[boomboomsubban]

Statcounter's data is based off hits to a certain websites. Not unique hits but hits. It's really hard to see their data as worthwhile from my perspective.

This is already desktop stats, a market that isn't really growing. Then figure they're unlikely to be partnered with anyone like Facebook, Tiktok, or Discord, the sites that 90% of users spend most of their time on.

Think of the remaining internet they might have deals with, does that seem representative of users as a whole? Now add in that it's based on hits, with no reason to suspect they're effectively filtering out bots. I don't see the value of such data.

[hansvm]

> filtering out bots

For all but the most advanced of bots, the only such filtering comes from something as simple as a user-agent or other maybe some other header fingerprinting. What percentage of bots even bother to switch from the default of "python-requests/2.25.0" so that they might be detectable as Linux?

[boomboomsubban]

If it's 1% of bots that switch, it could still be fairly meaningful when tracking hits.

[7e]

Yes, their data is garbage. macOS losing 20% share in one month. ChromeOS losing 50% in one month. A huge "unknown" share. Lots of volatility. This is just the Linux faithful seeing what they want to see.