{Things We Love} A Statistician Trying to Take the Guess-Work Out of “Game of Thrones” Death Predictions
0If you’re like 95 percent of the people we talk to on any given day, you’re currently chewing through your fingernails with anxiety over the fate of your favorite character in the next “Song of Ice and Fire” book. Author George R.R. Martin loves killing off major and beloved characters almost as much as he loves taking five years between books, and he ended “A Dance with Dragons” on the kind of cliffhanger that should really be outlawed.
Never fear, though — University of Canterbury statistician Richard Vale is here to try to relieve your nerves. He performed a Bayesian analysis of the number of PoV chapters each character has had in previous ASoIaF books in an attempt to predict how many chapters they’ll get in the upcoming “The Winds of Winter,” and if that sounds like the geekiest thing a person has ever done, well, that’s kind of the point, isn’t it? Vale’s assumption (which seems sound) is that dead characters don’t get PoV chapters, so if his predictions show a high probability that a character will have zero PoV chapters, it makes it that much more likely that the character’s going to bite the big one.
As Vale points out, there are a lot of problems with his analysis; it’s based on a very small sample size, and it doesn’t take into account common sense (e.g., the model doesn’t care that Ned and Catelyn are already dead; it’s going to try to predict their chapters regardless). In Vale’s words, The question to be answered could be expressed as: “‘What could be predicted about future books if we knew nothing about the existing books except for [the number of PoV chapters for each character]?'” Now that we think about it, this doesn’t really help us alleviate our anxiety at all. Still, the paper is definitely worth it for the references section alone:
If you want to read Vale’s paper, which is really quite interesting if you’re into that sort of thing, it’s available in its entirety here. Be warned, however, that there’s very little dragons and warfare and a whole lot of Greek letters and phrases like “Poisson distribution” and “confidence interval.” Oh, and spoilers! Lots of spoilers.