Steam Survey Results For August Show A Linux Dip After A Very Exciting July

[phoronix]

Phoronix: Steam Survey Results For August Show A Linux Dip After A Very Exciting July

In July the Steam Survey results pointed to a half percent jump in the Linux gaming marketshare, taking it to nearly 2% thanks to the success of Valve's Steam Deck that runs their Arch Linux powered SteamOS platform. After the big boost in July you are probably wondering what happened in August... Well, a few minutes ago the new Steam Survey monthly results were published...

https://www.phoronix.com/news/Steam-Survey-August-2023

[sarmad]

That's a smaller dip than I was expecting to see, and it's still above macOS numbers (which I wasn't expecting), so overall it seems to be a healthy upward trajectory.

On a side note: Valve needs to start selling those Steam Decks in China.

[peterdk]

Isn't it the case that the survey is random regarding selection of users? It's not every user, but a subset. So how reliable is it for comparison?

[luno]

Is this survey automatic, or you have to fill manually ?

[Sonadow]

Linux users when Steam survey shows slight increase in Linux percentage: Great! Linux gaming is gaining momentum! Windows we are coming for you!

same Linux users when Steam survey shows drop in Linux percentage: the survey is random and therefore not reliable.

[sophisticles]

If the sampling were not random, the survey would lose meaning. Assume they survey 1000 users this month and the same 1000 users next month, you would expect the results to be the same.

You need to sample random users so that you can draw an accurate average use over time. The more data you have and the longer the time frame, the more accurate your estimation becomes.

So what we can do is say in January, Linux usage was X percent, in February it was Y percent and so on and we can get pretty good idea of what overall Linux use among Steam users is.

[NeoMorpheus]

same Linux users when Steam survey shows drop in Linux percentage: the survey is random and therefore not reliable.

Personally, i have always said its not reliable, but its the closest thing we have to a proper market pulse.

That said, the only way this can be reliable is if every sing client was counted.

Also, one thing that would help is real numbers, in this case, 1.9% of how many? Better yet, 1.9%=1 something millions.

That would give everyone a more concrete idea of the amount of people in Linux.

Lastly, I moved my gaming pc from w11 to ChimeraOS, how its that counted, via survey, automatically added, not counted at all?

[CochainComplex]

the best autocome is to make each month a sample of x out of y persons. You don't exclude previous once. By statistical chance its unlikely anyway. If you use 1/1000 which might be enough for the statistics for a entire userbase of a couple of millions the chance to be selected twice in a row would be 1/1000000 so you dont need to preselect. Besides the survey is very objectiv in its nature. So statistical errors are very low. there are no individual questions like:

Do you feel to have more an AMD or a Nvidia card? - choose a number between 0(AMD) and Nvidia(5).
Is you OS more windows or more Linux? choose Mac if you are not sure

The only factor which can skew the overall statistics if you have different nations included where a certain product e.g. steam deck is not available in a country and you try to project from this to e.g. western market - as already mentioned above in the comment.

[mirmirmir]

Who's that boomer at the bottom of the article? A cowriter? Where's mike?

[RejectModernity]

Who's that boomer at the bottom of the article? A cowriter? Where's mike?

Oh no