Is The Garmin Forerunner 965 Accurate?
Answer: Accuracy is good to excellent.
More: Garmin Forerunner 965 Review – opinion and the negative bits
This article is a full-accuracy report based on selections from a month of workouts with non-beta firmware on the FR965. It looks at accuracy across various data types including GPS/GNSS, HRV, Sleep Stages, Elevation and HR.
TL;DR – Accuracy seems to be an incidental concern for Garmin. Sure, GPS/GNSS accuracy is now market-leading but it’s taken years to get to this point. Heart rate accuracy is hit-and-miss and person-dependant…you might be lucky, but I often wasn’t.
I’ll focus more on HR as that is where most of the problems are. Let’s get cracking with some charts from each test and a quick sentence or two to highlight the key takeaway from each.
HR Accuracy Run
This is a 90-minute tempo run at a pretty consistent pace on a warm day. I’ll blame the rigid nylon band for Garmin’s discrepancy at the start, otherwise, it’s pretty good.
Here are some intervals of varying lengths on hard, flat ground. This is not good enough by Garmin (green) that’s beaten by Whoop and the Apple Watch. It’s just wrong when it counts. Again, this is not quite good enough from Garmin.
This is a 20-ish minute threshold run on hard ground. Garmin does well here and is better than Apple.
I think these were meant to be 5ish minute VO2max efforts but turned out to be a bit half-hearted. Once again, Garmin tracks too highly at the start and misses the peak towards the end.
This is one of the nicer runs on the pavement with Whoop thrown into the mix. All are great this time around.
HR Accuracy Bike
This ride is in the Surrey Hills and often includes rough roads. The purple line is the FR965 and there are clearly some discrepancies fro it to the others, although most of them seem to be at the cake stop! There is under-reporting therefrom Garmin but I’d just about take this level of accuracy as a basic log of the ride and an input for load calculations (TRIMP). For any kind of race pacing it’s not good enough, at least not on this day.
This ride is a 1-hour tempo ride on smooth roads. This is not really acceptable from Garmin and Apple Watch is superior albeit with a wobbly moment at 1hr 50.
This is another Surrey Hills ride but at a relatively hard, steady effort level for 2 and a half hours over smoother-than-average roads for the area. Garmin does a nice job here.
This is a super-easy one for Garmin on the Kickr indoor trainer. Nice job all around. When there’s no bouncing and wrist movement it’s easy to get accurate oHR.
HR Accuracy – Swim
Accuracy when swimming is troubling as Garmin’s HR curves always look plausible. Here is a typical example comprising a warmup, flat out 10 minutes and then 3x 2 lengths ramping up the intensity with paddles. The brown curve is the Polar OH1 and that is the one that is almost certainly correct even though on the very last length it flipped over on my biceps.
I haven’t yet completed any open-water swims with the FR965
These seem about right.
Note: Do this now. Walk outside your house and do a manual altitude calibration using the result from whatsmyelevation.com. This value will be used in all future runs and rides starting from home.
Sleep and Sleep Stage Accuracy
I’ve been reading so much nonsense about this from just about everywhere on the internet. Device manufacturers, bloggers, reviewers, YouTubers, scientists, the lot.
Here I wore 4 of the leading devices almost every night for a month and then spent a couple of hours compiling a spreadsheet of the time in sleep stages. It was a thoroughly thankless task as it is patently obvious that these devices are random number generators. Or at least 3 of them are. One could be correct. None of them agrees…at all.
I shall say no more than to reiterate once again that the maximum possible accuracy for sleep stages on a watch vs polysomnography is 80%. Every single device will be less accurate than that. From a cursory look, you might guess that they are all a lot less accurate than 80%. Still, each of them produces very pretty sleep stage charts; that’s what you are paying for.
HRV Sleeping Accuracy
I’ve taken a waking HRV reading almost every morning for the last month with a Poalr H10 using HRV4Training. I’ve correlated it to the nightly average HRV for Garmin 965, Oura Ring and Apple Watch 7. Here you can see the best correlation with HRV is simply the number of hours I’ve trained for followed closely by the distance I’ve covered.
That said, this is a strange result and I would have expected at least one of the devices to show a 0.8-0.9 correlation. It is what it is. Even though I’m not comparing like with like, these measures should all correlate similarly assuming the base data is correct.
All that I can think has happened is that I’ve probably drunk more alcohol than I usually do (once a week rather than once a month!) and halfway through the month I started using another sleep gadget that definitely has me waking up feeling fresher but sleeping less – that may or may not impact HRV.
Anyway, the conclusion is that you can’t trust any of these gadgets to capture nightly HRV anywhere near accurately enough. In months gone by, Oura and the Apple Watch have returned much higher accuracy results
Readiness To Train Accuracy
Many companies now produce readiness-to-train metrics.
There is no such thing defined as readiness to train, each time you see that phrase it means something different from one manufacturer to the next. It might be possible to compare one readiness number to another or to calculate which best correlates to ‘performance’.
I’ve not specifically tested for this and I overrode the Garmin values with a Stryd footpod from day 1 as I was using the 965 as a training device and wanted that metric to be correct.
I did have the GAP pace on permanent display as well. That did not respond quickly enough to true changes in grade but once stabilised seemed OK.
