Monday, February 10, 2014

The Algorithm for a Bestseller

The biggest problem any publisher faces is how to tell if a book is going to be a success or not. Every so often, after a book has proved to be a huge best-seller, they have an  'Ah-ha!  So that's what the public wants!' moment and think that at last they've found the winning formula and commission lots of books on that basis. But sooner or later – sooner in most cases – the unobliging public make it clear that the previous success was a fluke and its formula-derived successors crash and burn.

Of course, there are the formula series like Mills and Boon that do incredibly well by supplying books to an audience that sees them as simply a brand that must be the same every time and would be as unhappy with something different as I would be if my favorite Fruit 'n' Fibre breakfast cereal turned out to have chocolate and nuts in it one morning.

But in less prescriptive fiction, the 'How do you tell?' dilemma for a publisher remains. Mostly, the decision to accept or not to accept a book comes down to a feeling about it, a hunch – or what editors like to call a 'judgement.' This could be the sort of 'judgement' that turned Frederick Forsyth and JK Rowling away from their doors.

However, a recent paper in Comparative Linguistics from the Computer Science department at Stony Brook University, New York, has come up with an algorithm that claims to be able to predict a successful book with 84% accuracy.

It was based on books available from the Gutenberg Press whose degree of success was already established, and it came to some surprising conclusions, some of which very much run contrary to accepted  beliefs.

For instance, advice given to crime writers suggests that we should be using verbs, keeping the action going, avoiding detailed description. We should be writing clear, punchy sentences, not stringing them out with conjunctions. Right?

The trends the study found in successful books were to have lots of 'and's and 'but's. along with large numbers of nouns and adjectives.

 Less successful writing relied more on adverbs and verbs, particularly verbs describing specific actions or emotions – they instanced 'wanted', 'took' or 'promised.' On the other hand, verbs describing thought processes like 'recognsied' or 'remembered' did feature in the books that had sold well.

Mercifully, I don't think there's any evidence that editors have started using them since I can't think the way I write would fit into the 'successful' formula, or that I could train myself to write that way. It would be a blow to think I could compete in only 16% of the market.

The books I like to read wouldn't come into that category either. And whatever the scientists say, I can't convince myself it would actually work in practice. One of the elements they do admit has an effect in practice is luck - and I certainly do believe in that! 


Anonymous said...

Luck of the draw, then, and what appeals to readers at a given point in time? Would that we had a formula, eh?

Aline Templeton said...

It would be easy, certainly - but aren't all us authors gluttons for punishment?!