Firstbeat Metrics | What Next?
The Fenix 7 has all the Firstbeat metrics I’m going to briefly cover below. But where will Garmin take Firstbeat in 2021 and beyond with the next generation of Fenix 7 watches and others?
The easy pickings from the list of ‘new metrics’ have been incorporated into the latest, greatest Garmin GPS devices. Now what? Might there be a deep, new features requiring the same degree of complexity as the recent Daily Workout Suggestions?
Will the next Firstbeat metrics be evolutions of the existing ones like, for example, those that have been modified to take into account heat acclimation? Perhaps others can take into account your age? Or perhaps they already do?
This article offers the following
- A recap of the current Firstbeat metrics on Garmin devices
- A vote for your favourite Firstbeat features
- I’ll add some further thoughts and then
- You can add your comments on what you would like to see in the future from Garmin-Firstbeat.
Must Review: Garmin Fenix 7 Review & Thoughts
Firstbeat Metrics – An Overview
Firstbeat metrics are sensor-derived from either EPOC/HR/HRV or POWER with subsequent staging points like LTHR/AnT, FTP and VO2max. EPOC is comparable to TRIMP.
As always, these measures are only as good as the inputs. So those that come from power meters should be mostly good but those that come from heart rate will be variable in quality, especially if you use a wrist-based optical heart rate monitor.
Another way of classifying the metrics is to consider those which are updated live as you exercise, like performance condition, and those which are mostly valid once your workout is completed like an assessment of your VO2max or your recovery time. Some metrics say ‘hey you’ve achieved X’ and others put ‘X’ in context with the historical trends of your recent performances or the trending state of your fitness. Perhaps a final way of classifying them would be ‘PRO’ metrics like those I’ve just alluded to and then the more consumer-grade metrics like body battery which many more people might be interested in, including some wannabe athletes.
Hopefully, that’s set the scene. Before I recap the current suite of metrics and suggest future directions, please vote for which metrics you currently use or love. Please comment on what you would like adding at the end.
Real-Time Performance Condition
Performance Condition looks at the relationship between your running speed (or cycling power) and heart rate. It’s visible as a workout metric after 5 or 6 minutes.
Usefulness?: 7/10. It’s certainly ‘interesting’ but I’m not convinced about accuracy.
Training Status
This is an at-a-glance dashboard-like feature that is available after your workout. The two arrows are self-explanatory and the textual description of your current state is handily summarized in one or two words, like PRODUCTIVE.
It doesn’t help you by giving predictions or trends on the watch and it’s unclear to what degree FATIGUE is included.
Usefulness?: 7/10. It’s interesting how Firstbeat interprets your training. When following a plan Firstbeat mostly agrees with what I would expect to see. If you are ‘randomly training’ and not following a plan then is a useful feature.
Recovery Time Advisor
This is how long you need to wait until your next hard workout.
It is useful information that I find is overstated if either my zones are wrong, my data is incomplete or if I have had HR readings that are too wrong/high. This metric is only useful if you are doing endurance training more than 3 times a week, less than that and you are probably always ready to ‘go for it’.
Usefulness? 9/10 I certainly take it into account when training frequently.
Daily Workout Suggestions
Daily workout suggestions are the pinnacle of Firstbeat‘s achievements and pull together many of the individual metrics to guide you through suggested workouts and predict the training effect. If you are not following a formal training plan these are GREAT workout suggestions that cover both running and cycling workouts
Usefulness? 9/10
VO2max
VO2max is a definitive metric of fitness, although not a measure of effectiveness. Neither is it overly actionable. I am not entirely clear what Firstbeat is showing and my VO2max can change by 3 or 4 points in a relatively short period of time. So is this showing the VO2max my performances actually demonstrate OR the VO2max I could achieve now if I had a test? I think it’s the former and so I’m not sure this is as useful as everyone thinks. That said VO2max trends are interesting.
I’m certain that VO2max calculations will be expanded to cover other sports. This will be for running based sports before ones such as rowing.
Usefulness? 5/10. In itself, I would take the number with a pinch of salt and then just look at the trends. However, a lot of Firstbeat’s other metrics are derived from VO2max so it NEEDS to be correct. But is it?
