26 November 2009

A bayesian cinefeel…(*)

by manos_parzakonis

How imbd ranks movies?

The formula for calculating the Top Rated 250 Titles gives a true Bayesian estimate:

WR=frac{nu }{nu +m}R+frac{m}{nu +m}C

R = average for the movie (mean) = (Rating)

nu = number of votes for the movie = (votes)

m = minimum votes required to be listed in the Top 250 (currently 1500)

C = the mean vote across the whole report (currently 6.9)

Now that’s something unexpected! Going further than the simple arithmetic mean is something exciting, right?

The formula is the well-known decomposition of the posterior mean -distributed a priori as normal- compromising the prior guess (C) and the data (R) weighted by the sample (nu) and the pretend-to-be initial sample (m) under the conjugate prior setting.

tags: bayesian - conjugate - movies - prior