Pixel Watch 5: The Most Accurate GPS Yet, Not in the Water

Google says it built the most accurate GPS track on the market, more accurate than the Garmin Fenix 8 Pro and the Apple Watch Ultra 3. Ray Maker of DC Rainmaker flew to New York to check it, a place where every GPS watch usually gets tripped up by canyons of concrete and glass. The findings are strong. Before reaching for a wallet though, here is what this fix does not cover.

The correction happens after the workout, not during it

This is the key to understanding the whole story. The Pixel Watch 5 does not compute position better in real time. It corrects the finished file after an activity ends, in the cloud. Google uses three things to do it:

3D building models. Google has the whole world mapped in three dimensions and knows the height of practically every building, bridge and ridge. Knowing what stands next to a wearer, the algorithm can work out which satellites the watch physically could not have seen, in other words which signals are reflections off glass and metal. Those reflections, multipath errors, are the main cause of a track that appears to run through buildings on a map. Google discards the suspect signals and recalculates the file from scratch.

DGNSS weather correction. Reference stations report actual atmospheric conditions at a given location, and Google applies a correction to the file. Rain, thick cloud cover and storms degrade reception.

Cloud sensor fusion. Watches have combined GPS with an accelerometer and gyroscope for years. Google does this again in the cloud, now armed with the 3D models and the weather correction.

What matters here is what is missing: this is not snapping the track to the road. Maker deliberately tried to trip up the algorithm. He walked under a bridge, turned back after 25 metres, crossed a street and walked back up. Google reproduced it to within a metre. Apple decided nobody does that and straightened the track. On a loop drawn in the middle of a closed street, the same thing happened: Google hit the loop, Apple pushed it onto the sidewalk instead.

How it performed in practice

In the field, Google won the large majority of segments. On turns, where the Fenix 8 Pro landed on the wrong side of the street, the Pixel sat on the correct sidewalk. Around Times Square, Garmin mostly held to the middle of the road, Apple wandered into buildings, and Google stayed on the correct side.

It is not a clean sweep though. There were segments where Google threw the track into the wall of a building while Apple correctly tracked the sidewalk. Maker makes a point of highlighting this himself. The algorithm is not flawless, only clearly better than the competition.

On the bike, differences were much smaller. At 30 km/h a watch has more confidence in direction, so the raw track already comes out reasonable. Under an overpass, with no sliver of sky overhead, all three devices performed similarly. On a bridge, only Apple landed in the correct lane; the Pixel placed Maker on the railway tracks.

Google bases its claim of route mapping twice as accurate in difficult terrain on a comparison against reference backpacks accurate to the centimetre, the same backpacks Garmin and Apple use. They measure the average distance of every track point from ground truth. Maker notes, reasonably, that the multiple depends heavily on location. In Manhattan, hemmed in by tall buildings, the number comes out higher; over open, flat terrain it would likely come out much lower.

The catch: open water swimming is outside the system

This is where the article gets into something not found in the press materials. Google’s post processing works for walking, running, hiking and cycling. It does not cover the remaining modes, including open water swimming.

That is exactly the discipline where a GPS track is most scrambled, because the watch loses signal underwater every few seconds and only catches a position when the wrist breaks the surface. Anyone irritated that a watch shows 2140 m for a 1900 m lake swim, tracing a route shaped like a drunk caterpillar, will find the Pixel Watch 5 does nothing about it.

For a triathlete, that changes the proportions. The improvement covers two thirds of a race, while the technically weakest segment stays exactly as flawed as before.

Prices and what else is new

Preorders opened on 12 August and general sale started on 20 August, through the Google Store and the usual carrier and retail channels.

In the US this is the first price increase in the line’s history: $399 (41 mm Wi-Fi), $499 (41 mm LTE), $429 (45 mm Wi-Fi) and $529 (45 mm LTE). That is $50 more than a year ago on the 41 mm and $30 more on the 45 mm.

Specs: an AMOLED display at 3000 nits brightness, an LTPO Actua 360 panel at 1.2 or 1.4 inches with a 1 to 60 Hz refresh rate, Wear OS 7. The processor is 12% faster, RAM is up 60%, and Google claims the system runs 20% faster overall.

On the training side, the watch adds a workout builder in the app, set and rep counting, voice cues and a rest timer between sets. Google also claims 15% more accurate sleep analysis and adds a smart alarm plus a new nighttime breathing quality score.

A separate bucket is Health Guardian, an umbrella for health features. It includes detection of a prolonged drop in blood oxygen, a blood pressure trend report and an insulin resistance report, both slated for autumn. None of them require a subscription, and none of them carry medical certification. Maker is openly sceptical of the last two, comparing Google’s approach to Whoop’s: pushing the boundary of what still does not count as a medical device, without the certification to back it up.

Battery life

This question decides everything over a long distance. Google states up to 30 hours of use for the 41 mm version and up to 40 hours for the 45 mm in normal use.

With GPS running, that drops to 8 to 10 hours. On a 70.3, that is margin to spare. Over the full distance, there is no margin left. An amateur finishing an IRONMAN in 12 or 13 hours simply will not carry a full recording to the finish.

Will older models get this?

Probably, though Google will not confirm it officially. The company repeats its standard line about trying to bring new features to earlier devices. Nothing technically stands in the way, since the Pixel Watch 4 already has multi-band GNSS. The antenna differences between generations are cosmetic.

The scale of this jump is similar to what Garmin did by introducing multi-band GNSS in the Fenix 7. It took the competition years to catch up. That earlier jump worked from the start, though, in every mode and without a cloud. This one happens after the fact and skips the water.