HRR zones incorrect

More info: only way I can get the watch to respect the heart rate zones is to chose HRR calculated zones in Connect, then switch to BPM zones in Connect (which just expresses the previous HRR zones as a BPM vs percentage, then sync to the watch. 

The watch interprets HRR and HRMax as the same thing (and lactate threshold % as well, possibly). 

I'm experiencing the exact same issue - all settings seem to be ignored except for the max HR, and zones are calculated on the default percentage of that (i.e. zone 1 = 50-60% of max HR).

I've tried using HRR, then when that was ignored, I've tried using BPM on the device and manually setting the zones, which was ignored, and finally have tried using Max HR as the setting, but changing the zones, which again was ignored in favour of the 50/60/70/80/90/100 zone limits.

As it stands, heart rate settings are useless, and it uses the default settings for reporting in both connect app and connect web and uses the results to calculate training effect etc (which is consequently giving incorrect results).

This has been happening from purchase, and I've joined the beta program hoping it will have been fixed, but it's still ignoring any setting other than Max HR on v4.19

Any suggestions would be warmly welcomed, because as it stands, the watch is not doing the primary thing I purchased it for!

Apologies for the long screed below, just that you might actually know some answers to a lot of my questions! Appreciate if you can help me, or point me in the direction of the right documentation.

I just was reading through all the posts, and saw this:

"Heart Rate Zones or power zones are just guide posts for the user to help with workout design, but the watch will always use real-time physiological and work output for any metric model, in comparison with VO2 Max HR/HRV and corresponding pace/power." 

Wow, OK! I've been really, really struggling to make sense of how HR zones map to Training Effect, because it seems hard to predict accurately from workout to workout when doing what I'm pretty sure are solid anaerobic HIIT routines. And from what you wrote, that finally makes sense! Because HR Zones <> work measurement. Got it.

But then.. now what? I can pretty easily predict that if I run for a while in mostly high Zone 3, I will register a Base Run and see TE in the mid 3's.

If I step it up, and run with the majority of my time in Zone 4, I will see register a Tempo Run, and see TE in high 3's, maybe low 4's.

Then step up further, get into high Zone 4 and some solid Zone 5, and see VO2Max. All good.

But trying to kock off some anaerobic stats is really, really squishy.  I've had HIIT workouts that have Aerobic TE of 4.6 and Anaerobic TE of 3.1, but then done a workout that was CRAZY HARDER (2x more time in Zone 5, way less in 4) but while the Aerobic TE was 5 (OK, hey) Anaerobic was 2.2!! What??? I spent 25 mins above lactate threshold, and only 15 mins below and I don't rate serious anaerobic #'s?? What is going on? 

So your note has some explanatory power here. Because HR Zone <> the actual work the watch is doing. It is a PROXY for that work. But... even if just a proxy, if I'm stomping my system into mostly Zone 5, that IS harder work. It just doesn't show up reliably in the Training Effect.  What am I doing wrong, or maybe more usefully, how can I better predict how my workout will register as anaerobic TE?

Lastly, in Training Status/Load Focus, there is a purple bar for Anaerobic, which might/might not creep up based on whether or not my over-the-top HIIT routine maybe registers some Anaerobic TE.  But nothing I've ever done has put a purple bar up on Exercise Load. I almost don't believe that is an actual thing, since no matter how insanely I push above lactate threshold (according to HRZ 5) all I ever see is a giant 290 benefit VO2Max orange bar.

I'm so sorry about blathering all this stuff. but the Garmin support folks just pointed me to the very, very thin docs. You seem like you must know how all this works. Help!!

The anaerobic EPOC estimation works well for running and cycling because the watch can use the pace (power) information to identify the intensity of the effort before the HR dynamics are affected. For other activities, like strength training, HIIT, rock climbing, etc, there is no such output metric. So the watch uses only the HR dynamics (rate of change combined with threshold augmented with HRV dynamics) and is looking for interval patterns. 

