Pop culture obsessives writing for the pop culture obsessed.

This algorithm predicts each Game Of Thrones character’s likelihood of death

Game Of Thrones has long warned fans that “all men must die,” but those with an emotional attachment to the characters probably wish the series would be a little more specific than that. While those looking to be spoiled on upcoming deaths no longer have the books to rely on, this brand new algorithm on the website A Song Of Ice And Data may be able to help them out. Which is good news for fans, but bad news for Tommen, Stannis, and Daenerys, who come out on top as the characters most likely to die this season.


Personally, we’re not sure we agree with you a hundred percent on your police work, there, Lou. For one thing, Stannis is already dead (although, to be fair, we didn’t know that for sure until the sixth season premiere, so perhaps that just means the algorithm is working). More importantly, it seems incredibly unlikely the show is going to kill off Daenerys Targaryen, even if it is famous for unexpectedly killing off its main characters. But, hey, we could always be wrong in our optimism and, regardless, A Song Of Ice And Data’s algorithmic process remains fascinating.

Relying on the Game Of Thrones wiki, the creators fed character data based on 24 different features (age, gender, nobility, culture, relatives, etc.) into their “Machine Learning” algorithm in order to “find the feature set that can best distinguish dead from alive characters.” From there they did a bunch of analysis we’re just going to have to assume is correct because how the hell are we going to double-check—we’re pop culture writers, not mathematicians.

It should be noted that most of this data is pulled from the books, although the site does collect information on the TV show as well. And, of course, the algorithm can’t deal with meta factors like, say, the popularity of Peter Dinklage as an actor. Which is why it gives Tyrion Lannister a 70 percent change of dying even though we’re all pretty sure that’s never going to happen.


But regardless of actual predictive accuracy, Song Of Ice And Data now serves as an impressive resource. Searching by character you can assess information on their title, house, spouse, and the books in which they appear. Plus their likelihood of death, information on how often they’re being discussed on Twitter, and an interactive map that charts their progress through the series (which is kind of similar to this one). And all that data can be used for more than just death predictions too. For instance, it turns out there are twice as many male characters as female ones in the books, with men more likely to be nobility than women. But women actually have a higher rate of survival, probably because they aren’t being led into battle all the time.

The whole site is a treasure trove for data nerds. But those who don’t want to dig through it can refresh the main page to get a new “match up” between two random characters, which compares how likely they are to die:


And for those hoping for some sort of happy ending to the bleak series, the site also predicts who has the best chance of living through it all. The top five survivors include Sansa Stark (3 percent chance of death), Jon Snow (11 percent), Cersei Lannister (16 percent), Mace Tyrell (18 percent), and Roose Bolton (28 percent). So apparently they’re exempt from the whole “valar morghulis” thing.

Share This Story

Get our newsletter