from 3s" but at the top in the "Target" section the range "143 - 143W" is displayed. That is, a fatal solution practically impossible to achieve because it would require maintaining the power in the block every watt.

May I ask a question: are you Polish (a rather Slavic name ;))?
If so, we can somehow click to exchange experiences more in a relaxed way?

greetings
Artur

Now I made attempts to create a workout in "interval.icu" according to the rule:
- power level expected: 67%
-- I entered in the training assumptions: 64% - 70%, 3 second power.
With this assumption, the "Edge1040S" now displays in the center field "power from 3s" and at the top of the target the power range "147 - 161W" - where the required training power is this "154W" which gives 67% TP here.

At first glance, dry - it looks as if your considerations have been met. Well, I'll check this afternoon how "Edge1040S" will evaluate the training performed this way ;)

Nevertheless, I have a great discomfort that "FR955" with the original workouts correctly evaluates the workout performance, and "E1040S" very badly :)

I will let you know how it comes out today ;)

Cheers
Artur

Hi Marcin.

So here I am after today's training, this is what it looks like on "interval.icu" after our discussion:

After the workout, the "E1040" (paired with AssiomaDuo - it doesn't control the trainer!) displayed a rating:

While the same workout recorded on "FR955" (paired with power from Kickr v6 - does not control the trainer!):

The trainer is controlled by the "XERT" app on a smartphone.
This is the summary of today's workout on "interval.icu":

In summary, now on the "E1040" is a better rating, but on the "FR955" is worse :D
Garmin does not support any ramp, and warm-ups are mostly based on ramps, so here it may slightly miscalculate the performance rating.

But on the other hand, "FR955" has the same workout rated :)

There's no denying it, garmin to evaluate the performance of the workout uses some kind of computational technique on the order of the cosmos, with differentials, potentiation, fields of volume, density and stick knows what else ;)
On common sense - you fit within the block/interval fork then you should be OK.
E.g. Zwift "stars" blocks/intervals as long as they are within +/-5% from the assumed power. If it does not fit - then there is no asterisk ;)

Thanks Marcin for your suggestions and participation in the discussion :)

greetings
Artur