GPS Heat-Map Positioning Error in Team Sports: Why Roofs and Stands Distort the Map

A performance analyst at a club with a part-covered main stand pulled up a winger's session heat map and found a dense cluster of touches sitting three meters outside the touchline, inside the stand itself. The player hadn't wandered off the pitch — he'd spent the second half hugging that flank on overlapping runs, and the GPS unit clipped to his back had simply logged him somewhere he never stood. The analyst's first instinct was to flag the unit as faulty and swap it for the next session. It came back with the same cluster in the same spot. The unit wasn't broken; it was doing exactly what a GPS receiver does when a roof overhang and a steel stand structure block half the sky it needs to triangulate a position.

This gets lumped in with ordinary GPS noise constantly, but it's a different failure with a different signature. Random measurement noise scatters a player's logged position in every direction roughly equally and averages out over a session. Roof- and stand-induced error is directional and location-locked — it pushes positions toward the open side of the sky and shows up every single time a player occupies that zone, which is exactly why it reads as a real hot zone on the coverage map rather than as noise. Below is the mechanism, two validation studies that quantified it, and the check-and-filter routine that keeps a distorted zone from quietly rewriting your coverage numbers.

A player heat map is built from thousands of individual position fixes, each one a latitude/longitude estimate the receiver calculates by timing signals from multiple satellites and trilaterating a position from the delay. Modern multi-GNSS units used in team sport (10 or 18 Hz, tracking GPS, GLONASS, Galileo, and BeiDou constellations together) record a fix several times per second, and each one gets binned into a small grid cell — typically 0.5 to 2 meters square — and the heat map simply counts how many fixes land in each cell over the session.

Two things determine how good a single fix is: how many satellites the receiver can actually see, and how those satellites are spread across the sky. The second factor is expressed as Horizontal Dilution of Precision (HDOP) — a low number means the visible satellites are spread wide across the sky, giving a well-conditioned triangulation; a high number means they're clustered in one region of sky, which mathematically amplifies any timing error into a much larger position error. Under a fully open sky a receiver typically sees 12 or more satellites spread evenly overhead, HDOP sits under 2, and horizontal position error stays in the sub-meter to roughly one-meter range. Take away half the sky and both numbers move against you at once: fewer satellites and the ones left are bunched toward the open half, which is precisely the condition that inflates HDOP and, with it, positional error.

Open training pitches rarely run into this because nothing overhead blocks the sky. A stadium with a roof over any stand, an indoor arena, or a venue boxed in by a covered structure is close to the worst case a receiver can face, combining a partially or fully blocked sky with steel and concrete that bounce signals back at the antenna after a longer delay — multipath error, which is worse than a simple blocked signal because the receiver still returns a fix, just the wrong one.

The distortion is fixed to a location rather than random across the session, so a fullback overlapping down the covered side for twenty minutes gets the same skewed fix twenty minutes running — which is exactly why it survives a full-match average as a false hot zone instead of canceling out.

Two frequently cited GPS-validity papers, from different angles, explain why this happens and how large the effect can get.

Witte and Wilson (2004, Journal of Biomechanics) tested non-differential GPS accuracy for tracking speed and position over ground and modeled how error scales with satellite geometry rather than treating GPS accuracy as a fixed number. Their central finding was that positional and speed error tracked HDOP directly: under good geometry, horizontal error stayed small and speed estimates were reliable, but as HDOP rose — the exact condition created when a roof or stand blocks part of the sky and pushes the remaining satellites into a narrower cluster — error grew several-fold rather than gradually. The limitation: their testing used early-2000s non-differential single-constellation receivers, well before modern multi-GNSS chipsets tracking four constellations at once, which have more satellites available to compensate for a partially blocked sky and soften — though not eliminate — the relationship.

Duffield, Reid, Baker, and Spratford (2010, Journal of Science and Medicine in Sport) took the applied angle, validating GPS-measured distance and speed against a criterion timing system for court-based, direction-change-heavy movement in a confined, partially enclosed venue rather than an open field. They found measurement error running substantially higher than figures typically reported in open-field GPS validation work, enough that the authors flagged confined, enclosed environments as a distinct accuracy problem rather than a minor variation on outdoor use. The limitation: the study used a single 5 Hz GPS model, since superseded by 10–18 Hz multi-GNSS units, at one venue and one sport, so absolute figures don't transfer directly to a modern unit — but the core finding has held up as newer units have been tested under similar conditions.

Read together, the two papers explain the same failure from opposite ends: Witte and Wilson show why blocked sky produces error, and Duffield's group shows what that error looks like inside a confined, partially covered space. Neither paper is about heat maps directly — distortion is the downstream consequence of the mechanism both measured.

You don't need a survey-grade receiver to find this out. Most GPS units used in team sport log HDOP and satellite count alongside every fix — you need to look at that data instead of the heat map alone.

Start by standing a unit at midfield, as far as possible from any roof or stand, and let it log for two minutes. Note the average satellite count and HDOP — your clean baseline, since satellite availability shifts somewhat with time of day. Then walk the unit in five-meter steps toward the covered side, logging both numbers at each stop. In a clean zone they barely move; near a roof or stand, satellite count drops and HDOP climbs steadily as you close the distance, and that gradual, distance-correlated shift — not a sudden jump — is the signature of geometry-driven error.

Repeat toward every structure near the pitch — the stand, the floodlight towers, the video screen — and mark where each first pushes HDOP past 2. Re-check periodically: a concert stage, a newly installed screen, or a seasonal shift in overhead satellites can move the distorted zone without anything about the unit changing.

Once you know where the clean and distorted zones sit, three changes fix most of the problem without new hardware.

First, filter by fix quality before generating the heat map, not after spotting something odd. Most tracking software lets you set a minimum satellite count or maximum HDOP as an inclusion threshold; HDOP under 5 with at least 7 satellites, based on the walk-out test above, drops the worst fixes out of the aggregation instead of averaging them in.

Second, treat total distance and speed separately from fine-grained position clusters near a distorted area. Distance and average speed over a full session forgive occasional bad fixes far better than a heat map does, since a heat map's entire purpose is showing exactly where a player stood — so a club with one covered stand can usually still trust that session's load metrics while treating the positional cluster near the stand as unreliable.

Third, when an accurate read genuinely matters near a compromised zone — confirming a fullback's real overlap distance down the covered flank, say — cross-reference against video or a local positioning system (UWB- or radio-based, independent of satellite geometry) rather than forcing more accuracy out of GPS than the physics allows. Filtering can stop a bad reading from contaminating the report; it cannot recover accuracy the geometry never had.