I recently came across the Internet Book List, a community driven book database which aspires to be as comprehensive for books as IMDB is for movies.
Picking a book at random, I did not find Dickey’s Deliverance in the database. I was hoping for a useful research tool for my 20th Century handicapping project, but it is not quite there yet.

One thing that makes me sad about the IBList is the book rating system. Once you register, you can rate books on a scale of 1 to 10 (a type of range voting). And, like IMDB, the ratings are weighted to compensate for some books being rated more often than others.
If you think about how your rating will affect the rank of a book, and if you want to maximize the influence of your opinions, you will quickly conclude that the logical thing to do is to give any book you like a score of 10, and any book you dislike a score of 1. Your opinion will be mixed in with that of dozens — or hundreds — of other voters, and making a nuanced choice between “8″ or “9″ is just a waste of time; a difference of 1 point lumped with 10 other voters has a net effect of .1, and with 100 has a net effect of .01.
When you rate only 1 or 10, this becomes a form of approval voting. A nice feature of approval voting is that the distribution of votes is a binomial distribution (although range voting looks like a bell curve, it is not binomial), which makes it extremely amenable to statistical methods. Standard deviations can be calculated with a simple formula, and these deviations can be used to compare entries with different sample sizes, without having to rely on an artificial weighting mechanism.
But the IBList people have chosen range voting, because that is the popular rating scheme that everyone else uses on the web. Everything ends up with a 7, or, on Amazon, every book gets 3.5 stars. These ratings provide no information.

Filed under musings · Tagged with

Comments are closed.