In The Bestseller Code: Anatomy of the Blockbuster Novel, Jodie Archer and Matthew L. Jockers make a provocative claim: A book’s fate in the marketplace can be determined in advance, and not by the opinions of smart literary critics, book publishers or savvy writers. Instead, they argue that bestseller status can be predicted, with more than 80 percent accuracy, by a computer.
To an avid reader, attuned to the seeming incongruity and unpredictability of the weekly New York Times bestseller list, such a claim may seem akin to heresy. But the book’s co-authors, armed with a secret algorithm, unpack precisely how a book like Fifty Shades of Grey can reasonably, accurately and persuasively be compared to something else entirely, like, say, Donna Tartt’s Pulitzer Prize-winning novel, The Goldfinch.
There’s an argument here about mass reading that is undeniably pleasing: The highbrow and the lowbrow are not, in fact, so far apart as most would believe. The book proceeds with more seemingly impossible facts, such as this: Algorithms can predict, with surprising levels of accuracy, whether a book was written by a male or female author only by looking at the writer’s use of pronouns. Seemingly insignificant details add up. And computers excel at this kind of granular counting.
Using a corpus of just over 5,000 books (500 of which are NYT bestsellers), the researchers have trained the computer to track more than 20,000 discrete characteristics. These items reveal patterns about all sorts of things—from topic to plot, from character to style. And along the way, the researchers unpack how various titles and authors you already know—from Danielle Steele to John Grisham—exemplify the patterns they are tracing, even as they move toward solving a particular and engrossing mystery: what working writer today best exemplifies popular approaches to novel writing. For readers interested in books about books, this is a title not to be missed.