Heat & Altitude Adjustments – VO2max & TRAINING STATUS adjustments
Temperature acclimation or altitude acclimatisation are adjustments to the standard models in more extreme conditions
Usefulness? 6/10: I guess it makes the models more correct so it must be good.
Training Effect
Generally, the post-workout, training effect feedback data feels correct.
Workout Labels
The two Training Effect measures are combined on one summary screen to which is added the WORKOUT LABEL (‘Tempo’, image below).
Aerobic & Anaerobic Training Effects
Endurance athletes and people trying to lose weight will want mostly aerobic training effect. Balancing the effective introduction of anaerobic workouts is tricky but important. I believe I am aerobically very fit but if I have not properly recovered when I attempt a harder workout I find it hard to score anaerobic points. That’s Firstbeat telling me I’m doing my training wrong.
Usefulness? 9/10: I really like these summary screens and the data they give.
Training Load
Training Load (TL) is super-important.
And look. It even shows a rolling 7-day period. You don’t know how many other companies simply show the standard 7-day week on many charts.
Firstbeat gives us an indication of TL as an accumulation and trend of EPOC but strips out periodized effects. It also gives few insights into the anaerobic breakdown and breakdowns by sports for multisport athletes.
Usefulness?: 7/10. Training Load is useful but masks the complexity that lies beneath.
Training Load FOCUS
TL Focus attempts to classify your zone-based loads against the optimal. It’s wrong. It doesn’t show the optimal for your race duration nor for the periodised state of your training right now.
Usefulness? 7/10. Useful and needs to be MORE useful.
Lactate Threshold
Fistbeat uses the AnT/LTHR Lactate threshold and HRV data from a chest strap. It’s the point above which your performance becomes increasingly inhibited by lactate buildup. Handily you also get the running pace at which this happens.
Usefulness? 9/10. This is a fundamental metric for endurance athletes and yet I feel Firstbeat is often wrong and obviously influenced by erroneous HR readings.
FTP
This is effectively the same as LTHR but translated into a power value. FTP does not properly account for fatigue when you want to use it to determine how to perform. Firstbeat now seems to be able to determine FTP from time periods other than 20 minutes, which is how it should be.
Usefulness? 9/10. FTP and power are super-useful but not a panacea for world domination.
Race Time Predictor
I’ll stick with the 645’s image, above, as it’s flattering 😉 Being more honest, the newer predictions from Firstbeat’s latest race-time algorithm appear accurate.
Usefulness? 7/10. It’s nice to know what you might be able to do with specific training.
Calories Burned
Hey. You know what calories are. And you can see them trended in lots of exciting ways.
Usefulness? 5/10. I mean calories in/out are obviously important but…you’re an athlete worried about training too much rather than eating too much, right?
Respiration Rates
I’ve always wanted this metric and knew it was possible YEARS ago from HRV-derived data. Now that I have it, I’m not quite sure what to do with it.
In some sports lab tests, Ventilatory Threshold (VT) is marked. Maybe there is usefulness related to identifying VT?
Usefulness? 5/10. I mean, breathing is obviously important but….do I target a higher breathing rate or monitor it over time?
All-day Stress and Recovery
This is HRV-based and superficially interesting data but I’d like to see the science that validates it (it doesn’t).
Usefulness? 3/10 – It seems to identify and mirror the stresses in your life.
Quick Stress Level Test
These HRV readings can indicate stress of some form but I wouldn’t place too much value on any stress reading other than one taken when you first wake up.
Usefulness? 3/10. Unless you are a stressed-out person, then it will stress you out some more.
Body Battery
Body Battery combines All-day Stress data, physical activity, and the restorative power of sleep into a single & easily understood metaphor. It’s perhaps indicative at best.
Usefulness? 4/10
It’s highly unlikely that Firstbeat will incorporate data from any move external sensors eg fat/carb consumption from Lumen or core body temperature from GreenTeg
The Future of Firstbeat Metrics – 2021
2021 will see evolution, not revolution. For example, with VO2max being calculated for new sports
The most likely inclusion of new Firstbeat metrics will be around running power as a means of determining effort, new metrics will mirror the equivalent Firstbeat metrics derived by cycling power. This will happen only if Garmin introduce native running power and wrist-based running power calculation (or their own pod to rival STRYD). I think we shall see a native running power by Q2.2021 with the Firstbeat components following later.
