As a new-ish smash player - but stats major in college, I disagree with much of what the OP suggests. I'll say this now, I'm nowhere near close enough to the metagame to be able to have an opinion on whether or not MK should be legal.
HOWEVER, I can say that the sample size of the poll referenced creates a far more accurate depiction of the smash community than the OP suggests. While it is true that both confirmation bias and skewed polls do exist, the large sample size gathered almost completely eliminates any possibility of a completely wrong poll. I would say that the margin of error is no larger than 6%, and more likely closer to 3%.
edit: In response to your literary digest example, I've taken elementary stats (obviously) just as you have, and let me tell you that you are cherry picking pretty hard there. While it is true that voluntary polls are not the optimal method of polling, the book uses that specific case in order to prove a point, not necessarily to be accurate. The poll you mentioned is mentioned in every single stats book in the country because it is one of a kind. While there is no concrete evidence as to WHY the poll failed, most statisticians conclude that the poll was given to a biased survey base - possibly in an attempt to change the votes of their readers. Literary Digest is an isolated incident that textbooks use to further their point that voluntary polls are not optimal - however most statisticians would agree that the poll's sample size should have given them the right prediction, and therefore one of two things happened. They either got corrupted results due to bad sampling (ie. asking a majorly democrat California who they would vote for) or a one in a million occurrence that will never happen again. If you're using that example to defend your point, I would hope that you're also arguing for bad sampling area at smashboards (even though we ARE the smash community) or perhaps a one in a million thing (not likely as most/all polls ever taken with a decent sample size suggest a desire to ban MK).
The OP also suggests that players may have voted for a ban "for the lulz" - but makes no mention of the possibility going the other way. There are many people who may have voted against the ban just to piss people off, just as there may have been people who did the same for the other side. To assume on is to assume the other - but regardless they are both speculation and most likely inconsequential. Furthermore, if the results were ever mentioned before a person cast their vote, a human trait is to try to balance things they see or do (often what people call OCD). When a player who otherwise does not have an opinion on a poll sees that it is skewed in one sides' favor, it is in their human nature to try to balance the poll. Combined with the people who may have voted to keep MK legal in order to piss people off, I don't see an argument on trolls legitimately effecting the vote.
I also saw mentioned several times in this thread the idea that "well you don't know how it would have turned out if EVERYONE voted - who knows they could mostly be against the MK ban". That is logically AND statistically wrong. Once a large enough sample size is collected, and say... 60% of people vote for the ban in the sample. It can be assumed (and in all likelihood correctly assumed) that roughly 60% of the people who did not vote would ALSO be for the ban. The only time a situation like this would occur would be in a situation where those who are AGAINST the ban are placed in a situation where they are less likely to vote. Considering that there are no outside factors acting on the people AGAINST the ban, it is logical to assume that the sampled people are valid.
Lastly, as a stats major it is frustrating to see something like this (posted by Tuen) -
"I'll revise my statement from above: It is very unlikely that even 10% of our total community responded to the poll.
If we had a larger percent response, we could get closer to solving some of the problems that come with using a voluntary response poll. However, with the response cross section so small, there is nothing we can really say about the community at large... no generalizations can be made."
This only shows that you are like a normal person who does not quite understand stats. 10%, sheesh only 5% of a population is most definitely a big enough sample size! Have you ever heard of the gallup poll? These people poll less than .01% of the US population (usually around 1500 people), and ONLY take answers from people who choose to respond (for those of you voluntary poll people), and they have a margin of error less than 3%! Realistically even about 150 people would make an accurate representation of the smash communities' opinion. The fact is, whenever we see these polls come up, they always tend come out in favor of the MK ban - by a fairly significant majority.
As I said, I'm not educated enough myself to know whether MK is actually banworthy. However it is obvious to see that the solid majority of smash players DO believe that he should be banned, and that much you really cannot argue.