I started by using all the contestants from the current poll and putting them into a spreadsheet. Yes, there is a survivor bias here, but I was willing to accept that inaccuracy. But hey, this is LJ Idol, so the Survivor bias should be canceled out, because we're a spin-off of a different reality show. ;)
I took the numbers from everyone's mutual friends. If that data wasn't available, used just the generic friends or friend of category. (Insert joke about Friends here.)
I then took the number of votes everyone got in the last 5 polls.
Then what to do with all that data? Hmmm.
My first attempt at an equation was this:
(Total votes for that person that week - Average number of votes for all that week)/Friends list
I looked at those numbers and then decided that it put too much weight on folks who did better than the average number of votes. There would still be a large friends list bias in those numbers.
Then, I decided to change it to this:
(Total votes for that person that week - Personal Average over the 5 weeks)/Friends List
I liked that number a little better, but I felt that put too much emphasis on just how they did over their average, this is a contest after all.
So finally I came up with:
(Total votes for that person that week - (Personal Average over the 5 weeks +Average number of votes for all that week)/2 )/Friends List
That seemed a good balance of doing better than your average, but also incorporating doing better than the group.
Now, I'm sure if I did more research, I could get a better number for the F-list bias. But, this was just for my personal use.
But after all that, I decided I was too lazy to go back and actually read all those posts.
So instead, I give you the top 5 for the last five weeks, based on the numbers described above: