Quote Originally Posted by Eyelet99 View Post
Does anyone know if this is correct and Garmin are actually using the CEP statistic?
It was historically accepted this was the case and will certainly be same or similar given the numbers I've seen compared back to back with professional grade equipment. You could check with Garmin support but either way it will be very rubbery in a Garmin/consumer device for a few reasons.

Firstly there will always be some error and a high confidence number would be off-putting and impractical for a typical Garmin customer. My professional Trimble mapping equipment and software is selectable for confidence readings - and even that only has a 68% default, and it is also selectable for 95% or 99% (and can be further selected to consider either an autonomous reading, real-time corrected or post-processed). However the 99% number is significantly larger and can be useful for visualisation and altering strategy when collecting data but would not be practical for general purpose use.

Secondly consumer devices also have limited rejection of bad signals so I would be wary of reading to much into the number. Pro grade gear is configurable to ignore satellites below specific angles above the horizon (more atmospheric error), above specific noise levels (SNR), or poor geometry (DOP) etc. And the hardware and software is also designed to reject multipath (signal bounce that lengthens the distance and adds error to the position), and a large high grade external antenna on pole can perform much better e.g. stronger multipath rejection, avoid your head blocking the signal etc. This is all good for making the data collected accurate but not so good when you simply want to find a large campsite and everything is stopped waiting for better signal.

So the Garmin's are optimised for practical productivity rather than precision and the confidence strategy will reflect that e.g. it's rubbery but works most all of the time because that's what you want.

They will certainly have a software strategy that considers some effects but things like multipath are difficult detect and deal with because the signals can still look ok and for most consumers it will be present because we typically use the devices in forests or amongst buildings etc. So regardless of what the device is telling you regarding the accuracy, a lot of the time the situation will be a lot worse and neither Garmin or you will know by how much. So there would be no point in them trying to give you are higher confidence number, 50% is totally practical in the use case.