My survey of bloggers called two out of four fiction categories correctly, the No Awards for Best Novella aand Best Short Story, with two near misses; The Three-Body Problem was essentially level-pegging with The Goblin Emperor in my survey, as in real life, and No Award, which was the winner in my survey, actually got most first preference votes in the count though was overtaken on transfers by the winner. (Actually my first survey at the start of July had Three-Body Problem ahead, and No Award with less of a lead in the Novelette category. I have sometimes found in election camapigns that one's gut feeling in the early days is a better guide to the outcome.) Incidentally none of the bloggers in my survey got all four outcomes that they wanted, though there were several with three out of four.
My original reporting of the Goodreads/LibraryThing stats for the Best Novel shortlist omitted the eventual winner, which was added only at a later stage. I've rerun the numbers, and the order is about the same, with The Three-Body Problem in the same rank as Marko Kloos' Lines of Departure:
|Skin Game, by Jim Butcher
|Ancillary Sword, by Ann Leckie
|The Goblin Emperor, by Katherine Addison
|The Three-Body Problem, by Cixin Liu
|The Dark Between The Stars, by Kevin J. Anderson
Not very surprisingly, this reveals that The Dark Between The Stars was a very weak choice, with by a long way the lowest rankings both of ownership and of user rating on both systems. Skin Game clearly has a lot of fans who rate it highly, but I don't think it would have won even if it had reached the ballot without puppies; Hugo voters have very rarely backed a later book in a series where they have not already given the award to an earlier one.
A little more surprising is that The Three-Body Problem is second last on both metrics on both systems. On Goodreads it's within the margin of error, but on LT it is quite some way behind in ratings despite a similar level of ownership. This of course demonstrates only that Hugo voters have different opinions from users of Goodreads and LibraryThing, and that consequently statistics from the latter are of limited value when trying to predict the behaviour of the former.