Originally posted by JeffLowell
View Post
It would be very impressive for readers to be within two points of each other if scores were evenly distributed from the 1-10 scale, with roughly 10% receiving each number, but they're not. Reader scores have a bell-curve distribution. If we look over the 2013 stats, we find this distribution:
82 472 1022 1781 2258 2011 1288 460 83 2
That means 0.02% of scripts got a ten and 23.9% -- almost one in four -- got a five.
Here's yet another way to look at it: 88.4% of scripts scored in the three to seven range. That means if you programmed a computer to arbitrarily assign every script a five, that computer would be within two points of the human reader 88.4% of the time. It wouldn't have quite the same internal reliability as the average Black List reader, but it would be pretty close.
I still believe The Black List is a great service, and as far as I can tell it has the best possible execution for this particular idea. But I have yet to see evidence that reader ratings deserve the level of esteem some people here have given them.
Leave a comment: