Comparison based formula for best reviewed albums of year: April
This is a formula I came up with to factor out the skewed distribution of scores albums get, especially near the top. The idea was to even out the power of every vote, and here are the results so far.
I thought I'd either get this out way before or way after mega-critic. I was only waiting for Portishead to get reviews.
Does anybody know if there's any escape characters or something I can use to preserve the white space? No matter how many spaces there are it seems to always reduce it to one. It makes it very hard to make the chart easily readable.
Band - Album . Wins - Losses - Ties . %comparisons won . reviews/total
1. Nick Cave - Dig Lazarus, Dig! . 45-1-1 . 70% . 27/36
2. Portishead - Third . 37-2-0 . 73% . 19/36
3. Vampire Weekend - Vampire Weekend . 39-6-2 . 61% . 30/36
4. Bon Iver - For Emma, Forever Ago . 28-1-2 . 68% . 16/36
5. British Sea Power - Do You Like Rock Music . 37-8-1 . 56% . 27/36
6. Hercules and Love Affair - Hercules and Love Affair . 26-1-1 . 74% . 17/36
7. Drive By Truckers - Brighter Than Creation's Dark . 34-7-1 . 58% . 23/36
8. Sun Kil Moon - April . 22-4-1 . 60% . 14/36
9. Elbow - The Seldom Seen Kid . 26-6-0 . 57% . 17/36
10. Los Campesinos! - Hold On Now, Youngster . 30-7-1 . 59% . 21/36
11. Erykah Badu - New Amerykah: Part One(4th World War) . 19-3-1 . 63$ . 18/36
12. Black Mountain - In The Future . 26-15-5 . 52% . 30/36
13. Why? - Alopecia . 17-8-4 . 54% . 16/36
14. Destroyer - Trouble In Dreams . 22-14-6 . 53% . 24/36
15. The Gutter Twins - Saturnalia . 23-15-6 . 50% . 24/36
16. American Music Club - The Golden Age . 20-11-7 . 53% . 17/36
17. Dodos - The Visiter . 11-1-2 . 59% . 11/36
18. Hot Chip - Made In The Dark . 21-21-4 . 48% . 30/36
19. Sons and Daughters - This Gift . 16-17-4 . 46% . 20/36
20. Fuck Buttons - Street Horsssing . 12-11-8 . 51% . 16/36
Re: Comparison based formula for best reviewed albums of year: April
Here's how I sort this list again.
First, for a win or a loss to count, there has to be at least ten publications that reviewed it in common.
After that, what I do is take the one with the lowest score out, recalculate the scores, take the one with the lowest score out, etc, until I'm down to 1.
Using this formula:
Win over album still in: 1 point
Tie with album still in: 0.5 points
Win over album already out: Between 0 and 0.5 points depending on how far down it is.
This way ones that are more reviewed get a bonus, and ones that have wins against the higher ranked albums get a bonus.
Re: Comparison based formula for best reviewed albums of year: April
Very interesting concept, Bill.
A question to understand the recalculating of scores correctly: With 'the lowest score' you refer to single review scores and with 'recalculating the scores' you mean the average of review scores per album, right? Or do you mean something else?
Re: Comparison based formula for best reviewed albums of year: April
No, it's not based on an average. That's the idea, I figure average-based calculations are already covered by metacritic and mega-critic.
I wanted to find a formula that wasn't based on the average, because I thought the negative skew of the review distributions was throwing off the results, giving single extreme negative reviews too much influence.
Here's the process. You look at every pair of albums by themselves, and for every publication that reviewed them both, just look at which gave it a higher score.
So, if it was...
Pitchfork: Album A 88, Album B 84
Popmatters: Album A 70, Album B 90
Blender: Album A 60, Album B 50
Album A would win over Album B 2-1. Every publication gets equal influence no matter how far from the mean their score is.
So you compile a matrix of wins, losses, and ties, and then you calculate the scores as previously described, going up the list one position at a time, taking out the lowest each iteration.