Running Power Accuracy (Running Dynamics)
I’ve not specifically tested for this either. And, in any case, there’s nothing to compare it to that could be accepted as correct.
Garmin’s running power is certainly a usable training metric but will only be as accurate as the accuracy of the individual sensor inputs. You can get these inputs from sensors in a variety of locations on your body ie RD-POD, chest strap and now even fully from the FR965 itself. A calibrated foot pod is the best source if you want as much accuracy as you can get, the wrist as a source will likely be the worst.
This link explains Garmin Running Power in detail and this link gives lots of test results for the FR955.
The bottom line is that there are two ‘correct’ methods of determining running power and each gives a significantly different result. Garmin and Polar use variations of one method and Apple, Coros and Stryd use variations of the other. The chart above clearly shows those similarities when looked at over data smoothed by 10s of seconds – runners looking to action 5s average power, or similar, will find very significant discrepancies between the different running power systems.
The Forerunner 965 has a battery-hungry GNSS (GPS) mode that uses every constellation of satellites AND two signal frequencies from each satellite. That means LOTS of available satellites and a way to determine individual satellite signal accuracy by looking at the time taken for the different signal frequencies to travel what should be the same distance, then eliminate the signals that have either bounced or refracted in the atmosphere. Even better, Garmin FR965 has a special SatIQ mode which powers up the super accurate mode only when it is needed.
As always, if you simply want a pretty picture of where you’ve been then just use the battery-friendly GPS-only mode. It’s fine for that purpose.
In practical terms, GPS accuracy impacts several aspects of your device
- Instant running pace
- The start and end times/positions of your Strava segments
- Your actual track under trees and in built-up areas and mountainous areas – the track could be extremely inaccurate in those scenarios
- The accuracy of navigational prompts in those tricky areas of reception
- The timeliness of the same navigational prompts. These will be (appear) delayed if you are either travelling fast or if the watch is underpowered.
GPS Accuracy – Open Water Swim
I’ve not yet braved the cold spring water to test this.
Open Water swimming should be easy for GPS as there will rarely be trees or canyon walls obscuring the satellite signal. Of course, those pesky satellite signals don’t travel underwater and that’s why many sports watch companies struggle with open water accuracy as specially modified algorithms are needed to avoid having a power-hungry beast continually searching to acquire signal.
Garmin’s history here is that they were initially producing ‘alright’ levels of accuracy. Then it ‘improved’ the algorithm and which just didn’t work ;-). Having fixed that a couple of years ago the result was market-leading. I’d expect that to be the case now on the FR965.
GPS Accuracy – Cycling
I expect that most of you will also have a bike computer. If that’s the case, Garmin lets you set the GPS accuracy level differently for each sport profile…just use GPS-only for all kinds of cycling on the watch and then whatever you want for your handlebar-based solution.
I didn’t cycle through mountainous passes but forests, open roads and urban canyons were all super-fine in max accuracy mode. Here are images from 3 different rides showing the general awesomeness apart from the 60m tunnel in Kingston…unsurprisingly there was no GPS signal to find in a tunnel.
GPS Accuracy – Running
I like this test, or at least I like the useful results it tends to produce rather than the monotony of running around the block to execute it! It exemplifies suburban running in the UK with minimal trees and houses that are probably too far away to make much of a negative difference to the GPS reception. Yet some devices struggle on this test. The killers are the sharp 90-degree bends and the need to stay on track on relatively long and straight pavements. I’ve given one of many examples of a corner, below, where Garmin just seems to nail this every time. The final image shows the Fr965 having a mini-wobbly in someone’s garden but that one slip-up is fine and only off-course by a couple of metres.
These are in max accuracy mode.
the FR965 scored an excellent 83% over a standard 10-mile test that I’ve put just about every GPS watch ever through. FR965 was nearly identical to the best-ever track from Epix 2 at 87% except the FR965 messed up one particularly long section where it ran parallel to where it should – other than that section if anything, it was very slightly superior to the Epix (it has later GPS firmware). The slip-up is likely due to the incorrect handling of signals that bounced off closeby buildings…ie precisely what multiband is supposed to correct for! Oh well.
Test methodology and FIT files are linked here.
The current level of technology across the key sports watch companies – Garmin, Apple, Coros, Polar and Suunto is the best it has ever been. If implemented properly each vendor could produce the good to excellent results we’ve seen here from Garmin.
GPS accuracy can still be improved but it must surely be about as good now as it is ever going to get
oHR is never that great for me and I can’t be alone. Other reviewers and athletes clearly get better results than me but others get the same or worse than me. This is the area that Garmin and the industry now need to focus their efforts on. That’s not just because we all want more accurate HR metrics but it is that HR is used to power pretty much all the physiology metrics and if it isn’t correct then all the physiology stuff we like to pore over will simply be wrong.
Resting HRV is the basis of a mixed bag of metrics used across the industry. It should be mostly correct but often is just wrong as all the companies, including Garmin, apply a mixture of scientifically-grounded algorithmic tweaks and then some pseudo-made-up ones to give a nice-sounding metric like readiness to train.
Finally: Content like this that shows accuracy is of very limited interest to most people. It’s mostly made for YOU and is extremely time-consuming for me to produce. If you’d like to say thank you below that would be great and if you’d like to support the work here and the site so that I can produce more of these accuracy reports then please become a supporter.
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.