Who will live and die, according to AI

Scared of losing your favorite Game Of Thrones character? Technology could hold some comfort. 

A group of what we can assume are extremely nerdy computer science students at the Technical University of Munich have developed an algorithm that predicts who will live and who will die in Game of Thrones.

Using data they extracted from the GoT show and book wikis, the students created an algorithm that weighs factors like gender, age, and house to determine the Potential Likelihood of Death (PLOD). 

For example, people in House Stark are more likely to die than people from House Lannister (sorry but it’s just true!!), so the algorithm considers house affiliation as one factor. 

The students used two types of models for the project, modeled after survival-rate algorithms used in medicine. The factors considered in one are house, lovers, marriage, titles, major/minor character, and gender. The other also considers things like popularity, gender, page rank (on the wiki), number of relatives, age, culture, house, region of house, allegiances, books the character was part of, episodes the character was part of, locations, and titles.

So what does it have to say about the fates of our favorite guys and gals this side of The Wall? 

Apparently, the algorithm predicts that Daenerys Stormborn of the House Targaryen, First of Her Name, the Unburnt, Queen of the Andals and the First Men, Khaleesi of the Great Grass Sea, Breaker of Chains, and Mother of Dragons… has a 99 percent likely survival rate!

Here are the top five characters most likely to survive, according to the algorithm: 

1. Daenerys Targaryen

2. Tyrion Lannister

3. Varys

4. Samwell Tarly

5. Jaime Lannister

On the other hand, everyone’s favorite boozin’ and whorin’ sellsword, Bronn, is 93.5 percent likely to die. Bronn had to know he was never gonna get that castle :/

You should probably say goodbye to these characters:

1. Bronn

2. Gregor Clegane

3. Sansa Stark

4. Bran Stark

5. Sandor Clegane

As Game of Thrones approaches its televised conclusion, we’re all wondering who will come out on top, and who will end up six feet under (my colleagues certainly have their share of predictions). This AI project is fun, and plumbing the algorithm for answers gives us something to do as we wait for the Sunday night Season 8 premiere. 

Still, Game of Thrones is at its best when its characters and viewers come to understand that the “Game of Thrones” is not just about the binary win-loss game of power, but is instead about understanding how power can bring out the best, and more often, worst, in leaders and followers alike. The sad, disgusted, and heartbroken determination Jaime Lannister exhibits when he leaves Cersei to aid with the defense of humanity, or the disappointment Tyrion feels as he watches Daenerys barbecue dissenting noblemen, show that the title Game of Thrones is inherently subversive: the quest for power is not a game at all to the people who die as a result. 

The students’ algorithm plays on a lot of what’s made Thrones, and other shows, such an internet phenomenon: often, high-budget dramas become about teasing apart mysteries and using crowdsourced obsession to ferret out likely outcomes on forums like Reddit. Networks only add fuel to this fire when they use the morally uncomplicated ~drama~ to market the heck out of their shows, something that HBO has not shied away from this season. But this tendency to favor plot over substance can erase the finer points a show is making about its topic, whether that’s the ultimate quest for power (Thrones), religion and afterlife (The Leftovers), or the nature of consciousness (Westworld).

At its best, Game of Thrones demonstrates that power is not merely a game; instead, the whims of rule affect the lives and fates of all people. Reducing GOT to an algorithm plays into the least unique aspects of the show, while ignoring what makes it good.

Still, the website‚ A Song of Ice and Data, is pretty dang fun to play with.