We collected 11 data points (the categories on the main page) on all of the previous winners, the shortlist this year as well as previous losers. By feeding this information in to a machine-reading algorithm, we were able to use previous winners to 'train' the algorithm as to exactly what a winning Man Booker Prize novel looks like. As a result of the algorithm using neural networks, it was able to weight different data points – working out what was important to the winning books and what was not. Once the algorithm was trained, the shortlist for this year was run through, and each book scored on its likelihood of winning. As soon as the 2016 shortlist is announced, we will use the algorithm to predict a new winner.
Whilst we collected the information, some interesting statistics emerged:
Author nationality - Over half have been British (56%)
Gender - 66% have been male
Age - 49 (average)
Number in canon 7 (average)
Genre - Historical fiction (48%); contemporary fiction (38%)
Length - 370 (average)
Protagonist gender - 84% have been male
Perspective - Third person (66%)
Setting - 22 out of the 50 previous winners have been set in the UK and Ireland
Goodreads.com score - 3.7 (average)
Readability - 75.0 (average)
Readability is classified using the Flesch-Kincaid scoring method, based on sentence and word length. The higher the score, the more 'readable’ a piece of text is e.g. 90-100 can be fully understood by an average 11-year old, 60-70 by a 15-year old etc. To calculate the scores, we entered the first two pages of each winning novel into readability-score.com, which produced the figure. As such, it is not representative of the books as whole bodies of work; rather it is a simple benchmark with which to rank them.
This is where the winning book comes in the order of the author’s adult fiction body of work.
Synopses, scores and author statistics: goodreads.com
Book covers: books.google.com
Readability: readability-score.com and amazon.co.uk