Our Linux & SteamOS gamer survey results for January 2016

The graphs are blank for me, except the first one.

Hmm seems Firefox rendering is a bit pants for an svg included as a normal image. Works fine in Chrome.

The graph lines don't even show up in Firefox when set as an html object, wtf is Firefox doing hmmm.


Last edited by Liam Dawe on 23 January 2016 at 11:59 pm UTC

Yup, the graphs are broken in Firefox. Very interesting results, though (the graphs are working fine with chromium). One thing that is noteworthy is the "How many games did you buy last month" result: Looks like the winter sale did indeed boost game sales for us. Or was it the new ports?

I missed this survey; I am probably not alone in that situation.

EDIT : graphs are blank in inkscape too, but not in kolourpaint, FYI. (the fact that it could actually handle svgs actually surprised me).


Last edited by MayeulC on 24 January 2016 at 12:12 am UTC

The graphs are blank for me, except the first one.

My bad, I wrote a comma where I shouldn't have so the lines in the svg file are not valid, Chromium seems to be less strict so I didn't notice the mistake. I've just fixed it and I'm working on another couple of minor improvements, I'll send The Boss the updated files some time tonight. Sorry again.

Thanks dude :D

Please don't post the numbers especially with 1/10 000th accuracy. With a sample size of ~1000 this makes absolutely no sense.
One survey per month is OK for me.

I tried so hard to make the move from Gnome 3.18 to Plasma 5 yesterday, so far no go.

The SVGs are also broken in palemoon. :(

While cool and all, I think these new graphs really show the utter stupidity of some people. I mean, some people have seriously submitted "Windows 7" as their current Linux distro? And despite having a dedicated Arch-based option, some idiots have still chosen "other" and entered variants of "Arch" and "ArchLinux"?

There is definitely some merit in removing the safety labels from products like bleach. Let the problem solve itself. :P

It works right now with a survey every month, the results are quite boring sadly but I'm afraid that if it's only every two months people would forget about it and the number of respondent would drop.

Please don't post the numbers especially with 1/10 000th accuracy. With a sample size of ~1000 this makes absolutely no sense.

It depends on how you read the data, under the assumption the respondent are the target population they do make sense. Please ignore them if meaningless to you.

While cool and all, I think these new graphs really show the utter stupidity of some people. I mean, some people have seriously submitted "Windows 7" as their current Linux distro? And despite having a dedicated Arch-based option, some idiots have still chosen "other" and entered variants of "Arch" and "ArchLinux"?

There is definitely some merit in removing the safety labels from products like bleach. Let the problem solve itself. :P

Some questions were fixed after a few surveys to be more clear, for example the variants of "Arch" have almost completely disappeared (only 2) in the last year.

Here some more curious answers:

OS question: "sadly windows", "Windows. Linux sucks", "uninstalled Linux", "Too frequently distrohopping to say", "n/a", "Nothing", "Windows, when there are more games i will switch to Linux", "Reverted back to Windows", "OSX", "Linux From Scratch" (all my respect!)

For the DE question: "what", "Command line", "Several", "No DE, I control my machine directly via opcodes", "explorer.exe", "Aero (Win7)", "on laptop i use unity in Ubuntu--because Ubuntu comes as default like this--but BEfore I very much liked Gnome 2, thats why I switched to Mint with Mate on board", "I don't own a primary Linux gaming PC.", "OSX"

For the retailer question: variants of "didn't buy any game" (I guess many don't know that an answer can be left empty), "Trading steam keys", "gift", "Linux gameing sucks!", "Support to open source projects", "Local electronics store", "I was broke last month...", "pregnant wife. The time wasn't there for new games." (this was on Oct/2014, late congratulations!), "Didn't buy any game last month duo to school" (good choice!)

all blank on firefox

In Safari on iOS the graph's titles are very tiny. So tiny I didn't see them in the first place and thought there weren't any.
Imho a monthly survey is better than a survey every two months, as it will give more fine grained data to look at over time and in the future. I don't think it will affect the number of people participating.

Nothing new about firefox, no graphs.

Regarding monthly or every two months survey, I vote for the monthly option. It's a five minuts or less survey, that could be filled using the mobile.

The new graphs are up, thanks to fedso for fixing them all up. I will respond to suggestions and stuff when I have more time later tonight.

Some feedback:
The previous survey results were much better presented than this. Anyone who works with statistics or data display can tell this presentation could be easily improved. Two suggestions I can offer you;

i) When working with normalized data (as in this case) the axis which refer to the distribution, the %, should have a fixed scale along all graphs for easier understanding. The vertical axis in this case was automatically generated for every graph I suppose, but fixing the top value to 100% would be much better;

ii) Makes absolutely no sense to work with more than two or three significant figures. "0.56%", "1.4%", "25%" or even "5.78%" are fine. Now, "78.94%" makes no sense.

Cheers


Last edited by khalismur on 24 January 2016 at 1:23 pm UTC

Maybe there could be two kinds of surveys. One full version every second month (like we have now) and a smaller and quicker version for the rest. The small version could remove all the boring questions like desktop, gpu, drivers and then spice things up by adding a few more of the interesting questions (new ones each time).

i) When working with normalized data (as in this case) the axis which refer to the distribution, the %, should have a fixed scale along all graphs for easier understanding. The vertical axis in this case was automatically generated for every graph I suppose, but fixing the top value to 100% would be much better;

I'd rather stick to the rule that graph space must be used efficiently. There is no reason to compare graphs side by side anyway so different scale shouldn't affect readability. Of course if the general consensus is to to go up to 100% regardless of the maximum value I'll be happy to comply.

ii) Makes absolutely no sense to work with more than two or three significant figures. "0.56%", "1.4%", "25%" or even "5.78%" are fine. Now, "78.94%" makes no sense.

Any one else would like to shoot at the 4th significative digit? ;)

i) When working with normalized data (as in this case) the axis which refer to the distribution, the %, should have a fixed scale along all graphs for easier understanding. The vertical axis in this case was automatically generated for every graph I suppose, but fixing the top value to 100% would be much better;

I'd rather stick to the rule that graph space must be used efficiently. There is no reason to compare graphs side by side anyway so different scale shouldn't affect readability. Of course if the general consensus is to to go up to 100% regardless of the maximum value I'll be happy to comply.

No need to go up to 100% as long as the bar numbers are rounded nice and even. Instead of 24% use 25%. 69% could go to 80%. 82% could go all the way up to 100%. It makes things a lot more consistent and more easily readable.

ii) Makes absolutely no sense to work with more than two or three significant figures. "0.56%", "1.4%", "25%" or even "5.78%" are fine. Now, "78.94%" makes no sense.

Any one else would like to shoot at the 4th significative digit? ;)

Having the numbers alongside the graphs makes no sense and makes the presentation look to cluttered. Hide them or remove them.

This new graphics are more difficult to read than the old way.

Hmmmmm, Arch is 4x present in this survey?

Why?
And why only Windows 7 and not 8 and 10?

On the Linuxside, why not only .deb and .rpm based Destributions?

Conclusion - the old way was much better.