Data has value and this data is particularly valuable to Valve and its competitors. So Valve share the raw data which has gross numbers but they don’t share the useful data. The useful data is the processed data - the corrected and weighted data based on the other information Valve has about its users and install base. That way can weight this months survey responses to expected proportions of the whole user base and see actual user wide figures and trends.
What Valve shares is akin to a polling company sharing the raw data from the people who completed a polling survey. It’s relatively meaningless and even misleading until they correct the data to weight it to make it representative of the whole population.
So this month there were 1.15% fewer linux users in the survey pool, not 1.15% fewer linux users overall. They will correct the data to see an actual proportion of Linux users. For example: they have data on every use of Proton and every install of Linux versions of software; and how many times each user installs a game (occasional vs heavy users). They don’t share that but they can use that to help correct the data and get much more accurate picture - one they don’t share as it gives them a commercial advantage.
I’m no survey expert, but why doesn’t Valve get a mixed distribution from all countries?
No clue, Think it’s one of those ‘Worked well enough in the past’ things they just never bothered to re-work.
I personally wish they would update it to be more globally accurate but it’s Valve so.
Data has value and this data is particularly valuable to Valve and its competitors. So Valve share the raw data which has gross numbers but they don’t share the useful data. The useful data is the processed data - the corrected and weighted data based on the other information Valve has about its users and install base. That way can weight this months survey responses to expected proportions of the whole user base and see actual user wide figures and trends.
What Valve shares is akin to a polling company sharing the raw data from the people who completed a polling survey. It’s relatively meaningless and even misleading until they correct the data to weight it to make it representative of the whole population.
So this month there were 1.15% fewer linux users in the survey pool, not 1.15% fewer linux users overall. They will correct the data to see an actual proportion of Linux users. For example: they have data on every use of Proton and every install of Linux versions of software; and how many times each user installs a game (occasional vs heavy users). They don’t share that but they can use that to help correct the data and get much more accurate picture - one they don’t share as it gives them a commercial advantage.