Before I get into my 2012 Favorites List for real, there are some preliminaries I need to deal with.
First, I want to give a shout-out to a few of the records that I had hoped would make the top 100, but which to my surprise just didn't make the cut.
In no order:
- Replacements - Let It be
- Bob Dylan - Highway 61 Revisited
- Alison Krauss & Union Station - Every Time You Say Goodbye
- Buena Vista Social Club - Buena Vista Social Club
- Digable Planets - Reachin
- The Sundays - Reading, Writing and Arithmetic
- David Bowie - Hunky Dory
- Sam Prekop - Who's Your New Professor
- Rogue Wave - Out Of The Shadow
- Sunny Day Real Estate - Diary
All fine records. Definitely in my Favorites list. But the Top 100 was just out of reach for them, this time around.
And now the boring technical stuff...
The past two Lists used a simple C++ program to perform a sort of my input list, using me as the comparison function. It worked OK - I got lists out of it. But it wasn't ideal. I needed something tolerant of subjectivity - because I knew that record A was better than record B most of the time, but not all of the time - and that scheme just wasn't designed for logical inconsistency. And I wanted something that could do the comparisons between pairs of records independently from the ranking itself; with the previous method, I could eventually figure out the list as it was being built, and that made the whole exercise feel a lot like just doing it by-hand in Notepad: tedious.
And then I saw "The Social Network". And that clued me into the Elo ranking system, which is used in chess and Scrabble and video game rankings. It's what Zuckerberg used for his 'Facemash' girl-ranking app in that movie. The match-ups are random, and can be done separately from the actual scoring; you can do all the voting up front, then run the stats later. So, you don't need to see the list being built as you go. And it's very tolerant of inconsistency - if A beats B six times, and B beats A six times, it just means A and B are closely ranked - it doesn't mean the final result is logically meaningless. And so I wrote a web app similar to Facemash, to use for the 2011 Cleek Readers Record Poll. And it worked. And that app is what I used this for this List. And it's also what we'll use when we do the 2012 Cleek Readers Song Poll, later this year.
So... did it make a difference? Stay tuned, to find out!
136 records were in contention for 100 spots. 4,000 votes over seven weeks.