Spoiler-wary fans of the show might be horrified to learn just which beloved character has a more than 70 percent chance of dying in the coming books or TV series.
But that's just what can be found on got.show, the website where the team have published the results of their analysis.
The project was the brainchild of Dr Guy Yachdav, a guest lecturer at TU Munich and self-described “devoted fan, consuming any piece of information I can get about the characters and the plot” of Game of Thrones.
“When I was done reading the books and the show was over, I was left with the question of which of my favourite characters are going to survive,” Yachdav told The Local.
That's when he had the idea to set a 40-strong class of programming students to the task of creating a programme to trawl the text of the books and the internet for clues as to who might be next to die.
Once the project was approved, the class had just 50 days to design and build the program.
An algorithm crunches together 24 properties for every character – such as how many people linked to him or her have already died – to produce a probability that they might be next for the chop.
“We use the same techniques as in our research group which does analysis of genomic data, proteins, DNA and genes, using computer algorithms to answer really complex biological questions,” Yachdav explained.
One of the hardest jobs was building code to analyse public sentiment about the different characters on Twitter – a “fairly new” technique that's also been applied to politics and the stock market by private-sector companies.
“Twitter is not really equipped for us to ask for three million tweets and analyse them in two minutes,” Yachdav said.
Not just a jolly
The enormous body of freely-available information about Game of Thrones was a first step into handling big data for the students – and one that made it a bit more fun for those of them that were fans.
“It was very important to have the students learn how to handle this data, construct it, clean it, organize it and then apply algorithms,” doctoral student and teaching assistant Tatyana Goldberg told The Local.
“We know that not everyone is so crazy about biological data as we are, and so finding this trove of Game of Thrones data online was a real goldmine.”
And while fans of the show might be horrified to have such concrete data, students say the intensive exercise has taught them a lot.
“Getting to apply what I studied in my Bachelor's was hugely important for me,” Christian Dallago, who managed the project and reviewed other students' code, told The Local.
“We already have a project that uses the ideas from this Game of Thrones work and applies them to bioinformatics.”
As for how confident the team are in the results of all that hard work – would they put a bet on their predictions being fulfilled?
“We're confident in our algorithm,” said Yachdav, “but I'm not really a gambling person.”
The program has successfully predicted 74 percent of all the deaths in the show so far.
But “the way that [Game of Thrones author] George Martin and the showrunners work, nobody can make a guess,” he went on.
And as a confirmed fan, Yachdav is not immune to the hope that the project might get the team noticed by Martin himself.
“If the big almighty Mr. Martin would come down to us humans, that would be the biggest compliment,” he said.
In the meantime, the Game of Thrones dataset could be used for further projects – or the team might move on to something involving Star Wars…