The Ranking Method

SCORING

Similar to last year, each voter starts out with a total of 55 points, with their favorite album getting 10, second 9, etc. Voters who didn’t rank 10 albums get both my disapproval and a proportional reduction in the value of their votes (10% less points for ranking 9 albums, 20% for 8, etc.).

Voters self-reported how many of the 59 best album candidates they listened to, and their scores were weighted by the following table.

Number of Albums Vote Weight
0 – 10 100%
11 – 20 125%
21 – 30 150%
31 – 40 175%
41 – 50 200%
51 – 58 225%
59 275%

SIMILARITY

The similar users section of the voters page is determined by a method that A) adds weight to any person who had an album also on your list , B) adds more weight if that person has the album in a similar position to you, and C) intensifies the relationship if you ranked that album highly. Because of that last stipulation, you might not be reciprocally similar to someone else. Given that the most similar users are spouses, girlfriends, and siblings, I feel it works pretty well.


=Sum(A*B + C) / Total Possible Points
=Sum( Person1Rank * (11-Abs(Person1Rank-Person2Rank)) + 5 ) / 600

The suggested albums calculation uses similar users to predict which albums you might like. I’d tell you the calculation, but A) I’m (still) holding out for the Netflix Million Dollar Prize and B) it’s terrible.