Quoting: Liam DaweI don’t get the fuss against it.

You can literally always find a polynomial that goes exactly through every point. You can literally always find a sine function that goes through every point.

Should you be plotting your points against some function your model needs explanatory power for why "percentage of Steam users running Linux" should be some factor of e raised to the power of "number of days since 2017," say, or "the tangent of days since 2017," or "the inverse of 1 plus e to the power of the negative of the number of days since 2017." Or whatever bullshit function looks a bit like it.

"Tee-hee I like de dot-to-dots" is meaningless gibberish.

That's why.

View this comment - View article - View full comments

Quoting: PlayingOnLinuxphoneThere is no real exponential trend. I did the math some hours ago and there was a linear trend before Steam Deck release, than an accelerated trend between Steam Deck Release and Win10 EOL and a 5.5 times increase after Win10 EOL. Steam Deck OLED did not change the linear trend. November and December 2025 had exactly the same 5.5 times increase as the whole 2026 until April. January and February were Chinese New Year, March somehow bad data, so these should be ignored.

If it would be a real exponential curve, March data would be real and April even above that. So it would be better to use 3 linear lines, clamped between those relevant dates.

PS: If it is linear we will see on Windows 10 extended support end of life. It should be around 6%. It should be far above 6% if it is exponential.

Here's a visual aid.

As you say, it's all pretty flat before the Deck, then we get added a pretty linear Deck signal, and then we get a pretty linear desktop signal as the Deck flattens out.

I've found the 5-month moving average to be quite good at separating out the signal from the noise.


There's no real justification for splatting the data onto an exponential curve, although we get someone demanding that it's done pretty much every month so I'm not surprised that Liam's given it to them to make them stop. One could splat it into a logarithmic, sinusoidal or tangent curve just as easily.

View this comment - View article - View full comments

Over to you in the comments - what have you been playing the most recently?

I've mostly been playing Bloons TD6 on the Deck recently. Being able to do co-op on that with the little one when we had car trouble on a long journey was an absolute sanity saver.

View this comment - View article - View full comments

There are a few times were "Verified" titles just don't perform well enough, and that does need to be made clear to developers. And, ideally, Valve need to be just a little more strict on that 30FPS performance floor.

Absolutely this.

View this comment - View article - View full comments

Ars has an interview [External Link] with the Framework CEO for those that are interested.

View this comment - View article - View full comments

While India has a large population, it isn't a large Steam market. India's statistic (for comparison with the others I was adding as you were writing your post) is just 15.6 PB. That's the same sort of size as Indonesia or Argentina. Not nothing, but not a market mover.

View this comment - View article - View full comments

Quoting: LoudTechieThnx for the information.
In which case I'll ask how come only the recorded number of Chinese users visibly fluctuates?
With access one can point to some kind of government interference, with measurement issues that doesn't work.

In countries where household PC ownership is standard (such as Europe and North America - the other significant regions on Steam) the measurement issue isn't a factor.

In other countries where PC bang gaming is standard (such as Japan and South Korea, as I mentioned) the Steam market just isn't as big.

China is a huge market, and the historical console ban among other factors has made playing online games in Internet cafes the norm, where measurement is hard (potentially impossible).

Edit: here are some stats from Steam for comparison.
Download traffic over the last 7 days.
UK: 53.9 PB
Japan: 35.9 PB
South Korea: 39.7 PB
China: 380.8 PB

View this comment - View article - View full comments

Quoting: LoudTechieIt could be, but I'm not seeing significant fluctuations, which is what surprises me.
I would expect Indian Steam access to fluctuate comparable, but with a different rhythm, to Chinese Steam access.
Simplified chinese spikes are easy too see, yet there're no visible english and/or Hindi spikes.

The number of Chinese users on Steam doesn't fluctuate.

The number of samples recorded varies, on account of sometimes being wildly wrong.

The issue is that it's difficult to count each machine in a PC bang once and only once per year. You can't count them server-side because they're all behind one IP address. You can't count them client-side because the clients get periodically wiped, which erases your means of seeing that you already counted it. Valve put their hands up on the issue - for a while after that they corrected the issues in the data when they came up, but they stopped doing that a couple of years ago so we get the spikes again.

Japan and South Korea could well have the same issues with data collection, but they aren't ~ a third of Steam, so no one notices.

View this comment - View article - View full comments

Quoting: LoudTechieThe improvements each of them adds make it better for everybody.

“The more you share, the more your bowl will be plentiful.” -- Anderson Dawes

View this comment - View article - View full comments

Quoting: PlayingOnLinuxphone

Quoting: CatKillerNothing has fundamentally changed about Linux gaming since then - there have been incremental improvements, sure.

That is half right and half wrong. Linux itself just improved, nothing fundamental new. That is correct. But the society started to fundamentally change and that is the core point. That many people switched to Linux shortly before W10 EOL and all the months after is a direct result of the society change in 2025 (which is the result of all what you wrote).

Let's explain it in another way: what would have been if 2025 would just be a year as every other one? Nobody outside the Linux bubble is speaking about Linux beside Steam Deck. Linux would just die again over time, because Steam Deck is no topic any longer at some point. Some people may switched and that's it. The few people still switching would not make a huge difference.
But all the things happened in 2025 (nothing fundamental new, but all together seen) made Linux a topic to speak about. Even my dad who was never interested in Linux told me he wants to try it (I did not even ask). That is the core of 2025. It is not just the speech itself, what I value here. It broke through a wall that cannot be closed again. Even if Windows 11 becomes "fixed", they cannot return the time and people know about Linux now and a lot of them are still interested to at least try it out.

That is also my point. To quote myself from earlier in the thread:

Quoting: CatKillerThere's been such a difference in the conversation about Linux gaming from the gaming press, YouTubers, and general tech people since Valve released their high-profile Linux gaming demonstration device. Even if you've never held a Deck in your hands you know that Linux gaming is awesome - something that we've known for ages, but the wider population needed to be shown.

I agree with you that things have changed - to borrow gradyvuckovic's metaphor, the snowball is gaining momentum now. If Windows were marginally less terrible, that change would still be happening. You can see it in the data. There are certainly some people that are contemplating switching now because of the change in mood about Linux gaming that potentially wouldn't be if Windows were more comfortable for them than it is - I don't doubt that Microsoft's choices are an amplifier for Linux gaming growth. But Windows has always been pretty bad, and we didn't have that growth before the mainstream conversation changed.

I'm glad you decided to join us in 2025, though.

View this comment - View article - View full comments