The blue dot is the most trusted object in modern navigation and it is a probability distribution wearing a costume.
Understanding roughly how big that distribution is, and what makes it bigger, is the difference between a tool you can rely on and a tool that will eventually tell you something confident and wrong at the worst possible moment.
How the position is worked out
A satellite broadcasts two things: where it is, and what time it thinks it is. Your receiver picks that up, compares the timestamp against its own clock, and works out how long the signal took to arrive, which gives it a distance to that satellite.
One distance puts you on a sphere. Two puts you on a circle. Three narrows it to a couple of points. But your receiver's clock is a cheap one and is not synchronized with the atomic clocks on the satellites, so there is a fourth unknown, and a fourth satellite is what lets it solve for position and time together.
That is why four is the minimum and why more is better. It is also why the whole system is really a very precise clock experiment that happens to produce coordinates.
Modern devices do not use one system. They combine the American constellation with the Russian, European and Chinese ones and sometimes regional systems on top, which is why the term is now GNSS rather than GPS. More satellites visible means better geometry and a better solution.
How GPS Works: From Satellites to SmartphonesWhere the error comes from
Several independent things degrade that solution and they behave differently.
The atmosphere slows the signal, and by an amount that varies. The ionosphere is the larger effect and it changes with solar activity and time of day. Receivers model it, and models are approximations. A receiver that can listen on two separate frequencies can largely cancel this error by comparing them, which is why dual-frequency capability is the single biggest step up in consumer positioning accuracy in recent years.
Satellite orbit and clock errors are small and are corrected by broadcast data and by augmentation systems, which transmit corrections that meaningfully improve accuracy over the areas they cover.
Receiver noise is the cheap electronics doing its best.
Geometry is the one people never think about. If all the visible satellites are clustered in one part of the sky, the intersecting spheres meet at a shallow angle and the solution is imprecise even when every individual measurement is good. Surveyors call this dilution of precision, and it is why accuracy varies through the day at a fixed location as the constellation moves overhead.
Multipath, the one that actually gets you
The error that dominates in exactly the places outdoor people go is multipath.
A signal that travels straight from the satellite to your receiver took a known path. A signal that bounced off a cliff, a wet trunk, a rock face or the side of a building took a longer one, and arrived late. The receiver has no way to know it bounced. It treats the late arrival as a longer distance, and the computed position moves.
This is why your track log wanders while you sit still in a narrow canyon, why the blue dot jumps around in dense timber, and why positions taken against a rock wall are systematically worse than positions taken in the open. Deep drainages are the worst case: fewer satellites visible, and the ones you can see are more likely to be reaching you by way of the opposite wall.
What is Multipath?What the accuracy circle actually is
The number your device shows, and the circle it draws, is the device's own estimate of its own error, computed from the signals it is receiving and the assumptions its software makes.
It is a useful indicator and it is not a guarantee. It is also not consistently defined between platforms, so a figure from one device is not directly comparable with a figure from another. And it describes the expected spread rather than a boundary: a reading with a stated accuracy of five meters can be further off than that.
The right way to use it is as a relative signal. When the circle grows, your position is getting worse, and you should trust the dot less. When it shrinks, the opposite. Treating it as a specification is a mistake; treating it as a mood indicator is about right.
Vertical is worse than horizontal
Consistently, and by a substantial factor, and it surprises nearly everyone.
The reason is geometry. Satellites are distributed across the sky above you but there are none below you, so the vertical component of the solution is always constrained from one side only. Horizontal error benefits from satellites on opposite sides of the sky; vertical error does not have that option.
The practical consequence is that satellite elevation readings are the weakest number your device produces, and that a barometric altimeter, which many watches and some phones have, is considerably better for relative elevation change even though it drifts with weather and needs calibrating.
If you are counting elevation gain, the barometric number is the better one. If you are trying to establish what elevation band you are in for an avalanche forecast or a snow line, calibrate at a known point first.
Why your watch says you walked further
Track logs have a systematic bias and it always runs the same direction.
A receiver sitting still does not report a still position. It reports a cloud of positions scattered around the true one, and if the device is naively adding up the distance between consecutive fixes, it accumulates distance while you eat lunch. Over a day with several breaks, in terrain where the error is large, that inflation is real.
Devices apply smoothing and filtering to counter this, with varying success. The result is that recorded distance is an estimate with a bias toward being too long, particularly in canyons, in forest and on days with a lot of standing around.
The related trick worth knowing: if you want a good coordinate for a specific spot, most devices and apps offer waypoint averaging, which takes many fixes over a period and averages them. Standing still for a couple of minutes while it does that produces a substantially better position than a single instantaneous fix.
Your phone is not only using satellites
Worth knowing, because it explains behavior that otherwise looks like magic or malfunction.
