** Predictive analytics and machine learning are now potentially spoiling popular TV shows. Don’t read any further if you’re not caught up on Game of Thrones season 5, or just plain dislike spoilers and speculation about your favorite characters.**
As long as the Internet has been around, there have been people comparing and compiling details about every topic of interest across the globe, cramming the data into wikis, building web pages chock full of facts, and using it to one-up each other on message boards. It’s how we’ve arrived at information banks like IMDB, Wikipedia, and Numbeo. One of these massive online collections is the Wiki of Ice and Fire, a fan-created website based on George R.R. Martin’s A Song of Ice and Fire (ASOIAF) series and HBO’s popular television series Game of Thrones.
The Wiki of Ice and Fire is very comprehensive; it has over 7,000 articles summarizing information from across five novels, five seasons of the television series, and other related media. It’s a rabbit hole for fantasy novel aficionados. But recently, a group of students and their advisors at the Technical University of Munich decided to make real use of all that data. They’ve designed a machine learning algorithm that analyzes online data from Game of Thrones wikis and Twitter, and predicts the likelihood of character deaths.
Before I explain how analytics has become the new spoiler, let’s have a quick refresher on what it actually is. In a nutshell, analytics is the distilling of large amounts of data into easily consumable summary information, and predictive analytics tries to look into the future by leveraging historical data and statistic technique. But back to the mortality of our beloved characters.
By pulling data from the 2,028 characters that span the story, the team in Germany has pulled some interesting statistics that help them predict outcomes:
So according to the data available, gender, rank and age all play a part in predicting character death likelihood. Their use of machine learning directs computers to make predictions by learning from a large number of examples from the past, then compiling statistics to predict likelihood in the future. In this project, the team aimed to find features that are common to already dead characters, then use these same features to predict likelihood of death in alive characters.
But what about Jon Snow? When we last saw him at the end of last season’s cliffhanger finale, he was looking quite dead-ish. Everyone associated with the show has insisted that he is “deader than dead”, but what does the data have to say? According to the algorithm, Jon Snow’s likelihood of death is only at 11%!
Farewell, Stannis. It was statistically inevitable, I guess, what with you leading an army and all.
Check out their rankings. Basically LOTS of people are more likely to die before Jon Snow. Based on the algorithm and their Twitter sentiment analysis, Jon Snow is way too popular to kill off at this point. Not that it’s stopped the show before.
GAME OF THRONES finale: that was...was...let me just say it was not what I expected. :-(— Stephen King (@StephenKing) June 15, 2015
OMG!!!!!! NO!!!!!!!!!! 😥😥😥 NO!!!!!! Why???? Not John Snow!!!! 😢😢😢 I'm refusing to believe this!… https://t.co/Is1RVGyw0X— Karina Smirnoff (@Karina_Smirnoff) June 15, 2015
But in this instance, the hugely negative sentiment and response from fans is an indicator that the showrunners have something up their sleeves, and that Jon is still a contender for the Iron Throne. Thus, if Kit Harrington is not coming back to the show in any form, and Jon Snow is truly dead, it would be a statistical shocker.
And this is all real analytics telling us the likely outcomes of a fictional character’s death based on actual data extractions and statistical comparisons. There’s no telling what we can discover when we start looking more closely at the data in our world, instead of in Westeros.
Download the State of Analytics report now to get insights on analytics trends from more than 2,000 business leaders worldwide.