Written by Anodot

Who Will Win and Who Will Die: ‘Game of Thrones’ Fans Use Data Analysis to Predict What Happens Next

In the era of Peak TV, there is probably no more emblematic version of worthy binge watching than “Game of Thrones”. At the cusp of the series’ eighth and final season, the internet is buzzing with some painstaking analysis – using algorithms, AI, and big data sets, to find hidden Easter eggs.

The numbers capture the mind-boggling detail author George R. R. Martin weaved into his novels and how easy it is to get lost in a fictional world as intricate as our own.

Here are some of the best uses of data analysis in Game of Thrones. (Advisory: spoiler alert.)

Death, by the numbers

The jaw-dropping deaths of the show’s central characters are among its guiltiest of pleasures. But, is there a way to prepare yourself for seeing your favorite character’s head on the chopping block? As it turns out, data-savvy fans across the web have been addressing that very question.

Technical University of Munich students have developed an algorithm to help you predict who will survive the coming episodes, relying on scraped data from the Game of Thrones Wiki and Wiki of Ice and Fire, and factors such as gender, age, and house affiliation for a project they’ve titled “A Song of Ice and Data.” With a machine learning program that’s learned from its past errors, the project now predicts that Daenerys Targaryen, Tyrion Lannister, Varys, Samwell Tarly and Jaime Lannister have the best chances of staying alive, while the alcoholic, womanizer Bronn, on the other hand, is predicted to die by a full 93.5 percent. Interestingly, their Potential Likelihood of Death (PLOD) algorithm is based on another, admittedly more serious survival-rate system applied in the medical field.

A data science expert at DataRobot has used a newly-developed machine learning approach known as light gradient boosted machines (LightGBMs) to further hone in on the reasons why our protagonists keep getting killed off. According to his analyses, 67 percent of those affiliated with the House of Targaryen were killed off, while only 5 percent of the Tyrells met the same fate. He concluded that the number of dead relatives, or, in other words, the amount of turmoil within any one particular house, was a major predictive indicator for death.


For each machine learning algorithm in DataRobot, the platform builds a blueprint that optimizes the data for that specific algorithm. (source: DataRobot)


Using a technique called feature impact, DataRobot analyzes the extent to which each characters’ house of allegiance, age, whether they had dead relatives or not, and their gender may have contributed to their death. (source: DataRobot)


Over at The Washington Post, their elaborate illustrated webpage allows fans to relive every one of the shows 5,862 deaths, starting with the shocking beheading of the central protagonist, Ned Stark. Each death – which have to be updated frequently as each new episode drops – are categorized by house allegiance, place and time of death and method and circumstances of death. What are the deadliest places in the seven realms? That would be Winterfell (3,709 deaths), beyond the wall (993 deaths), and King’s Landing (334 deaths).


It also gives us a good chronicle of really how much bloodshed has been shown since the show first aired in 2011. In the first season started an impressive 59 deaths quickly set the tone. Season 3 featured a similarly “modest” death count, at 87, but one of the most memorable scenes: The Red Wedding, in which four major characters were killed, and the multi-layered Arya Stark was established as a ruthless warrior.

Season 6 was especially bloody, with 540 deaths, a full 168 of which happened in the Battle of the Bastards, and a near annihilation of full houses: Baratheon, Bolton and Tyrell.

Season 8, which already has 3,523 deaths, promises to be no less brutal.


What’s in a name?

It seems that GOT fever has also seeped into nearly every facet of our non-TV watching lives. Since its 2011 premier, it’s amassed one of the widest and most avid fan bases, including many who chose to name their newborn babies after their favorite, fiercest characters. According to an analysis of USA Name Data, a public dataset created by the Social Security Administration, and published on the website Kaggle, the name “Daenerys” made its first appearance ever in 2013 – just after Daenerys emerged as a protagonist in Season 3. The use of the name continued to rise to 60 children in 2016 and 2017. The name “Brienne” joined as a first-timer to the name database, as the name for 5 children in 2016 and 2017.

Unsurprisingly, “Khaleesi” has gained in popularity, becoming the name for some 400 American children in the past few years. As Arya Stark, too, garnered her fearsome, bad-ass reputation, the numbers of baby “Arya”s spiked sharply.


In 2017, more than 400 American children were named Khaleesi, after GOT character Daenerys Targaryan’s title in the series (“Khaleesi” is used to refer to the wife of the khal in Dothraki culture). (source: Kaggle)


Winter is here, but what’s next?

It’s unknown to what extent this cultural phenomenon will continue after the highly anticipated finale. But what we know is that the show continuously appeals to a truly broadly-spanning demographic group. Season 7’s opening episode received nearly 98,000 mentions on social media, equally spread among men and women. In an example of massive crowd-sourcing and a rare pop-culture context for the Swarm platform, GOT fans are gathering to predict that Jon Snow will take his seat at the Iron Throne.

Of course, the show is famous for defying expectations, and fan theories have proven wrong in the past. But at a time when virtually everything is divisive in America, such an escapist debate is clearly a welcome change of pace.

footer background