Test process and summary results
For any meaningful analysis to be carried out, it is essential that the raw data you capture initially is as accurate and repeatable as possible. That's why we created 3 tests to measure the GPS accuracy with movements fully representative of games that involves rapid change of directions and sprints.
We tested 8 GPS units by walking and running through these grids and extracted two key elements:
- how accurately does the GPS replicate the expected shape by looking at the GPS traces on a computer
- what is the difference between the reported distance and the known reality, expressed as a percentage (e.g. 80% indicates for that the GPS measured 80m out of the 100m length of the grid).
The GPS we tested were:
For each GPS we had 18 results (3 grids x 2 paces repeated 3 times) and the table below shows the average of these results. These tests were conducted over 2 days that had similar weather. On day 1 we had sprintZ GPS + 3 other GPS strapped to the arm of one single person going through the grid. On day 2 we repeated with sprintZ GPS + 4 other GPS.
You can see that our own sprintZ GPS is only 3% short of the exact grid distances while all other 7 GPS are between 20 and 80% off. For full details, keep scrolling....
Detailed results and traces
On all traces, check the low sampling of most units: the more arrows make up the traces, the higher the sampling and the better potential for fidelity to the speed. The colour on the GPS traces represents the speed variation: when there are too few samples, the speed information between two distant samples will miss lowest and highest speed and show a dull average medium speed.
All self explanatory beyond the general comment above.
We were puzzled for a bit as to why the sprintZ shape didn't look quite perfect. But the shape drawn actually makes complete sense... The unit was worn on the left arm so all right turns are more rounded (arm on the outside doing a bigger loop) and all left turns are sharper (arm on inside).
This was a fun one, in a way looking to "trick" the GPS by building up some momentum one way, doing a short shuttle run, then going back the other way. All units are fairly good detecting the straight lines but not many pick up the shuttle runs at the ends.