When you kick in a hard interval, the output is high but the HR is low. The HR ramps up to somewhat high level (maybe up to zone 5 for longer anaerobic intervals - let's say more than 2mn). With pace and power available, the watch can see the intensity of the effort right from the beginning. Without, it might end up underestimating the anaerobic component.

jrbrett said:

I spent 25 mins above lactate threshold, and only 15 mins below and I don't rate serious anaerobic #'s?? What is going on? 

Being above lactate threshold from a HR perspective doesn't mean you are in "anaerobic" zone from an output perspective.

An anaerobic effort is short (let's say 30 to 1:30mn) at *maximal effort". For running and cycling, anaerobic efforts are even better recognized if the rest periods are really rest periods (we all tend to run fast after a very fast interval). For the watch to give you full credit, it is better to walk for 30s then jog slowly for the rest. It is also better to rest long enough. If you do 30s/30s intervals you will get less anaerobic credit than 30s/2mn for example. The watch might rate it a "tempo" workout in the first case, but an "anaerobic" one in the second one. From a training perspective, either one is good though!

jrbrett said:

Lastly, in Training Status/Load Focus, there is a purple bar for Anaerobic,

See above, the watch give you points to add to your training focus bars in all cases, but for the exercise load, the color is determined by the overall workout label. So if the watch gives you a 3.0/3.0 rating and call this tempo, it will add some points to the anaerobic focus bar, some points to the aerobic focus bar, the total of these points will be the exercise load. However, the overall workout is tempo, so it will show as yellow in the exercise load.

The lesson here is that the watch is probably underestimating the anaerobic contribution of some non-running and non-cycling activities. I say probably because it might be hard to get anaerobic in some of these activities. But one example is strength training. Traditional strengh training is anaerobic in nature, but the HR will never give the watch the info matching the effort. HIIT actitivies can probably be anaerobic. I am guessing 1mn AMRAP burpees or box jumps do qualify, but the watch might not give you full credit. It doesn't mean it is a bad workout :-)

Wow. That is all fantastic information. I think you've nailed it. My low anaerobic score elliptical HIIT was 1m/1m, and I wasn't really recovering. The workout with the higher anaerobic score was 1m/5m, so way, way more recovery. I think you've explained it. And yes none of that is an assessment of the workout quality, which was terrific. Just not easy for the watch to figure it out. You are a genius. I will stop fretting about it and just hit it hard at the gym! 

Hi - I’m sure you have better things to do than answer my questions, but you are so knowledgeable about all this.

I get what you’ve said above about the ability for the watch to interpret “anerobic” activity when it lacks speed or power input and has to rely on HR alone. I’ve done a lot of experimentation, and that certainly seems true. However even so, the algorithm appears to be non-continuous. I can do 6 reps of HIIT on the elliptical, and in one rep the Anerobic Training Effect notches up .6 (way!) and in the next, close to identical rep (duration, HR BPM) it will only notch .1 Eh? That is just way too finicky. But so be it, as you said, not its strong suit. 

And yet… the following is absolutely mysterious to me.

Last week, my long easy outdoor run: (Clear open horizon, no buildings nearby to juke with the GPS. Chest strap for HR monitoring).

Training Effect 3.7, 1:02:45 in HRZ 3, 9:19 in HRZ 4, and 1:42 in HRZ 2. Load 139. Registers as “Base.” All is well with the world.

Today, my same long easy run:

Training Effect 3.7, 1:07:17 in HRZ 3, 5:51 in HRZ 4, and 0:00 in HRZ 2. Load of 145. Registers as TEMPO. What???

I spent MORE time in HRZ 3 vs 4 and it upshifted the label? What is going on? Yes, I finished the same distance 45 seconds faster (yay for me!) and the load is *slightly* higher, but was that enough to tip a truly classic base run into another category? My experience of this is that the Algo is “capricious” if not random. There must be very steep step-function involved, and the factor model (if it is a factor model) has all sorts of non-continuous rules.  

