Algorithm predicted who will bite the dust next in Game of Thrones

Students at the Technical University of Munich have created an algorithm which predicts the likelihood of death percentage for Game of Thrones characters. If you think computer-generated guesses will be accurate, then this article could contain spoilers.

Game of Thrones Hall of Faces
Credit: HBO

This article could involve spoilers, if you believe an algorithm can accurately predict who will be killed off in the Game of Thrones. Of course if you were to believe the Game of Thrones season 6 poster and trailers, then it looks like all the big players are dead and in the Hall of Faces.

After being curious enough to check out the algorithm’s predictions, I thought, No! Say it isn’t so! The first thing I saw was that Daenerys Targaryen, Mother of Dragons, “got a pair of cement shoes.” Clicking through though, shows a page explaining why she might win the Game of Thrones and sit on the Iron Throne. The reasoning behind the two extremes is that the editorial staff believes she can win the Game of Thrones, but their machine learning algorithm predicts Daenerys has a 95% likelihood of death.

Just great…why not go ahead and kill off Jon Snow? Oh wait, that happened; maybe I’m not just in denial as the algorithm predicted Jon has only an 11% likelihood of death. The same algorithm, created by computer science students at the Technical University of Munich in Germany, accurately predicted 74% of characters’ deaths in the HBO show and books.

If you watch the show, then you are all too aware that characters drop like flies; 61 characters were killed off in the first five seasons. The students’ “A Song of Ice and Data” machine learning project used common features of dead characters to predict the percentage likelihood of death for the remaining characters. First, the students extracted data from the Wiki of Ice and Fire on the 2,028 characters in Game of Thrones. They came up with more than 30 features to describe each character. Then they used machine learning to statistically compare features of dead and alive characters in order to predict which characters are going to kick the bucket.

At least it sounds like it wasn’t a boring class. Dr. Guy Yachdav, who came up with the project for his class, called it “a lot of fun.” He added, “In its daily work, our research group focuses on answering complex biological questions using data mining and machine learning algorithms. For this project we used similar techniques. Only this time the subject matter was a popular TV show. The epic scale of the worlds created by George R. R. Martin provides an almost endless resource of raw multi-dimensional data. It provided the perfect setting for our class.”

Unlike the computer-generated prediction for the Mother of Dragons, some of the other algorithm’s “who is most likely to die next” predictions are far less upsetting…unless you are rooting for the bad guys.

House Lannister

Get ready for the boy on the Iron Throne to cease to exist as the algorithm rates Tommen Baratheon at 97% most likely to die next. Like it or not, Tyrion Lannister has a 70% likelihood to die; Jaime Lannister has a 64% chance to croak. If the algorithm is correct, then there is only a 16% chance that Cersei Lannister will die.

House Stark

“Arya Stark will surely bite the dust,” according to the algorithm which marked Arya with a 65% RIP tombstone; going to her page shows Arya has a 68% likelihood to die. Her little brother Bran is rated at a 71% chance of death. In fact, it appears that the most hopeful survivor for House Stark is Sansa Stark who only has a 3% likelihood to die, which is even a better chance than the 11% given to Jon Snow.

If you are curious about other characters, then look up their predicted death or survival chances. It will be interesting to see how accurate the algorithm was as the new Games of Thrones season progresses.

Computerworld's IT Salary Survey 2017 results
Shop Tech Products at Amazon