Game of Thrones has been famously protective of its storylines, going so far to protect its secrets that they’ve stopped handing out scripts to actors, instead delivering dialogue to characters line-by-line through an ear piece.
Could a Boston-based startup ruin all the fun?
DataRobot, which uses automated machine learning for predictive modeling, analyzed roughly 2,000 Game of Thrones characters to predict who will die in the next season.
The company applied a newly-developed tree-based algorithm approach to figure out who would get "Ned Stark’d" in the upcoming eighth and final season, which is set to air in 2019.
Here are the results, listed according to chance of death:
- Daenerys Targaryen – 83.77%
- Jaime Lannister – 72.91%
- Tyrion Lannister – 70.76%
- Bran Stark – 66.02%
- Cersei Lannister – 60.39%
- Jon Snow – 58.99%
- Euron Greyjoy – 54.95%
- Sansa Stark – 50.28%
- Arya Stark – 49.04%
- Gendry – 39.87%
"One of the most unique aspects of the series is its fearless approach to eliminating major characters," DataRobot wrote in a blog post. "This defining trait has led to some fascinating predictive analyses, such as creating Bayesian survival models and leveraging machine learning techniques to predict the likelihood of death for our beloved characters."
The blog goes on: "In particular, the latter analysis scraped data from a fan-created Wiki in order to make predictions on approximately 2,000 characters using various descriptive characteristics such as their house of allegiance and gender."
The full analysis, including its methods, is available here (beware of spoilers!).
