I disagree. When you are working with a closed aperture of poll respondents (such as subscribers only) and it is that specific set of respondents you want to gather data from, any bias from voluntary poll response is marginal. Why? Because there is in fact randomness in the samples in an interest group, yet no fly randomness that tends to bias.
Completely disagree, the people who are going to be attracted to his poll and to his particular sub forum in general are not a fair cross sample of the player base or even the Subscriber Player Base. Or rather I should say there is no proof that that the interests of the people who bother to come to the General Forum Thread are a fair cross sample of the Subscriber User base. The Bias is built into how and where the poll, which is why Self Selected Polls have a major source of bias that you just can't get rid of. It's the same reason why putting a poll for customers of say McDonalds in the dine in booth section of the stores would be a bad idea, you are completely ignoring Drive through orders and people who come in but don't sit down and eat, and to pretend that people who eat in the restaurant automatically have the same break down and interests as the other 2 groups is foolish.
In other words, polls work BEST when the percentages of the basic demographics of the people polled matches the general population of people involved in the activity being polled for closest. ((Likely voters in the case of elections, and Subscribers maybe + preferred users in this polls case.)) Self Selecting polls do nothing to insure that, they can't.