Scoping Out the Hugo Nomination Stats

By Kyra: [Reprinted from a comment by permission.] Some random observations about the voting statistics released with the 2017 Hugo finalists

Clear Favorites: The following categories had clear favorites during the nominations:

  • Almost 60% of nominators voted for the top pick in Dramatic Presentation Long (DPL)
  • Almost 50% of nominators voted for the top pick in Dramatic Presentation Short (DPS)
  • Roughly 50% of nominators voted for the top pick in Semiprozine

This may in part be due to fewer options to vote on in the case of Semiprozine — the votes were distributed among 103 nominees, the least of all categories by a significant margin. However, DPL had a number of nominees comparable with many other categories (206), and DPS had one of the highest number (569), so it seems likely that there were certain nominees that were either obviously standout or widely well-known. Probably true of Semiprozine as well — 103 is still quite a few!

Mild Favorites: Several categories had mild favorites:

  • About 36% of nominators voted for the top pick in Novella
  • About 37% of nominators voted for the top pick in Related Work

Novella nominations were spread over a relatively low list of nominees (187), but the number of Related Work nominees was fairly high (344).

In most of the other categories, about ~25% of nominators, +/-3%, voted for whatever the top pick was. Once again, there was a wide range for distribution of nominees for these, ranging from a low of 152 for fanzines to a high of 652 for novels.

No Favorites: The following categories arguably had no clear favorites:

  • Only about 14% of nominators voted for the top pick in Short Story.
  • Only about 17% of nominators voted for the top pick in Professional Artist.
  • Only about 15% of nominators voted for the top pick in Fancast.
  • Only about 18% of nominators voted for the top pick in Fan Writer.

For Short Story, this was almost certainly in part related to the wide number of options available – votes were distributed across 830 nominees! Professional Artist was also reasonably high at 387, but Fancast and Fan Writer were pretty typical at 253 and 275, respectively.

Moral: To have the lowest chance of winning a Hugo, write a short story. To have the highest chance, edit a semiprozine. If that sounds difficult, then simply writing or directing a widely-distributed SFF Hollywood blockbuster will also give you a decent shot.

7 thoughts on “Scoping Out the Hugo Nomination Stats

  1. This makes me even more interested in nomination stats than usual! Thanks, @Kyra. 😉

  2. I took a look at the low limit too. I.e. for novel, there’s 2078 ballots and a finalist range of 156 to 480. Which says that 23% voted for the top novel, while 7,5% voted for the least popular finalist.

    I spot-checked a few, most categories seems to be around 7%. There are a few over 10%, and the highest one I could spot was Editor short form, where the least popular finalist was on 15,6% of the ballots.

    There seems to be some correlation between popularity of the most popular and the popularity of the least popular, but it’s not absolute:
    – DPS-L have the most popular favourite, and also a relatively popular #6 at 13,8%.
    – DPS-S have the second most popular favourite, but #6 is “only” at a 7,8%,
    – Fan Writer have a low number for the favourite at 18,9% – but a relatively high number for #6 at 9,9%.

    Someone with more time and statistics skill kan certainly make more of this…

    (I’ll also note that any talk about “clear favourite” is questionable when we don’t know the popularity of #2-#5. For example, it is plausible that there’s two or more movies with nominations from “almost 60%” of nominators.)

  3. In no category would the 5% rule have been invoked, if it were still in force. This, however, may be to some extent the result of the slate. If we suppose that in a number of categories the lowest vote was the slate vote, which seems to hover around 80, this exceeds 5% in the smaller categories, but it’s possible that the next highest would not have.

  4. Power-law distributions are fun!

    The dominant effect isn’t how many total candidates there are but how many “popular” ones there are. So although there are a whole lot of TV episodes, there are just a handful that were really popular, hence the steep curve. Contrast short stories, where there are a great number of them and a broad range of opinions.

    If anyone wants to play with it in Excel, power laws give you a way to estimate how many votes there were in position #2, #3, etc. It’s a rough estimate at best, and the puppies mess it up a bit, but it’s still useful for some purposes.

    Let’s call the two numbers we’ve got “max” and “min.” We want to use them to compute an exponent, s. Then the number of votes at “rank” will be max*rank^s, which you can compute easily in Excel. The formula for s is

    s = (ln(min) – ln(max))/ln(6)

    So for Best Novel, min is 156, and max is 480, so s = -0.62728
    That gives us the following estimates for the ranks:

    1: 480
    2: 311
    3: 241
    4: 201
    5: 175
    6: 156

    Again, this should be roughly correct. You can do this same experiment with the data from previous WorldCons, although the heavy slating presence makes it harder to use for 2016 and 2015.

  5. @Greg: OK, you can come back on 11 Aug (US time) and quantify how far the actual nominations deviated from a power curve…. In the meantime: how does data from 2014 &prev fit such curves?

Comments are closed.