OK, look, the watch is just one tool the training arsenal. I don’t “train for the watch.” I had a great, strong, easy run, which is all that matters. That said, the watch (and Connect) are very noisy, and this stuff sort of spams you with “Load Focus is skewed toward High Aerobic” and I need to mix up my routines and such. Which I did!  Sure, I can just ignore it all, but honestly it makes me feel like all this top-tier FirstBeat stuff is a thin veneer and not that great or even baked. If the algo is not a closed-form regression factor model, how does the ML training f up a simple, obvious base run? I’m hard put to tell my friends to run out and get the watch when something so simple just doesn’t work. I get the anerobic bit, even though it SHOULD be able to be slightly more predictable between close to identical reps in a workout. I can accept that. However I’m really struggling with the botched base run bucketing.

Obviously, FirstBeat isn’t going to publish its formulae, but the very simplistic narratives in the help docs don’t have any explanatory power.

Would be thrilled if this was user error, and I just need to get with the plan and grok the algo better!

Appreciate any insight! 

(Later.. also Mr. Grumpy-pants here - my run times for a set distance have been going down, with the same or even lower heart rate. Also a good amount of weight loss (5 lbs) over the last month or so, but my VO2Max hasn’t gone up. I know this is supposed to be a slow moving statistic and not jump around, but if the numerator is going noticeably up (ml O2) and the denominator is going noticeably down (kg), at some point the result ought to change!)

BTW, I’ve re-read your comments from the start of this thread, and completely understand that the watch doesn’t “use” HR Zones for its calculations: it is using direct physiological measurements, as you described. I will assert however, that HR Zones are a reasonably *proxy* for those measurements, particularly across a long enough period of time, such as my hour+ run. Further, during a workout, those proxies are pretty much the only observable data in order to calibrate my effort during a workout, no? I did add “Training Effect” for Aerobic and Anerobic, but those are cumulative, not moment-by-moment levels.

What would be great (I think) would be for the watch to show the the raw measurements in a real-time display of how that data is interpreted - a screen that shows all the values going in to TE and Load, and how they are getting bucketed. The inputs (for running) are speed, HR, VO2Max, and HRV? There isn’t anything else the watch *can* measure, correct? Speed (effort) is the only thing I can control directly, but it would seem that the interactions between all four variables is what the model interprets, and generates a number of output statistics. Thoughts?

jrbrett said:

I can do 6 reps of HIIT on the elliptical, and in one rep the Anerobic Training Effect notches up .6 (way!) and in the next, close to identical rep (duration, HR BPM) it will only notch .1

This is because the watch is looking for peak EPOC. It is possible to reach the peak in reps before the end of the exercise.

https://assets.firstbeat.com/firstbeat/uploads/2015/10/white_paper_epoc.pdf

I observe the same thing when I do anaerobic 400m speed repeats on a track. I get 80% of the anaerobic benefit before the last 3 or 4 reps, which give me only 0.1 or 0.2 additional anaerobic TE.

Vice versa, if I do 400m to exhaustion up hill repeats, I get almost no anaerobic TE and a tempo rating although the RPE is clearly higher (I don't run the 400m on the track to exhaustion). This is the proof that pace plays a significant role in the anaerobic estimate. If only Garmin would switch to running power...

jrbrett said:

I spent MORE time in HRZ 3 vs 4 and it upshifted the label? What is going on?

The watch is looking at the rate of increase and the level of peak EPOC and the reps. It is possible that in the second run, the rate of change of EPOC kept it in base territory while the level of the peak tripped over in tempo territory and that locked it in.

Also, tempo seems to be a catch all for the workouts where specific criteria, thresholds or rules used to assign the other labels don't quite work out. As I mentioned above, if your rest periods during an anaerobic workout are not long enough or clear enough (you run too fast for a true rest interval), you will most likely get a tempo label although the TE components will show a nice anaerobic number.

jrbrett said:

FirstBeat stuff is a thin veneer and not that great or even baked

I disagree. What is nice with the Garmin approach is that the metrics are linked to the physiological costs of the training, not the output (like TSS or RSS with power or pace). All of these metrics are estimated from models with a pretty low error rate (about 5%, maybe worse if your HR Max is off). Of course there are glaring issues with anaerobic estimates except for running on flat grouns and biking. 

jrbrett said:

I will assert however, that HR Zones are a reasonably *proxy* for those measurements,

Yes they are, notwithstanding the HR lag and HR drift phenomena. Nothing wrong with training with HR or even simply with RPE either. 

jrbrett said:

Obviously, FirstBeat isn’t going to publish its formulae, but the very simplistic narratives in the help docs don’t have any explanatory power.

