You misunderstand inferential statistics. This is common. (I'm a part-time statistics professor).
The incredible beauty of the math behind the field of statistics is that it -- amazingly -- doesn't matter how big your population is. The percentage in the sample generally follows the percentage in the population, regardless of the relative sizes between sample & population. That can be a mind-bender for a lot of folks.
For example, our poll is currently over 300 votes. Standard polling "margin of error", at the 95% confidence level, is computed by E = 1/sqrt

, where n is the sample size (notice that population size has no effect on this formula). So for our poll you'd calculate E = 1/sqrt(300) ~ 0.06 = 6%.
For example, the poll now says that 33% of respondents play all 4E now. Our analysis would say that in the larger population (regardless of how large it is) there is a 95% chance that the percentage of people playing all 4E is within 33+/-6% = between 27% and 39%. So it's an excellent bet that the population percentage is within that fairly narrow range.
Look closely the next time you see a political tracking poll. They usually only poll around 400 people (for margin of error = 5%), even though they're making inferences for a voting population of hundreds of millions.
Now, there are other legitimate critiques that can be made about our poll. It is in fact likely to be biased and reflect the opinion of people who attend ENWorld and like to vote in polls on the issue of 4E, for example.
But that's a separate, distinct criticism. The issue that we've got 300 votes and the larger population is in the millions has actually proven to be a non-issue by the mathematics behind inferential statistics.
http://en.wikipedia.org/wiki/Margin_of_error