That's not how I interpret it works. I see it as one full poll, reduced down to a second poll, reduced down to a final.
Plus, I really don't want some fancy-pants statistimagician to prattle on if there's a difference in number of voters scattered among three polls, and hear about unequal distribution and needing to weight each poll accordingly.
Thanks for the discussion, all.
I'm going to construct a poll of all entries, and allow for five selections. The top half will roll to round two, and I'll allow three selections to get to the final 10. Then a third poll to select the final three.
Probably won't have the poll ready until tomorrow morning.