10,000 in a population of close to 13 million concurrent users per day (at least Valve reports the latter number). Are you serious? That's representative of very little. Furthermore, we don't actually know how many people are being surveyed.
'murican political pollsters try to justify their polling in the same manner. Sample sizes of 10,000+ in a nation of 320,000,000, with a third of this actively voting. That's representative of absolutely nobody.
A sample size of 10,000 for a population of 13,000,000 has a confidence level of over 95% and a confidence interval of +/- 1%.
For example, If you wanted to just be within +/- 5% for the entire Steam population (125 Million), you would only need to survey 400 people.
Edited by AmericanLoco - 10/3/16 at 5:08pm