My other hope is to see closer attention paid to training load and its constituent parts plus the presentation of its data.
The Firstbeat metrics are all closely intertwined behind the scenes. However, in my opinion, they do not gel together as a coherent feature set from the athlete’s perspective. I can’t offer any specific advice to Garmin but hope that Garmin will ‘tidy up’ the display and presentation of all the metrics both individually and how they might flow from one to another.
Less likely, we might see Garmin introduce onboard body temperature sensors. This would come with the ELEVATE v4 oHR sensor and the data could improve the quality of the sleep and recovery-based metrics.
What do you think?
Reader-Powered Content
This content is not sponsored. It’s mostly me behind the labour of love which is this site and I appreciate everyone who follows, subscribes or Buys Me A Coffee ❤️ Alternatively please buy the reviewed product from my partners. Thank you! FTC: Affiliate Disclosure: Links pay commission. As an Amazon Associate, I earn from qualifying purchases.
I’m waiting for 955, April time me thinks, 945 has been superb ?
And announced next week at CES ? I hope so since I just lost my 935 and the Coros Pace 2 feels a bit “simple” (although it does a fine job).
something will be announced soon.
655 is overdue
945lte/955/955lte are the only leaked products that have not surfaced.
deduce away.
yes, i’m coming round to that thinking on timing
As long as those metrics are reported to garmin connect over the integrated lte modem I’ll be happy.
I think the VO2Max estimate is in the ballpark of +-10% in the real world. Maybe worse in some cases. Maybe better for others. The Firstbeat documentation is thin but acknowledges +-9% if your max HR is not correct.
https://assets.firstbeat.com/firstbeat/uploads/2017/06/white_paper_VO2max_30.6.2017.pdf
Some kinds of workouts and terrain are much better than others for its algorithm. Implicitly Garmin acknowledges this as the Trail Run app doesn’t calculate VO2Max and also if you accidentally use the Run app on a technical trail, your VO2Max estimate is guaranteed to plummet.
I think the fluctuation we see is evidence of the error in the algorithm. I see it fluctuate in a week within +-5% regularly. That seems likely to actually be just noise.
I also am sure that if you are only using the built-in optical HRM, the value is going to be total nonsense.
The larger concern that I have is that Performance Condition is often significantly at variance with VO2Max estimate and they are supposed to be based on the same thing. Garmin says 1 point of PC is 1% relative to “baseline” VO2Max.
I don’t think the actual VO2Max estimate is precise enough to be a useful input to these other algorithms. And for me it doesn’t correlate very closely with Stryd threshold power estimate or Training Peaks lactate threshold estimate like it should but those two correlate closely with each other.
It suspect there isn’t enough memory and processing power in the watch to do as good a job with estimates that TP and Stryd do on the server (with machine learning?).
I think the Garmin VO2Max estimate is useless as a training tool. It’s only a game to compare with your mates. I have removed the widget from my watch and GCM and would turn remove the VO2Max page from the post-workout summary if I could.
interesting, i hadn’t seen the +/-10% doc. so if it says 60 it’s between 54 and 66? or between 50 and 70. former I assume.
inbuilt hrm…nonesense…yep (often)
performance condition sometimes seems great and sometimes doesn’t make sense for me.
machine learning: Yes, I’m not sure who exactly does this where and when. I think it is more widespread than we imagine. eg WHOOP does it. so whilst some people will decry the point-accuracy of whoop, with machine learning the noise can be removed to some degree.
The f6 generation watches still have a tiny and low powered ARM processor with no GPU or ML neural accelerator type of silicon and they are severely memory constrained. There is no way the on-watch calculations of VO2Max are using any ML. And furthermore the memory constraint must also influence the type of algorithm it can use.
I just think the technology on the watch has no hope of having a reasonable estimate of VO2Max to the 1L/(kg*min) that people think they get and certainly not to the 0.10 that people want.
Yeah. I think it is in the ballpark of 5L/(kg*min) for most people and also can have meaningless noise fluctuations of +-2 or 3.
Within that level of precision it is a cool technology but it would be useful only for comparing people who are totally unfit and overweight on the couch to reasonably fit. Like for a fitness device aimed at couch to park run participant or something.