A phone's location is a fused estimate. It combines the satellite solution with known positions of nearby wireless networks and cell towers, with the accelerometers and gyroscopes that track how you have moved since the last good fix, and with map matching that snaps you to a road when it thinks you are driving.
Most of the time this makes the position better and faster. A phone that has network access also gets assistance data telling it which satellites to look for, which is why a first fix indoors on wifi can be nearly instant while the same phone in airplane mode in a remote valley takes minutes.
It also produces two effects that confuse people. In a place with no signal and no recent fix, position acquisition is genuinely slower than you expect, so turning the phone on at the moment you need it is the wrong time. And in places where the wireless database is wrong or stale, the fused position can be pulled somewhere the satellites alone would not have put it.
The practical response is to open the map and let the device settle before you need it, rather than pulling a cold phone out of a pocket in a whiteout and expecting an answer.
What this means for boundaries
This is where the abstraction meets the consequence.
A device reporting a position with several meters of uncertainty, standing near a line whose own position carries error from an old survey and a coordinate transformation, cannot tell you which side of that line you are on. It can tell you that you are near it.
We laid the boundary side of this out in why map boundaries are wrong, and the ownership data behind those lines in how to read a parcel map. The receiver error described here stacks on top of all of it.
Nowhere is that clearer than corner crossing, where the entire question is whether someone occupied a specific point. A phone is not the instrument that settles that, and treating a screenshot as evidence of precise position misunderstands what the screenshot is.
The habit that follows is simple and it is the same one as everywhere else in this category. Give yourself margin. If it is close, it is close, and being right by three meters is not a position anyone wants to defend.
What "good enough" actually looks like
It is worth being concrete about when the error matters and when it does not, because the answer is usually that it does not.
Finding a trailhead: the error is irrelevant. Following a recorded track: irrelevant. Knowing which drainage you are in: irrelevant. Estimating how far you have come: mostly irrelevant, allowing for the bias toward long. Getting back to a car in fog: fine, since a parking lot is much bigger than the error.
Relocating a specific fishing spot on a big flat: this is where it starts to matter, and where averaging a waypoint pays for itself.
Confirming you are on the public side of an unmarked boundary: not good enough, and no consumer device is.
Establishing a survey position, or documenting where you stood for any purpose where somebody may disagree with you later: not the right tool at all.
The list is worth having in your head because the usual failure is not people being fooled by the error. It is people applying survey-grade expectations to a device that never claimed them, and then being surprised. For the great majority of what anyone does outdoors, a few meters is a rounding error and the dot is exactly as good as it looks.
The failure modes that are not accuracy
Two ways the system stops helping that have nothing to do with error size.
Battery. A cold phone loses charge fast, and satellite positioning is one of the more power-hungry things it does. A dead device has infinite error, and cold is the condition where you are most likely to need it. Keep the phone warm, carry a battery, and do not rely on a single electronic device for navigation. That whole argument is navigation without signal.
Interference. Satellite signals arriving from orbit are extremely weak by the time they reach the ground, which makes them easy to disrupt. Interference happens near some infrastructure and in some regions, and the symptom is not always an obvious failure; sometimes it is a confident wrong position. If your device insists you are somewhere the terrain says you are not, believe the terrain.
The gear
A device with dual-frequency GNSS support, if you are buying. It is the single largest accuracy improvement available to a consumer and it helps most in the canopy and canyon conditions where phones have always been worst.
A barometric altimeter, in a watch or a handheld, for elevation that is better than the satellite solution and for the elevation change numbers that actually matter.
A map and compass, and the skill to use them, because they fail in ways you can see rather than in ways that look like success.
A battery pack, kept warm, plus the discipline to not have every navigation tool depend on the same charge.
Waypoint averaging, which is a feature rather than a purchase, for any coordinate you actually care about.
A habit of checking the reported accuracy before acting on a position, so that a large circle makes you cautious rather than going unnoticed.
And terrain association, which means continuously matching what you see to what the map shows, so the device confirms your understanding rather than replacing it. That is what reading a topographic map buys you and it is the skill that catches a wrong position immediately.
The short version
Four satellites minimum, more is better, and geometry matters as much as signal quality. Multipath is the error that dominates in canyons and timber, which is where you are. The accuracy circle is an estimate, not a boundary. Vertical is worse than horizontal by a lot. Track distance is biased long. And none of it is precise enough to place you confidently on one side of a legal line.
The dot is very good. It is just not the certainty it looks like, and knowing the size of the uncertainty is what turns it into a genuinely reliable tool.
It is also worth keeping some perspective about what has happened here. Within living memory, knowing your position to within a mile or two anywhere on earth required a sextant, a chronometer, clear sky and a competent navigator, and it took a while. Now it is free, instantaneous, and in everyone's pocket. The complaint in this article is not that the system is bad. It is that it is so good that people stopped asking how good, and the answer to that question is what decides whether you use it well.
