statCounter Has Outdated and Unfit-for-Purpose User-Agent Parsers
This chart does not tell anything meaningful:
WE'VE often cited statCounter here, but we occasionally point out the caveats. For instance, statCounter does not at all recognise all sorts of operating systems, especially Chinese ones. It's like they simply do not exist to statCounter. We can assume they're all classified as "other" or "Unknown" (see above). Maybe some are mis-classified.
Any time we mention statCounter it's because it lacks competition, except perhaps private and proprietary alternatives, which would have vastly smaller data sets to process (it's about even distribution rather than size, but judging trends by one single site/company wouldn't work either).
In some cases, the poor recognition or parsing or classification of User-Agent strings means that trends get distorted very badly. In the case of India, "Unknown" rising from 7.5% to almost 23% in a few months (as shown above) can distort global trends. Indians are the world's largest population, so it doesn't quite "average out"... █