10 January, 2014

How to Write a Bestseller - the Science!

I found this article in the Telegraph yesterday fascinating. Scientists in New York have used a technique called statistical stylometry to mathematically analyse the words and grammar used in popular novels to determine what makes them stand out.

Most analysis of what makes a bestseller focuses on the qualitative - themes and character-type for example. This study is purely statistical and it has an alarming accuracy. After using 800 books to create an algorithm they then used that algorithm to predict the popularity of other books. Prediction matched popularity 84% of the time.

I recommend you read the article (it's short!), but here are the common characteristics of successful and not-so-successful books.

Less successful: cliches, heavy uses of verbs and adverbs, descriptions of emotions and actions (e.g. 'wanted' or 'promised')

More successful: heavier use of nouns and adjectives, verbs more likely to describe thought processes (e.g. 'remembered' or 'recognised'), heavy use of conjunctions (and/but etc.)

On the whole this doesn't surprise me - describing actions and emotions is "showing not telling", and we must all know by now that heavy use of adverbs is bad form (and genuinely (adverb!) annoying to read). Although the study did take into account a wide range of genres - so this can be thought of as a list of general characteristics of good writing - I imagine there would be slight changes if it had focussed on one particular style. Thrillers, for example, I might expect to have more descriptions of actions.

The one thing that I wouldn't have guessed at is the use of conjunctions in successful books. Although we know we should vary sentence length for effect, in general it seems that longer sentences are preferable. I think I understand this. Short sentences for impact can be great, but can be tiring to read for more than a paragraph or two. Longer sentences makes for a smoother read, allowing the reader to become more engrossed.

I'm sure there will be some people who find it distasteful to analyse something creative in this way. What does success mean, anyway? Personally however, I find it fascinating at how accurate the predictions are from this analysis. It confirms something that we all know deep down. There will always be badly-written publishing sensations, and beautiful neglected masterpieces, but if we want to maximise our chances of writing something successful then the best thing we can do is work on those basics of writing style. Luck and contacts come into it, but first and foremost we've got to write good books. Nobody wants to live a beautifully-designed house built from shoddy materials!


  1. I hope this post - and the study - sparks some debate. I wonder how Dan Brown feels about critiques of his books? Does he, for example, just write what he wants to write, or does he believe that he has pushed himself to the limits of his writing ability and cannot develop any more stylistically. That's not a dig at him, as I've not read any of his books yet, just a question. I'm reminded of Stephen King supposedly referring to himself thus: "I'm the literary equivalent of a big mac and fries." When does style become formula? Answers on a postcard please!

    1. I love that there's room for all types of writers out there. I bet Dan Brown doesn't mind too much (although I'm sure still a bit) if his books aren't considered literary greats when he sees the sale figures! I've always said I want my books to be read and enjoyed, I don't care about whether they are masterpieces... but I wouldn't say no to writing something considered masterful!

  2. I'm surprised at the results because it seems to me that an interesting story badly written will tend to out-do a boring story brilliantly written.

    I think the mixing of different genres, not to mention the choice of specific books, might have thrown up misleading results.

    1. I guess that's the point though - this study takes out our personal concept of good/bad/interesting/boring and literally just looks at the building blocks in a mathematical analysis. And it's found that certain characteristics appear across all genres - which I find really interesting and encouraging: successful writing is successful writing whether you're trying to write a literary novel or comic caper. Probably if they chose a different set of books the results would be a bit different, but I'd be surprised with a sample size of 800 if it would be very different. However, if they limited the study to one genre only I'd love to see the results - especially with something like horror or sci-fi.

      Obviously there's tonnes a study of this nature could never say about something as subjective as books, but I think in the narrow scope of what it's tried to investigate, it's fascinating.