Steam On Linux Marketshare Edges Tiny Bit Higher In November

[F.Ultra]

They have OS usage (its included in every userAgent request),
Number of people playing each game

https://steamcharts.com/

and loads more "interacting with their servers" data in real time.

https://steamstat.us/

Whats presented in the monthly hardware survey:

https://store.steampowered.com/hwsurvey/Steam-Hardware-Software-Survey-Welcome-to-Steam

Really is 1.2million+ Linux users

"LA LA LA LA put your fingers in your ears it can't be right" isn't really productive or honest.

Only the last one comes from Valve and is from the survey, the other two are external sources trying various tricks to do stats since Valve does not release any official stats more than the survey numbers. Again, Valve might have exact numbers internally from telemetry but they are not sharing them publicly.

[mSparks]

Only the last one comes from Valve and is from the survey, the other two are external sources trying various tricks to do stats since Valve does not release any official stats more than the survey numbers. Again, Valve might have exact numbers internally from telemetry but they are not sharing them publicly.

Which, if I suspend belief for just a second, changes what and is relevant how exactly?
Even if it was just a sample, 1% of 120million would still be 1.2million and a 0.3% increase MoM is still 30,000 more monthly active linux users in one month.

[F.Ultra]

Which, if I suspend belief for just a second, changes what and is relevant how exactly?
Even if it was just a sample, 1% of 120million would still be 1.2million and a 0.3% increase MoM is still 30,000 more monthly active linux users in one month.

Because Valve is not showing that 1.13% of 120M users are Linux users, they show that 1.13% of the people that replied to the survey are Linux users. Which means that the number if within the error bars of the reported data. Actual Linux numbers could be stable at say 3%-1% but shown as fluctuating between 0.9% and 1.13% from time to time depending wholesale on who responds to the survey and when due to the inherent errors in the reporting.

[mSparks]

Because Valve is not showing that 1.13% of 120M users are Linux users, they show that 1.13% of the people that replied to the survey are Linux users. Which means that the number if within the error bars of the reported data. Actual Linux numbers could be stable at say 3%-1% but shown as fluctuating between 0.9% and 1.13% from time to time depending wholesale on who responds to the survey and when due to the inherent errors in the reporting.

So, firstly, you need to brush up on the law of large numbers.
and secondly if you think linux is loosing and gaining 2 million gamers on a month by month basis I really call into question

• your definition of stable
• where you think those 2 million gamers went and come from

[F.Ultra]

So, firstly, you need to brush up on the law of large numbers.
and secondly if you think linux is loosing and gaining 2 million gamers on a month by month basis I really call into question

• your definition of stable
• where you think those 2 million gamers went and come from

I'm not sure why you are so confused by why i write and reverse the meaning of it? I'm not saying that Linux is loosing and gaining 2M users from time to time, I'm saying that the numbers of Linux users are fluctuating so much between each survey due to our install base being inside the error-bars for what the survey is collecting. LLN does not make a play here since the survey is not a rerun on the 120M users each time, it's not even rerun on the same user twice. Myself I was counted once long ago but have since then never gotten the survey, nor have a lot of other Linux users, so quite likely is that Linux users are undercounted in the survey.

[mSparks]

I'm not sure why you are so confused by why i write and reverse the meaning of it? I'm not saying that Linux is loosing and gaining 2M users from time to time, I'm saying that the numbers of Linux users are fluctuating so much between each survey due to our install base being inside the error-bars for what the survey is collecting. LLN does not make a play here since the survey is not a rerun on the 120M users each time, it's not even rerun on the same user twice.

1% of steam users would be 1.2 million linux users
3% of steam users would be 3.6 million linux users

So when you said
->Linux numbers could be stable at say 3%-1%

You actually said
->Linux numbers could be stable at say 3.6 million to 1.2 million users

How are you not sure why I find that a hard sell?

Myself I was counted once long ago but have since then never gotten the survey, nor have a lot of other Linux users, so quite likely is that Linux users are undercounted in the survey.

I think any difference is more to do with PCs that aren't steam users.
60-80% of desktop PCs have intel GPUs, but only something like 5% of Steam users, Linux is going to have a much greater share of those desktops, but also, you aren't going to be gaming on an intel GPU, and users are much more likely to run windows for gaming because up until very recently gaming on Linux was just plain rubbish.

They also definitely wont send you a survey if your hardware hasn't changed, because they already know what it is.

[F.Ultra]

1% of steam users would be 1.2 million linux users
3% of steam users would be 3.6 million linux users

So when you said
->Linux numbers could be stable at say 3%-1%

You actually said
->Linux numbers could be stable at say 3.6 million to 1.2 million users

How are you not sure why I find that a hard sell?



