Related
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Acute Training Load Calculation

sierrahiker
over 2 years ago

I recently noticed that GC is now reporting Acute Training Load (ATL) for recent activities for my Fenix 6; ATL replaces "Training Load".  GC states that ATL is a "weighted sum of recent exercise load scores".

Does anyone know the mathematical calculation for ATL?  What is the weighting value that is multiplied by each day's exercise score (Exercise Load) in the process of calculating ATL?

Thanks!

Top Replies

All Replies

  • 0 over 2 years ago

    I do not know what the weight values for individual days are, but if you really want to know, you could find it out in the following way:

    1. Wait 7 days without logging any workouts
    2. Do an activity, note its exercise load (EL), and note the resulting ATL. The weight value for the day 0 will be equal to the ratio ATL/EL
    3. Stop recording activities for the next 6 days, and note the ATL for each of the days. The weight for the day -N will be again equal to the ratio ATL/EL

    You could even do it without stopping your regular training - by creating a secondary dummy GC account, and just recording a single activity there.

    If you do it, please post the results here.

  • 0 over 1 year ago

    HI, I had a look at the recently added Acute Load on both the Garmin Connect App and an Epix Gen 2.  With a bit of exploration using Excel it seems to me that the algorithm used for the Acute Load is a simple linear decay of the Exercise Load by 10% of the initial value each following day.  So, if on Day 0 there is a load of 100, on day 1 this counts for 90, on day 2 it counts 80, etc.  So, by day 10 the contribution from that Exercise Load to the Acute Load is zero.  The Acute Load is the sum of all the decayed Exercise Loads still within their 10 day decline on any given day.  So, while the literature talks about 7 days before an Exercise Load is "eliminated", in fact it is 10 days as per the algorithm.

    I had the added complexity that I was recording on two devices and so getting two Exercise Loads.  The Activity can be deleted as well as the Exercise Load, but it is already included in the Acute Load and this can not be removed.

    Going forward I will prevent my Epix from recording rowing on my Concept 2, as the RowPro I use gives more information than the Epix, and I will prevent my Edge810 from synchronising with Garmin Connect, favouring the Epix in this case.

    It would be nice to remove activities from the Acute Load also....

  • 0 over 1 year ago in reply to hobdell1
    hobdell1 said:
    So, if on Day 0 there is a load of 100, on day 1 this counts for 90, on day 2 it counts 80, etc.  So, by day 10 the contribution from that Exercise Load to the Acute Load is zero.

    Thanks for the feedback! Unfortunately, I tested your formula (factors 1, 0.9, 0.8, ... 0.1 for the ten last days) with a set of data for ~30 days, and it does not match any ATL values reported by GC, so it does not seem to be the right algorithm yet.

  • 0 over 1 year ago in reply to hobdell1

    i've just done a similar observation, the Acute Load cannot be removed.

    I had a training recorded for thursday on two different devices. On friday I decided to remove the training stored from my fenix 6 and upload the training from the other device (fit-file). Now i can see that the acute load increased on thursday and on friday (where i did not do any other trainings).

  • 0 over 1 year ago in reply to trux

    Okay, I have more data and a few points where I really only recorded once.  I used Excel to fit the data between Exercise Load and Acute Load.  I get a perfect fit but here is how.  Basically I need to add 28% to the Exercise Load, and then allow this elevated load to decay over ten days with a linear decline (factors 1, 0.9, 0.8....). 

    I really do get a perfect fit to the data so this must be the algorithm which is being used.

    Why does Garmin increase the Exercise Load by 28% before adding to the Acute Load table????

  • 0 over 1 year ago in reply to hobdell1
    hobdell1 said:
    Why does Garmin increase the Exercise Load by 28% before adding to the Acute Load table????

    Yes, it looks almost perfect. Thanks! With real-life data, and a factor of 1.275 instead of 1.28, I get the results almost the same as in GC (+/- 2 points).

    I guess they added the constant factor of 1.28 (or 1.275, or whatever) to make the new ATL better comparable to the old 7 days non-weighted sum.

  • 0 over 1 year ago

    Thanks for this thread - very useful.

    I want to add some observations. I do dual recording of my Zwift workouts to have power charts from my Favero pedals (recorded on Garmin Edge 530) and from Wahoo Smart Trainer. As a result, I get two almost identical activities in GC: same duration, recording time, HR and almost identical power.

    What is interesting:

    1. The Exercise load tab shows both activities but load values are very different: 189 from Zwift vs 222 from Edge

    2. Acute load seems to account for only one of the activities - in my case the one with the smaller load (before workout AL was 631, after 871. Diff 240 matches Exercise load of 189 + 28%)

    I suppose GC just takes the first uploaded activity and updates AL. If another activity for the same time period is uploaded, GC just ignores it.

    What I still don't understand is why EL is so different for the 2 activities. Any thoughts? 

  • 0 over 1 year ago in reply to profarvin
    profarvin said:
    What I still don't understand is why EL is so different for the 2 activities. Any thoughts? 

    Do you use the same HRM strap on both sides (same chest strap connected to Edge and Wahoo)?

    Please share the links to both activities. The detailed data (most of it hidden in the JSON structures), may reveal the reason better.

  • 0 over 1 year ago in reply to profarvin

    There may be more factors involved, but what I see is especially the difference in max power, and generally slightly different power curves. If you zoom the power graphs and compare side by side (you can even select small sections to compare them easier), then there are some differences. Not sure whether it comes from the different time points the samples are taken on both sides every 2s, or whether Garmin averages the 2s intervals, and Wahoo does not, but the peeks are slightly higher and more frequent on the power graph from Zwift. These sharper power peaks then lead to higher anaerobic load, and may indeed influence the total Training Load more considerably than the slight max power difference of some 2% may suggest.