Emilia Clarke as Daenerys Targaryen in Game of Thrones | GoT Facebook
Emilia Clarke as Daenerys Targaryen in Game of Thrones | GoT Facebook
Text Size:

Data scientist Taylor Larkin, relying on an algorthim, says a number of important characters may be killed, gives Stark girls better odds of survival.

Bengaluru: For fans of the House of Targaryen, winter is coming, if this computer algorithm is to be believed.

An American data scientist, relying on a machine learning (ML) algorithm, has predicted that a number of important characters will be killed off in the final season of the Games of Thrones. The final season of the phenomenally popular show, based on Song of Ice and Fire series by George R.R. Martin, is to air in 2019.

According to Taylor Larkin, a data science evangelist with DataRobot, leading the list is Daenerys Targaryen with an 83.77 per cent chance of dying, followed by Jaime Lannister (72.91 per cent), Tyrion Lannister (70.76 per cent), Bran Stark (66.02 per cent), Cersei Lannister (60.39 per cent) and Jon Snow (58.99 per cent).

Fans of the Stark girls can breathe a little easier as the odds of them dying are relatively lesser — Sansa Stark has a 50.28 per cent of being killed, while Arya has even better survival odds at 49.04 per cent.

Latest news on Game of Thrones | ThePrint.in
Illustration by Siddhant Gupta | ThePrint

The Game of Thrones is notorious for unpredictably killing off main, important characters. Now with the fight for the Iron throne set to reach its final lap, there has been a lot of speculation among fans on which of their beloved characters would, literally, face the axe.

Past deaths a pointer

As with any other ML algorithm, Larkin says, his was also fed past data. In this case, of who all have died in the show, which houses they belonged to, and how many people associated with them have died. Also factored was the gender, age, nobility status, and more of those left in the show. The algorithm then uncovered patterns in previous deaths, and applied those patterns to existing characters to come to results based on probability.

The most important reason in determining the probability of death in the majority of characters was whether or not they had dead relatives, Larkin has explained in his post about this algorithm’s result. He interprets this as the amount of turmoil in the family. Characters in the Game of Thrones universe are indeed intent on wiping out whole families and houses to secure their positions of power, and no one has had more family wiped out than Daenerys.

Larkin also points to age as a factor. Compared to the Stark kids, the Lannisters are much older, increasing their chances of death. The data scientist says he noticed a raging trend where the older a character grows, the higher their chances of dying, just as in real life. Someone aged 80 was twice as likely to die as someone aged 30.

Gender also plays a big role. There are more men in the series than women, and thus more men die, mainly thanks to them being shipped off to battles.

All about LightGBMs

The algorithm is built on a new tree-based approach to handle data, called Light Gradient Boosted Machines (LightGBMs). In machine learning, the algorithms delve deeper and deeper into data structured in a tree format, with branches extending downwards. But LightGBMs have wider trees with nodes containing several branches, but none of them too deep. This enables algorithms to find general patterns quicker in data.

Machine learning is the way of the future but not necessarily in the Game of Thrones world. Martin and the writers of the show have one big card in their pocket — that of creativity and imagination. Nearly every death in the books and the show has been unpredictable, purposely going against patterns that consumers would form.

ML algorithms very much reflect how we form patterns as well, so at the end of the day, it is definitely quite likely that all ML predictions are entirely inaccurate because of the surprise wild card. “We can expect the season finale to have twists and turns intentionally added in by the writers that will not be picked up in the historical data gathered from the previous seasons,” wrote Larkin. “Perhaps predicting the likelihood of death would be better suited for applications of game theory rather than via machine learning.”

Ultimately, it is going to be the author of the books who decides which characters to keep and which ones to wipe off.

Who are your guesses for the top three deaths in the next season? Tweet to us at @ThePrintIndia.

Subscribe to our channels on YouTube & Telegram

Why news media is in crisis & How you can fix it

India needs free, fair, non-hyphenated and questioning journalism even more as it faces multiple crises.

But the news media is in a crisis of its own. There have been brutal layoffs and pay-cuts. The best of journalism is shrinking, yielding to crude prime-time spectacle.

ThePrint has the finest young reporters, columnists and editors working for it. Sustaining journalism of this quality needs smart and thinking people like you to pay for it. Whether you live in India or overseas, you can do it here.

Support Our Journalism

Share Your Views

LEAVE A REPLY

Please enter your comment!
Please enter your name here