I think any difference is more to do with PCs that aren't steam users.
60-80% of desktop PCs have intel GPUs, but only something like 5% of Steam users, Linux is going to have a much greater share of those desktops, but also, you aren't going to be gaming on an intel GPU, and users are much more likely to run windows for gaming because up until very recently gaming on Linux was just plain rubbish.

They also definitely wont send you a survey if your hardware hasn't changed, because they already know what it is.

No, with "Linux numbers could be stable at say 3%-1%" I meant that: take a random number from that range, say 2.4M. So lets say that the number of Linux users in Steam have been exactly 2.4M for the last two years (and that the number of Steam users was also a constant for this period), that would still due to how the survey are done, be presented as fluctuations between 0.9% and 1.13%.

There is a range of uncertainty due to the inherent errors in the numbers, hence my take that we are somewhere in the range of 3% to 1% of all the Steam users, of course if we where fluctuating between 3% and 1% the number would not be stable so that alone should tell you that you interpreted that completely wrong.

All I'm saying is that the way the survey is done in combination with our numbers being inside the error bar for a statistic counted this way, means that any fluctuation that we see could very well just be random noise and not an indication that this month we got +200K Linux users and the next month we lost 150K Linux users.

AFAIK you are only included in the survey for the month that you are participating, and it's done this way to show stats for active users so that they don't get bogged down by all the bot or stale accounts. I have used Steam for aprox 108 surveys and have had been selected to be included just once. Had the trigger been new hardware I would have been included the first month I installed Steam but no, I have also changed CPU and GPU more than once and again no survey.

[mSparks]

All I'm saying is that the way the survey is done in combination with our numbers being inside the error bar for a statistic counted this way, means that any fluctuation that we see could very well just be random noise and not an indication that this month we got +200K Linux users and the next month we lost 150K Linux users.

So my point is "thats not how steam hardware survey works".

Steam is in a fairly unique position in that they know when you change your hardware (your account is bound to the hardware you log in on), if you change your hardware you have to validate it and they send pretty much most of them an automated hardware survey (I've done two in the last 12 months due to buying lots of new hardware, main machine which is FC35 on an RTX3070 and AMD5900, a mac mini, a macbook air and a ROG windows laptop with RTX3050 - that is currently back in the shop DOA because the keys started falling off the keyboard). Not sending them hardware data is also "opt out" iirc.

What is then presented in the hardware survey is "what hardware did people log in to steam on in the last month" - not your typical survey of "what were the answers to the hardware survey this month".

You can see they have all that nailed down if you try and log in to steam on a device you haven't before, you'll get a message that this is a new device and they'll send you an email to let it log into your account.

[F.Ultra]

So my point is "thats not how steam hardware survey works".

Steam is in a fairly unique position in that they know when you change your hardware (your account is bound to the hardware you log in on), if you change your hardware you have to validate it and they send pretty much most of them an automated hardware survey (I've done two in the last 12 months due to buying lots of new hardware, main machine which is FC35 on an RTX3070 and AMD5900, a mac mini, a macbook air and a ROG windows laptop with RTX3050 - that is currently back in the shop DOA because the keys started falling off the keyboard). Not sending them hardware data is also "opt out" iirc.

What is then presented in the hardware survey is "what hardware did people log in to steam on in the last month" - not your typical survey of "what were the answers to the hardware survey this month".

You can see they have all that nailed down if you try and log in to steam on a device you haven't before, you'll get a message that this is a new device and they'll send you an email to let it log into your account.

That is not my experience which match that of many other Linux users, in fact a common complaint among Linux users is that they have never gotten the survey to begin with. I have since I started to use Steam changed both GPU, CPU, Motheboard, amounts of RAM, drives etc and not a single one of those changes have lead to the survey popping up.

It did pop up once several months ago out of the blue though, not triggered by any changed from my side so by all accounts it does look like I was just randomly selected to be part of that months survey. But I digress and in the end only Valve truly knows how all this works.

[mSparks]

But I digress and in the end only Valve truly knows how all this works.

Which is why you should always draw on multiple sources for numbers like that. For example:

http://dashboard.x-plane.com/

BUT also, dont get lost in percentages, 0.13% in a month "sounds" tiny and insignificant, but when its out of a population of 120million people - every desktop gaming PC - 156,000 new Linux gaming machines in 1 month, its a lot more significant than 1.27% of the ~2,000 people who fly XP11 on a daily basis.