UCR Star reveals Google Maps poor quality in Beijing, China (Or may be not!)

We used to hear stories about some mapping applications failing due to poor data quality. However, it makes a big difference when you find major failure of the most prevalent mapping application in the country with the biggest user base. The story is that while exploring some datasets in UCR Star, I found the following view of building in Beijing, China.

Update (7/31/2019)

Turns out, there is a reasonable explanation for that. China uses its own coordinate reference system (CRS) GCJ-02, which obfuscates the location by adding an apparently random offset to each point. The shift that we see in the map is because Google Maps View applies GCJ-02 correctly while all the others do not. In other words, the building dataset from OpenStreetMap (OSM), the OSM view, and Google Satellite View, all use the standard WGS-84 so they match perfectly. Google Maps View uses GCJ-02 so it appears to be shifted. Google Maps uses consistent CRS on google.cn/maps.

The image looks completely wrong. It shows building in the middle of the sea. Explore it yourself. To describe this situation, the black drawing is the buildings dataset extracted from OpenStreetMap and the base layer in the background is provided by Google Maps. First, I trusted Google Maps more and thought that OpenStreetMap data might be outdated, especially that it is dated to 2015. When I replaced the base layer with the up-to-date OpenStreetMap image, it made more sense.

There seem to be more buildings in OpenStreetMap that are not yet in the buildings dataset, probably due to the outdated dataset. But the existing buildings match well with this background. Still, which data is more accurate, Google Maps or OpenStreetMap? To decide which one is correct, I replaced the baselayer with satellite images from Google. The resulting image is below.

Now, it is clear that the data from OpenStreetMap matches perfectly with satellite images. This might indicate that Google Maps data is outdated and completely inaccurate. You can further explore this with the official Google Maps website and switch between map view and satellite view. You can also investigate this area with the following view. If you found other interesting patterns, please share them in the comments below.
Update (7/31/2019) See the same location correctly on Google Maps Chinese version.

Links

UCR Star: A Star is Born at UCR
Apple Maps drove me over water
Restrictions on geographic data in China