Joel C Ryan/Invision/AP
There’s no way all of them will survive.
WINTER IS COMING

Charted: Machine learning says who’s most likely to die in “Game of Thrones,” season 8

By Hanna Kozlowska

Possible spoiler alert! The final season of the hit show “Game of Thrones” is upon us, with the first episode scheduled to air Sunday in the United States at 9pm ET on HBO, broadcast simultaneously at 2am UK time on Sky Atlantic.

The internet is teeming with theories on what will happen as the series, with its myriad plots, nears the end. Since the show is known for killing off both key and side characters, the question fans often pose is: Who will die next?

Students at the Technical University of Munich tried to answer the question using a machine-learning model. The math is not looking good for two of the Stark siblings, as well as the mighty Clegane brothers:

There’s better news for the Lannister family, and a couple of other beloved characters:

The team explains their modeling on a website. They extracted data from two major fan sites—The Wiki of Ice and Fire and Game of Thrones Wiki—allowing them to create a dataset that described each character using the same type of information. “Our next task was to find the feature set that can best distinguish dead from alive characters.”

They predict both the likelihood of death in season 8, and a character’s longevity for 20 years ahead of the current year in the show. Their analysis involved two methods. The first is similar to methods used in scientific research that examine the effects of treatments and complications for cancer patients, or the correlation between seismic events. The factors that they considered were house (Stark, Lannister), lovers, marriage, titles, major/minor character, and gender.

The second is a neural network technique that predicts the likelihood of a character dying in any given year, allowing for more complex analysis.

On the website, you can click through dozens of characters to view their likelihood of death, both in the TV show and book series, and learn how various factors like their heritage and gender affect their chances. See how your favorite characters fare in the algorithmic battle. Or just… watch the show.