My most recent bad take on distinguishing fantasy from science fiction used a simple case of judging a book by its cover. If the title used type with serifs then it was (probably) fantasy and if it used sans-serif then it was probably science fiction (or at least marketed as such).
The Goodreads semifinalists have been announced (http://file770.com/goodreads-choice-awards-2020-semifinalists/ ) and that present a perfect opportunity to test the serif-fantasy-spotting algorithm.
I count 14 out of 20 with serifs, 6 without and one (Seanan McGuire’s) which is hard to classify but which I put in serifed.
The genre defying works are:
- Magic for Liars by Sarah Gailey
- Storm Cursed by Patricia Briggs
- The Dragon Republic by R.F.Kuang
- A Little Hatred by Joe Abercrombie
- Gods of Jade and Shadow by Silvia Moreno-Garcia
- Black Leopard, Red Wolf by Marlon James
Meanwhile what does science fiction offer?
I count 15 sans-serif titles, 5 serif titles and one (Charlie Jane Anders’s) that is ambiguous but I’m lumping in with the sans-serif (the T’s have serifs and the other letters not so much).
The genre busting titles are
- Gideon the Ninth by Tamsyn Muir
- This is How You Lose a Time War by Amal El-Mohtar and Max Gladstone
- Dark Age by Pierce Brown
- Thrawn: Treason by Timothy Zahn
- The Deep by Rivers Solomon
So, if you tried to use the rule to classify books by genre while ignoring the content you’d get this table:
||Actually Science Fiction||Actually Fantasy|
That’s effectively terrible. It’s enough to show there’s method in the madness but practically it would be wrong a quarter of the time. Mind you, we should judge a method not just on basic results but how it compares with other methods and I think it holds up quite well compared to other schemes I’ve seen for genre classification. 🙂