Garmin Running Power accuracy improved with a new tweak
Source: doesn’t want to be disclosed
Garmin Running Power is a cousin of Stryd. Both calculate running power based on different algorithms and Garmin performs the calculation with sensors on the wrist whereas Stryd does the calculations based on its footpod and sensor package.
I won’t dwell again on the issues with Garmin’s Running power but its biggest problem is how to handle wind. Basically, it can’t. So it uses the most up-to-date weather report your app has before you start your run. Localised wind conditions will inevitably mean the wind component of the calculation is wrong. Wind is an important force and there is just no getting around this issue for Garmin, Polar, Suunto and Apple unless they actively and continuously measure wind.
Sadly, that hasn’t changed! (Sorry to get your hopes up). But what seems to have changed is that Garmin has, or is about to, tweak the power calculation to account for your stamina.
I don’t quite understand how this would work.
By comparison, cyclists will know that “power is power”. Which it is…but it’s also not! True, the work done is fixed but the impact on your body is variable. 300 watts for 10 minutes has quite a different physiological effect than 300 watts for 10 minutes AFTER 2 hours of hard riding or running. As a cyclist do you want to measure the actual work you are doing or how hard the work feels? Probably the former.
Maybe with Running Power, Garmin hopes to incorporate physiological difficulty into the calculation?
Garmin uses the GOVSS (Skiba) model which DOES seem to allow for the inclusion of difficulty.
GOVSS stands for Grade-Adjusted-Pace Optimal Velocity Smoothing Score. It measures the effort of a runner, taking into account various factors such as gradient, running speed, and environmental conditions.
Stryd, Runscribe and Apple probably use the EESA model, which is different.
So, I’ve said that Garmin plans to incorporate stamina into running power. Let’s quickly understand what stamina is in the Garmin context.
Garmin’s stamina metric is already available for cyclists and I go through phases of ‘quite liking it’ and ‘not liking it’. It is based on multiple factors including distance and heart rate. If we assume that Garmin will use the same stamina calculation as a tweak for running power then surely a logical conclusion is that everything which affects stamina also affects running power. Heart rate affects stamina according to Garmin and caffeine patenly affects heart rate according to me. Thus caffeine must affect running power.
That sounds wrong to me.
Take Out
This could be one of those tweaks that appears to work in some situations but blatantly fails in others.
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.
Mmmm, interesting, confusing and feels initially unnecessary.
Is this at all related to Coros’ “Effort Pace” ?
On a related note, I am finding Stryd Zones temp adjusted power quite useful, though still find the power targets a bit high when it’s 25C
Thanks Will
I’d completely overlooked coros when I was penning this.
to answer: IDK, but that’s certainly a good point
effort pace becomes ever more nebulous than running power the more you add new tweaks. honing in on runnin power must be targetting a ‘corret’ figure at some point even though no-one knows what the correct figure is, for effort pace surely there will be never-ending tweaks for nuances in people’s physiologies.?
Hot take: Running power is pretty much equivalent to grade-adjusted pace (maybe plus wind resistance) and does not account for compounding factors like road surface effects (smooth tarmac, brick, rough cobble, camber, gravel, grass, mud, sand, etc) or shoe mechanics. Also completely fails at pitches steep enough to involve hands on knees or scrambling and use of poles.
It’s really unfortunate that this metric is labeled power because it is an indirect estimate based on motion capture and inference and is very unlike cycling power which is relatively simple, captured directly, and not subject to all of these confounds (but still manages to have accuracy problems with “simple” strain gauges).
power is work-done, right?
Power is the *rate* of work done or Work / Time. Therefore the amount of work done during an activity is (Average) Power * Time.
I agree that running power is totally unlike cycling power in the sense that the latter is measured and the former is estimated. That’s why Stryd power ≠ Garmin power from wrist ≠ Garmin power from chest strap / RD pod.
At least with GAP, the value is related to “real” pace (which has a physical interpretation), so values from different measures of GAP can be compared (even though they probably won’t match: Garmin vs. Strava. vs. Runalyze GAP are all different). Doesn’t seem possible to compare values of running power from different vendors.
cycling power is not measured in the finnickety sense. It will be inferred from the defelction of a strain gauge or somthing similar. Pace isn’t measured but distance and time are! Sorry I’m being annoying.
running power should be comparable. There are two models and those operating each model should be the same. Clearly they aren’t. why? either the calculation or inputs are different, if the inputs are different it will come down to data capturing differences I would have thought,
then, even though the two models have different calcualtion they are still measuring the same thing and they should correlate well, at least at some level.
GAP uses Elevation deltas (easy to measure accurately), distance (quite easy to measure accurately in most cases), grade (often quite difficult to measure), and sometimes: surface type (difficult to measure), weight (easy to measure), fitness level (sometimes used, hard to measure), aspects of weather (often tricky to measure).
I guess scientists want precision but I’m more forgiving and want actionability. One (of two) problems I’m having with a bike power meter at the moment is that there is a delay in the display of the end result. the PM is probbaly very precise but it’s much less actionable when the data is, say, 5 seconds old.
There’s a similar thing with optical heart rate which can be lagged by multiple seconds. only dcrainmaker ever mentions that. I adjusts my presentational data sets to correct for it (which is probably wrong) but how many runners who train by heart rate even realise the error in their data? (I bet it’s less than 1%) of course there is a delay anyway in the physiological reaction of heart rate so it obscures that additional delay
Garmin fixed the complaints about weather on a watch isn’t very local weather.
(On my forerunner 965 local weather is an airport 30k away).
But garmin got a brilliant solution. Just remove the location info from the weather widget. Now you cannot see anymore how inaccurate the weather shown is.
“ As a cyclist do you want to measure the actual work you are doing or how hard the work feels?”
I don’t want to know how hard it feels, I want to know how hard I have to work. And of course when I’m chasing PR’s it feels horrible at the end.
Yes, this sounds like garmins marketing department demanding a new sexy (bs) number to attract data collectors.
I agree. I don’t want power being affected by fatigue or stamina. I ran with power yesterday and had several spikes of 350-400 when my target was z2 and 260. So all over the board. I am not convinced I can race on power but am trying it out and assessing usefulness. Garmin – Fenix 7 BYW)
the only way i can see stamina being relevant to true power is if they are adjusting for the reduction in efficiency as you fatigue. when fatigued you likely exert more power relative to your actual pace than when fresh due to power not actually pushing your forwards. i can’t imagine they can do this accurately as everyone will be different but it seems not unreasonable to apply some average adjustment – everything about running power (from all vendors) is full of assumptions anyway
yes !