Scientists claim to have written an algorithm that can predict bestsellers.
Lewis Jones isn't so sure.
In his entertaining Bestsellers: A Very Short Introduction (OUP, 2007), the indefatigable John Sutherland concluded that "to look for significant patterns, trends or symmetries" among bestsellers "is, if not pointless, baffling".
The truth of Sutherland's argument is demonstrated by how publishers, who are naturally keen to find significant patterns, trends or symmetries, have often been spectacularly baffled, and have rejected such future bestselling classics as Animal Farm (dismissed by TS Eliot at Faber as "not convincing"), Lord of the Flies, The Spy Who Came in from the Cold and the first Harry Potter.
The Bestseller Code
sets out to challenge Sutherland's axiom.
Jodie Archer, who worked in publishing before writing a PhD thesis on bestsellers, and Matthew Jockers, a professor at Nebraska University and a pioneer in "digital humanities", particularly "stylometrics", have devised what they call a "bestseller-o-meter".
Their starting point is The New York Times bestseller list.
Every year in the United States more than 50,000 new works of fiction are published, excluding self-published ebooks, of which about 200 make the list - less than half a per cent.
A fraction of those remains on the list for more than 10 weeks, and a few of that fraction sell a million copies in a year.
To find out what distinguishes the few, Archer and Jockers fed a computer almost 5,000 novels, including just over 500 NYT bestsellers (thereby altering the proportion of bestsellers in the mix from 0.5 to 10 per cent, naughtily), and then asked it to "predict" which would succeed.
It duly did so, "with an error rate of only 10 to 20 per cent". Doesn't that "only" look rather desperate?
Quite early - on page 29 - Archer and Jockers announce that theirs is not a "how-to" book, and "comes attached to no guarantee".
They might have announced this still earlier - on page one, say - but to have done so would have lost them most of their potential readers.
They claim to be interested in "widening access, potentially, to the career of writing", which is doubtless a noble ambition but somewhat at odds with their book's professed non-"how-to"-ness.
So what is the point of The Bestseller Code? Its authors hope that it will "lay some of that ineffable je ne sais quoi of talented writers bare", and reveal not only the "latent DNA", "core DNA", "topical DNA" and "foundational DNA" of blockbusters, but also the "difference between breakout affective DNA and lower-selling DNA". What this modish bilge boils down to is that bestsellers tend to have certain things in common, most of which you do not need a computer to detect.
In On Writing (2000), for example, Stephen King advises that one write from personal knowledge of life, friendship, love and work, especially the latter, and the bestseller-o-meter agrees.
King also thinks that it is much better to report speech with a simple "she said", rather than "she exclaimed indignantly", and so does the bestseller-o-meter. And so it goes on.
Christopher Booker's The Seven Basic Plots is translated into graphs, with the observation that it is advisable to have a clear three-act structure, with a regular rhythm.
Characters should do, look, tell, know and arrive, rather than wait, murmur, hesitate, halt and drop - though a worshipfully quoted passage from The Da Vinci Code includes this horror: "Sophie immediately plopped down on the stone floor..."
"Wouldn't it be fun," ask the authors, addressing the heart of their thesis, and missing, "if success weren't so random?"
No, it wouldn't.