Well, one of the big reasons for non-sincere voting (practically the only one) is that there are consequences; Vote for the guy you like, and the guy you hate will thrash the guy you mildly dislike.

In a poll, it doesn't work that way; there are secondary ways to game with polls, I know, like trying to persuade your opponent that to sit on a "lead", but they're much less prevalent.

The primary outcome is seeing your candidate do better in the polls, so you're most likely going to vote for your candidate. Especially on the ones that ask your vote for each of a series of possible pairings, so in each case it's a pure two-candidate race, and there's no issue with "do I help my real candidate or my practical candidate?".

In fairness, C doesn't apply here because it was not meant as a prediction of election results, but as a measure of the itT crowd, including those who were ineligible to vote. Would it be wrong to try to predict the election outcome with this? Well, only about a half-dozen different ways! That doesn't invalidate it for it's purpose though...

Maybe I'm a one-trick pony here with my election methods, but I don't see polls as being that big a problem. Are they unreliable? Yes. (Some are decent, but assume they're unreliable unless you know the methodology, the whole questions used, etc. Then, if they're clean, assume they're semi-reliable.) But they're nothing locking us into a 2-party system. It's the voting system, the whole voting system, and nothing but the voting system.