amok: It depends on what you mean by ‘reliable.’ Any survey method will have some degree of sampling bias. However, it’s still a reliable snapshot of the Steam users who agreed to take part in the survey at that time, since the sample is randomly selected and the sample size is statistically acceptable for the total population.
Syphon72: I know how surveys work. It's more of a jab at people saying they're really not reliable when Linux drops a little, but they are when showing Linux growing. It's statistically acceptable, but is it truly accurate? Technically, Linux could have a bigger market share because the survey isn't based solely on people who actually take it. I never once took the survey while using Steam, and I use multiple OSs.
Indeed. The Steam Hardware Survey only captures Steam users - and that’s exactly what it’s designed to do. So for what it is supposed to do, it is reliable.
There are two large populations missing if you try to generalize the data to a wider audience: people who do not use Steam, and people who do use Steam but have opted out of the survey. So it only reflects that specific population, which people often forget.
Outside the survey, though, everything becomes speculation, there simply are not any other reliable numbers, there has never been a rlaible census of types of OS people are using. The main problem is that there is wast difference between the OS people use for work (and those operating on for example servers and ruters but are still counted), and then the OS people use at home for playing games. So, flawed as it may be, it’s still the best data we have.