July Steam Survey: RTX 5000 surge, new top GPU, 4 in 10 participants using AMD CPUs

[HardReset]

What numbers do you have to be able to state either way?

Mindfactory sales numbers tell 9070 sell much better than 5090. While that is just one retailer, Steam is just one game launcher.

 

[Yppah]

if Valve included the steam deck then the OS's would read closer to 5%. 6m steam decks divided by 136m users = 0.044% linux users from steam deck alone. No, we don't KNOW, but the math allows us to extrapolate from the data that it like is not from the steam deck. If that angers you then idk what to tell you

EDIT:

Steam Survey For July Shows Linux Use Approaching 3% - Phoronix

www.phoronix.com

The share of SteamOS making up that of Linux users on the steam survey actually dropped with Ubuntu, Mint and Fedora making up the biggest increases in shares. That Fedora thing really threw me off when I saw it. But with SteamOS's share of the steam linux segment dropping I feel it is safe to say that the increase in linux users is not made up by the SteamDeck. Also, the Steam Deck been out for 3 years now, those who want one have mostly already bought one.

LOL you and your extrapolation. Excellent back track. If that angers you then idk what to tell you

 

[nnguy2]

LOL you and your extrapolation. Excellent back track. If that angers you then idk what to tell you

Right?!?

I have a graduate degree in statistics and my day job is manager of statistics for a large medical device company.

I love it when people argue statistics, surveys, and inference with me.

But keyboard warrior you keep on keeping on.

 

[yRaz]

LOL you and your extrapolation. Excellent back track. If that angers you then idk what to tell you

Right?!?

I have a graduate degree in statistics and my day job is manager of statistics for a large medical device company.

I love it when people argue statistics, surveys, and inference with me.

But keyboard warrior you keep on keeping on.

I extrapolated first and then I found the actual data to back it up which confirmed my hypothesis and the source was steams own survey statistics

 

[nnguy2]

I extrapolated first and then I found the actual data to back it up which confirmed my hypothesis and the source was steams own survey statistics

Nonprobability sampling which steam survey (online opt-in) is the most unreliable form of survey sampling there is.

But you probably won't take my professional advice. So there's that.

 

[yRaz]

Nonprobability sampling which steam survey (online opt-in) is the most unreliable form of survey sampling there is.

But you probably won't take my professional advice. So there's that.

I linked to a statistic that they gave where the Steam Survey said the percentage of SteamOS users dropped among Linux users. I'm actually interested in what your professional advice says about that.

Further, the math shows that if they included the steam deck in their hardware survey then the percentage of Linux users would be almost double what they're reporting

 

[nnguy2]

I linked to a statistic that they gave where the Steam Survey said the percentage of SteamOS users dropped among Linux users. I'm actually interested in what your professional advice says about that.

Further, the math shows that if they included the steam deck in their hardware survey then the percentage of Linux users would be almost double what they're reporting

Not much. We don't know if the responders are representative of the steam user base. How representative are the people who received the survey? What about the people who declined? For example if the 30% of people who declined, 70% have AMD gpu (extreme example) then the Nvidia card gpus are overrepresented and needed to be adjusted down (using weights), this is known as bias.

Or for example, made up extreme scenario: if the true population of users on Steam, windows are 90%, and Linux 10%. 100 people were invited, 90 windows and 10 Linux users, but 8 Linux users declined and all 90 windows users accepted, now you have 92 people total. The survey will show 2.1% (2/92) surveyed were Linux users. What adjustments were made to account for non-response of Linux users? What if Linux users tend to decline more often, did the survey adjust for that?

Without knowing the methodology: how the survey is sent and what weighting methods are used to reduce bias, caution should be used to make inference from the survey.

 

[yRaz]

Not much. We don't know if the responders are representative of the steam user base. How representative are the people who received the survey? What about the people who declined? For example if the 30% of people who declined, 70% have AMD gpu (extreme example) then the Nvidia card gpus are overrepresented and needed to be adjusted down (using weights), this is known as bias.

Or for example, made up extreme scenario: if the true population of users on Steam, windows are 90%, and Linux 10%. 100 people were invited, 90 windows and 10 Linux users, but 8 Linux users declined and all 90 windows users accepted, now you have 92 people total. The survey will show 2.1% (2/92) surveyed were Linux users. What adjustments were made to account for non-response of Linux users? What if Linux users tend to decline more often, did the survey adjust for that?

Without knowing the methodology: how the survey is sent and what weighting methods are used to reduce bias, caution should be used to make inference from the survey.

Sample size tends to account for those variances and the steam survey is massive.

 

[nnguy2]

Sample size tends to account for those variances and the steam survey is massive.

That's not how sample surveys work.

A large sample of bias data gives bias responses.

 

[yRaz]

That's not how sample surveys work.

A large sample of bias data gives bias responses.

But now you are infering that the data is biased to justify your argument without and proof to the contrary. The difference between my argument and yours is that openly stated the weakness in my hypothesis and why I think that weakness isn't relevant. You're now making conjectures to inbvalidate it.

 

[nnguy2]

But now you are infering that the data is biased to justify your argument without and proof to the contrary. The difference between my argument and yours is that openly stated the weakness in my hypothesis and why I think that weakness isn't relevant. You're now making conjectures to inbvalidate it.

I'm saying without steam releasing their methodology no one knows how good their data is.

Stop deflecting.

 

[yRaz]

I'm saying without steam releasing their methodology no one knows how good their data is.

Stop deflecting.

I'm not deflecting, I cited a source that showed steamOS market share dropping among Linux users and an increase in Ubuntu, Debian and Fedora users. The question was "are steam deck sales artificially inflating the amount of Linux users on the steam survey. With the share of SteamOS dropping among Linux users on the steam survey, I think it's safe to say that it's highly unlikely that the steam deck is artificially inflating Linux market share.

 

[nnguy2]

I'm not deflecting, I cited a source that showed steamOS market share dropping among Linux users and an increase in Ubuntu, Debian and Fedora users. The question was "are steam deck sales artificially inflating the amount of Linux users on the steam survey. With the share of SteamOS dropping among Linux users on the steam survey, I think it's safe to say that it's highly unlikely that the steam deck is artificially inflating Linux market share.

I would caution extrapolating what you see from steam survey to make general statement about the distribution of Linux market share of all pc users.

We need to know how many people offer survey, how many declined/accepted, were everyone offered the survey representative of Steam population, etc. Lots of things for statisticians to consider.

I doubt Steam has a team of survey statisticians creating and adjusting weights for each surveys.

Agian my main point is use steam survey with caution.