Mathematical model tackles 'Game of Thrones' predictions

Mathematical model tackles 'Game of Thrones' predictions
Posterior predictive distributions for the number of POV chapters for nine characters in The Winds of Winter. Credit: arXiv:1409.5830 [stat.AP]

Take events from the past, build a statistical model, and tell the future. Why not apply the formula to novels? Can contents in future books be predicted based only on data from existing ones? Richard Vale at the University of Canterbury in New Zealand, said The Physics arXiv Blog, has taken on the challenge in predicting content of as yet unpublished novels in the "A Song of Ice and Fire" series by George R R Martin. The novels are the basis of the television series, "Game of Thrones." The series has five books and two more are awaited. Before proceeding, it should be emphasized that the paper comes with a spoiler alert, so avoid linking to Vale's study if you have not read the first five books.

As The Physics arXiv Blog explained, "Each chapter in the existing books is told from the point of view of one of the characters. So far, 24 characters have starred in this way. The approach that Vale has taken is to use the distribution of characters in chapters in the first five books to predict the distribution in the forthcoming novels."

After creating a model, Vale runs a computer program to find the parameters in the model that best fit the data. He uses the model to find the probability distributions of the number of chapters that each character will star in, in book 6 and book 7. What kinds of predictions result? They include predictions about certain characters unlikely to star in any chapters and if one particular character is likely to be dead. Vale's paper submitted to arXiv is "Bayesian Prediction for The Winds of Winter." As Vale described it, "Predictions are made for the number of chapters told from the point of view of each character in the next two novels in George R. R. Martin's emph{A Song of Ice and Fire} series by fitting a random effects model to a matrix of point-of-view chapters in the earlier novels using Bayesian methods." There is also a "Spoiler Warning" that readers who have not read all five existing novels in the series should not read further "as major plot points will be spoiled."

The blog commented on how this is a "fascinating exercise in statistical modeling that will do more to introduce the process to a wider range of people than any number of textbooks or Wikipedia entries."

Vale is a Lecturer in the Statistics Department at the University of Canterbury. He has a PhD in mathematics from the University of Glasgow and was an HC Wang Assistant Professor at Cornell University. Vale acknowledged several shortcomings in his model— such as not dealing with the possibility of new characters being introduced and a model resting on a relatively small amount of data. Robin Kawakami of The Wall Street Journal, writing in the Speakeasy blog, quoted Vale saying "Game of Thrones" cannot be predicted using statistics alone. He said his project can be viewed as "fun data analysis." He said in an e-mail to Speakeasy, just as many people make fan art by drawing favorite from books, "As a mathematician/statistician, this mathematical is my version of fan art."

Explore further

Brazil's Vale to halve stake in Belo Monte dam consortium

More information: Bayesian Prediction for The Winds of Winter, arXiv:1409.5830 [stat.AP] (PDF)

© 2014

Citation: Mathematical model tackles 'Game of Thrones' predictions (2014, September 30) retrieved 20 September 2019 from
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

Feedback to editors

User comments

Sep 30, 2014
You might be able to predict the likelihood of a character being dead or alive, "promoted" or "demoted" based on their relationships to other characters. Build a web and rate relationships on a scale from positive 5 to negative 5 based on some combination of mathematics of relationships and "intangibles" the reader may infer.

Maybe 2/3 the score might be based on a computer algorithm which counts each interaction between the two characters, and each character's positive, negative, or neutral mention of the other character, and so forth.

About 1/3 of the score would come from the reader interpreting "intangibles" such as false leads, present and past persistent injuries, and so forth, and rating the interaction from -5 to 5.

After finding these values for all characters, the algorithm would total all positive and negative mentions and positive and negative conditions of a character and assign a probability that they have been killed or mortally wounded.

Sep 30, 2014
It might be useful to study how well the algorithm can "post-dict" the previous books.

That is to say, use book one data to try to predict book 2. Then use book 1 and book 2 data to try to predict book 3.

I assume this approach was already done, but sometimes people over-look things.

Another approach would be to see if the reverse is true. Can you predict the contents of book 1 using the data from books 2 through 5?

It seems to be under-predicting, because it makes only about 4.5 to 5 chapters among the 9 most important characters. This seems a bit too little.

I admit I haven't read past the second book of the series. Martin has a very graphic writing style which was at times disturbing in the case of two particular scenes, which really turned me off from the series.

However, I am familliar with a lot of high fantasy novels.

It would be interesting to see this same approach applied to the Wheel of Time and Sword fo Truth series, to see if it can predict character PoV there.

Oct 01, 2014
Bah. Futile exercise.
We already know how it will all end. I was able to steal the end scene from HBO HQ with extreme danger to my life. But here it is, behold!


Please sign in to add a comment. Registration is free, and takes less than a minute. Read more