There are several white papers and resources I found useful. This other one about EPOC-based training:

Microsoft Word - EPOC_based_training_effect_assesment_REVISED 2.doc (firstbeat.com)

https://www.nsca.com/education/articles/kinetic-select/oxygen-uptake-and-the-aerobic-and-anaerobic-contributions-to-exercise2/

https://www.youtube.com/watch?v=xUr6tS7QSdM

jrbrett said:

it would seem that the interactions between all four variables is what the model interprets, and generates a number of output statistics

This is it, but it is probably not a simple set of "if..then..else". It is even possible that machine learning and predictive analytics are used. This dissertation explain how neural networks can be used for a variety of applications from HR correction to VO2 estimates. We can assume a subset of these have been implement since then. The paper is arid but very interesting:

https://jyx.jyu.fi/bitstream/handle/123456789/13267/951391707X.pdf

Much gratitude. Wow. Thank you so, so much!

I’ve made it through the FirstBeat paper on EPOC. Epic! (OK, I’m sure that joke is really old…) I’m starting to grasp things, I think. It was helpful to see those graphs of the different exercise patterns and the EPOC load. That and your running experience on the track and the hills.

I take back my grumpy rumblings. I will read all these papers and may come back with questions if you don’t mind.

Am I correct then that a steady, hard pace at > 85% of HRMax is likely to generate more Anaerobic Training Effect than intervals with a lot of recovery? Just go flat out at race pace and stay in it for 30 mins or so? (Not so much as a training plan, but would that result in the likely highest EPOC prediction by the watch, given the design of the model?)

On the “Base” vs. “Tempo” question, I guess I prefer continuous variables to opaquely bucketed ones. (E.g. Training Effect/Load vs. “Benefit”) but I understand the need to bucket things eventually. I just wish I could “see” how the inputs were collated — e.g. did running slightly faster tip things from “Base” to “Tempo”? Where is that tipping point? I know you said that the system distributes the TE across the different buckets in Load Focus, but I’m not so sure. It looks to me that is true for anaerobic TE, but Base, Tempo, and Threshold go all/nothing into teal and orange. I got no Base credit for my Tempo Run (that I was expecting to be Base). Instead, the orange Highly Aerobic line got longer and I was inveighed by Connect that I need to mix up my exercises. 

Anyway, thanks again so much. A lot of reading ahead, but I will try and think in EPOC terms!

jrbrett said:

the system distributes the TE across the different buckets in Load Focus, but I’m not so sure. It looks to me that is true for anaerobic TE, but Base, Tempo, and Threshold go all/nothing into teal and orange.

There are 2 graphs, le training focus graph (horizontal bars, under the load focus sub widget) and the exercise load graph (under acute load sub-widget).

After you execute a workout, you will see that the total of the focus bars will increase by the exercise load of the last workout. That load will be distributed in 3 buckets from the 2 TE numbers. The distribution is not proportional to the value of the TE components (I checked) but it is related.

You will also see that the exercise load shows up in a single color in the exercise load graph. That color is wins it all depening on the label of the workout, regardless of the relative weight of the TE value. The (added) height of the bar is also the exercise load of the workout.

jrbrett said:

I got no Base credit for my Tempo Run (that I was expecting to be Base)

I am not sure indeed whether it is possible to have the aerobic load distributed between low and high aerobic for the same workout. But I am sure that the load of a base workout with some anaerobic TE, or of a threshold workout with some anaerobic TE will be distributed between low/high aerobic (respectively) and anaerobic.