I don’t think it is useful to tell if the training plan of even a 4 or 5 hour marathoner is effective.
And I think that is true across the industry for automated VO2max inference. I’d like to be able to totally hide it from the watch UI and have removed the widget but am vexed they I can’t adjust the post-workout summary pages to remove it.
Polar seems to be aiming at a ramp test which in principal would have a better chance of meaningful results. Garmin has a FTP test but in my experience it fails more than it works which makes it discouraging to use and not worth the effort.
By contrast Stryd’s FTP ML inference seems to work and Training Peaks Lactate Threshold also seems to work. These two independent systems generate values that corroborate each other for me.
yes, ‘people’ DO think they get +/-1 unit of accuracy.
Like you, I don’t think that 😉
Yet as the above survey results show, everyone seems to want VO2max #sigh.
FTP ML…do they do ML? I thought it was just a simple calculation based on avg power over various durations. I’m pretty sure??? it requires a bit pf processing to do that but not masses, i’d guess a watch can do it
I don’t know how Stryd FTP estimate is implemented. I guess I assumed ML since that is where technology is in the world today for this sort of trend analysis. But I don’t know how it actually works.
many of the platforms estimate ftp from efforts other than 20 or 30 minutes. essentially you apple a slightly different factor eg 30 mins is 100% and 20 mins is 95% of FTP. maybe the effort of 19:30 mins is 94.2%…something like that.
I like the body battery more than you appear to. I think they should do more in that vein and essentially replicate what Whoop does.
i’ll need convincing that 24×7 daily hrv readings mean something regardless of who does them
Maybe. I think there is some useful potential there and the continuous readings seem to m have the potential to overwhelm noise.
I don’t think Whoop can be successful long term as a blank band that is as large as a watch. It feels like technology demo that should be embedded in a general purpose fitness watch or else made as tiny as the Oura ring.
yes you are talking accuracy of the reading (agreed) however there is also the interpretation of what that means and interpretation is FAR from straightforward.
whoop/rings: a ring is unlikely to ever do in-sport readings. but it IS probably the best sedentary sensor type/location.
i agree that whoop is going to find the future tough. that said, there must be a next get piece of hardware due soon
Personally, I find Body Battery to be the ONLY First Beat metric of any value. It’s the only one that tells me something that’s not either laughably inaccurate or something I already know.
same to me, probably one of the most valuable metrics to me.
I found it quite accurate and often use it to plan my sleep and running sessions
Something like a fuel gauge in a car for use in ultra races – takes power meter data and notifies when to eat, drink, slow down if estimated time to empty is less than race distance.
xert.
THIS 100% THIS…. Thats a great idea! Have you submitted that to Garmin Ideas? (Share Ideas | Garmin)
https://www.xertonline.com
Purchase Xert in 2021?
At the moment most performance metrics and even the workout suggestions rely heavily on VO2 max. Given VO2 max is disabled during trail running, most of the stuff doesnt work or doesnt make too much sense when trail running (Workout suggestions BASE for trail running make my heart go boom).
So I think they should heavily improve everything trail running related… e.g. even learn from user specific trail running patterns (how much uphil/downhil and whatnot) and relate them to more helpful performance indicators. Machine Learning should make it possible.
Stuff like grade adjusted pace and climate adjustment are small steps in the right direction.
as per some of the other comments, the watches probably aren’t powerful enough for machine learning in real-time (IDK for sure). See: AI Endurance https://the5krunner.com/2020/10/26/artificial-intelligence-enabled-5k-cycling-triathlon-training-plans/, he’ll know (Markus).
the vo2max point then comes back to what I was saying in the article. what exactly is it? if it is your current capacity that you could do today in a lab test, then that wont change much from one week to the next. and so there is no point in calculating it during a workout.
yes the small steps are interesting and maybe some times insightful. but are they smoke and mirrors?
I think the watches are quite powerful already but then thats not really the point, as in typical supervised machine learning scenarios you have a “training” and a “productive” phase – e.g. a classification. Lets just give an example:
The power of these platforms is the central collection of data and garmin surely has lots of it. So what could they do e.g. for trail runners? They could analyse all trail runs for all members centrally (not on the watch). They will see patterns and underlying correlations, e.g. some could be classified as “hardcore mountain” others perhaps “easy forest runs”. They will notice, that depending on trail/type pattern the correlation to HR and whatnot will be different and they might know too, how the runner performs in ideal conditions (straight, normal street run) to have a baseline to correlate with vo2max or whatever central metric they have (this is actually something suggested by Markus). So in the end you train a machine learning algorithm to classify a run and adjust your performance indicator accordingly. This is something the watch doesnt have to do.
