A group of IT students at Munich's Technical University has developed an algorithmic piece of software to predict deaths of characters in the upcoming season of the cult television series Game of Thrones.
Speaking to dpa on Thursday, research assistant Tatyana Goldberg said the system analyses 24 variables, such as the number of murdered family members, to generate the probability of a character's death.
"The likelihood of a character dying increases if the mother is already dead," Goldberg said, referring to the unusually high number of characters that are killed off in the show.
The team says its algorithm has been able to confirm previous deaths in the series with a 74-per-cent accuracy rate.
"Many people have written to us that this project finally shows what machine learning can do - finally something not to do with finances," said Guy Yachdav, a doctoral student involved in the project.
According to the software, the most likely to die in the upcoming season, are Tommen Baratheon (97 per cent probability) and Daenerys Targaryen (95 per cent probability).
The show, which is watched by millions of fans around the world, begins its sixth season on April 24 in the US.