The trick with non-standard voting schemes is that the art of polling gets completely shaken up. When you can cast more than one vote, the polls won't add to 100% any more. Instead, assuming you had amazing polls that were able to get the full state of how voters would vote, and if you were using approval voting, then there would be some ability to vote strategically. If your top two candidates are tied, you could choose not to give any preference to your #2 pick, with all the risks that some unforseen event could render that a bad choice.

Indubitably, the presence of polls has a radical effect on the electorate. I wonder sometime whether polls "predicting" a strong showing by a candidate can create one, as people get on board with the winner or lower-valued candidates are convinced they should drop from the race.

There are a pile of papers linked from the sites I mentioned above. If I get a spare couple days, I'll try to read them all over. The mathematics of these systems have been studied for quite a long time...