But the trained model has to be put on your watch and every time after you run, it calculates the adjustment with all given trail and your personal parameters. That might be even possible realtime. The hard work will be done by the central servers though and the model will be only rolled out with firmware updates. There might be more sophisticated possibilities in the future but as I understand (and we learned the hard way) the watch should function pretty independent from servers in general.
In a way more simple approach, the watch is doing something like this already by putting your data in a model to calculate vo2max which in turn is fed into various pretrained models (rules rather) to give you ideas on how fast you might run a marathon. This shows too from my point of view, that the vo2max by itself is not a really relevant value to be taken litereally… its only valuable when used to predict other, more relevant KPIs or actionable insights. But then its way to simple so firstbeat has to introduce more and more workarounds/parameters.
that sounds pretty cool but Garmin won’t do it, it’s too niche right now (IMO)
I honestly expect more, given how they dominate the western market (besides Apple). They must keep up – I feel the rising Chines competition is very powerful and doesn’t have any reservations to make AI based experiments in any direction and on any scale. They have the data as well thanks to a huge inland market.
Agree with Ben, the watch could do a lot in real time as it would typically only evaluate the trained model (which was trained on server) which is basically just the ML function with a bunch of server-trained weight and bias parameters, around a few kB in size.
This is an interesting paper where they predict a cyclist’s heart rate from their power etc in real-time:
https://arxiv.org/abs/1808.00198
GAP is much better than raw pace for trail running (or any running for that matter) but still doesn’t take the ‘difficulty’ of the terrain into account, i.e. rocks vs smooth forrest ground. Potentially one could gauge a segments ‘difficulty factor’ if runners with power passed through it and looked at their ratio of pace to power. For covering a reasonable fraction of the world’s trails running power would have to be pretty pervasive though. Or one could make do with heart rate albeit probably more noisy.
They no longer restricted to doing Firstbeat calculations on the watch since they now own Firstbeat?
I dont know but I think its not needed anyway. If Firstbeat can acces all the Garmin data, they can train and optimize models before deploying to the watch. Whether they can access the data and how is a question how they legally work together – as long as they have other customers there must be some clear kind of separation.
I know it’s niche, but I’d love to see SmO2 data be incorporated into the metrics.
me too. it will never happen tho
My biggest wish would be better VO2max and better sleep detection, because so much hangs on these values… Both are fluctuating and very imprecise on my Fenix 6, esp. sleep is still not good (I’ve had a Fitbit lately for some weeks and it was totally on spot regarding sleep).
On the other side: I’ve become a big fan of the stress metric lately. I look on this more than on any other value to find out how fit my body is. Btw, I was really ill this week (belly probs, no corona) and it was fascinating to see how the stress levels where pretty much in sync with me becoming healty again over the days. I found this very helpful for a number of reasons!!
vo2/sleep…yes. exactly as you say
stress: yep I look at one regularly too. shame the data is wrong underneath 😉
ill: yep I’ve had EXACTLY the same as you over the last week. up to 9/10 for today as I feel good. the data is still wrong tho 😉
I think they just need to improve the sensor and some of the algorithms to be more accurate. The sleep tracking on the F5 is some of the worst I’ve seen across smart watches. If it’s that bad, then any metric to do with recovery based on quality of sleep is going to be useless. Also training status is a bit weird. Apparently firstbeat tells me that a long slow run is highly improving my aerobic endurance, but puts my training status into recovery or unproductive. I think some more clarity on training status and goals needs to be tailored to the individual. I’d like to see where they can take it. One thing I do find really good is the Garmin Coach training programmes. I’d like to see how they can build on those.
Unfortunately Calories burnt model of Firstbeat is used by Garmin in a mutant way. https://forums.garmin.com/outdoor-recreation/outdoor-recreation/f/fenix-5-series/141473/does-garmin-use-firstbeat-calorie-calc-method-when-biking-w-a-powermeter-paired/860509#860509
I contected Garmin and Firstbeat multiple times to correct it, the last occasion ws less than 2 weeks ago
For bad Lactate Threshold calculation are you referring to erroneous HR readings coming from chest straps?
Regarding the daily workout suggestions: AI Endurance now has the option to export your workouts directly to your Garmin so you can actually do a series of workouts optimized towards your goal event and time constraints:
https://aiendurance.com/blog/garmin-custom-workouts-directly-on-your-device
+1
Whoop has some very interesting reports on respiratory rates and increases from baseline as early indication of COVID.
Don’t have a Garmin so not sure if that’s something that could perhaps make it useful…
it’s def not an early indicator of covid!
it might be an indication of stress due to an upcoming illness
Should have caveated “from baseline as a possible early indication of COVID”.
We know that COVID impacts lower respiratory track more so than a cold or influenza (2020 Flu Season (webmd.com)). If you see significant deviations above baseline, I’d say it’s probably time for some isolation or a test.
Understand that it’s not a medical device but you seem pretty quick to dismiss it. Even if it’s just an early warning sign as you stated, probably still worth paying attention to.
yep, i’d definitely pay attention to a waking HRV reading that deviates significantly below baseline
Hi,
And above all, better hardware. I carry the simple Fenix 6. The USB port is so the slowest thing you can get on the market at all. Charging takes forever and 3 days because that only charges at 1V, maximum. Maybe even finally away from the bad Sony GPS chip.
USB always uses 5V (with USB-PD also supporting higher voltages), 1V isn’t possible. (perhaps you mean 1A?)
Looks like there’s something new from Garmin coming (more Fenix-like than Forerunner-like given the steel/titanium)
https://fccid.io/IPH-03992/Label/Label-and-Location-5072786
The full package of documents doesn’t reveal much more but does mention NFC:
https://fccid.io/IPH-03992
Digging a bit deeper, looks like maybe a few things coming:
https://fcc.report/company/Garmin-I-N-T-Inc
Cross-referencing here, looks like there’s an “Outdoor GPS Watch”, “GPS Dive watch”, and something else
fyi: (if you post multiple link it automatically goes in ‘spam’, on most WordPress sites)…its back now !!
Ah, ok, good to know. What do you make of all this?
the main watch doesn’t fit with my info.
fenix cycle is 18-24 months.
it looks like a fenix with solar/case. so that’s what it must be. so it could be a fenix 6 lte, fenix 7 (too soon but definitely possible), marq Solar, or Fenix 6 variant
my bet would be the latter.
cant be a fr9x5 solar with that case.
as far as other releases then that doesn’t surprise me. there will always ben accessories and other things that slip through without notice.
Any guess on timing? Historically, Garmin’s CES announcements have been on the first day I believe. Also, don’t know how long typically between regulatory filings and actual release but the FCC granted 90 day confidentiality on photos so I guess it could still be a while. That said, there was a thread on the Garmin forums where somebody mentioned today or tomorrow and that thread has since been deleted…
the dates on the FCC stuff is a bit of a giveaway ! #soon
yes i think the ‘something else’ is not very interesting. like a vivokids or something like that. i could be wrong. wasn’t the descent mk2 released in sept? maybe that is just a formalisation of that historic announcement. yes the other is an outdoor GPS watch
Id love to see many of the features that are currently only specific to some sports (Vo2 max for example) expanded out to other sports. For example Garmin recently released a firmware update that means the Fenix 6 recognises the Concept 2 PM5 as a power meter now so theres no real reason why indoor rowing couldn’t use that data
Id also love to see an improvement in the HR wrist strap, again for rowing the Garmin is useless and is really shown up by the apple watch which can take an accurate HR from the wrist
yep i guess that will come eventually (or sooner)
i think the idea was that rowers should be grateful with pm5 support rather than asking for more ! 😉
it has always been a mystery why garmin hasn’t treated rowers better, it must be a similarly sized market to ultra runners. ie not huge, but big enough.
Agree – it seemed pretty unusual as well since the C2 is so well established by adding support for it as a power meter youve essentially captured a large proportion of the rowing market, and all of the C2 bike and C2 Ski market – it seems a no brainer for any company providing power meters. Still, as you say good that the support is finally there even if its only on the top end right now