home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky rec.music.misc:30927 alt.rock-n-roll:20115 alt.music.progressive:5074
- Newsgroups: rec.music.misc,alt.rock-n-roll,alt.music.progressive,alt.music.symphonic
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!ira.uka.de!math.fu-berlin.de!mailgzrz.TU-Berlin.DE!news.netmbx.de!Germany.EU.net!mcsun!sunic!seunet!enea!sommar
- From: sommar@enea.se (Erland Sommarskog)
- Subject: UAP #28 - Peter Gabriel, reminder
- Message-ID: <1993Jan24.211135.5895@enea.se>
- Followup-To: poster
- Keywords: Usenet Artists Polls
- Organization: Enea Data AB
- Date: Sun, 24 Jan 1993 21:11:35 GMT
- Lines: 137
-
- As I've announced in a seperate article, Peter Gabriel will be the
- third-to-last Usenet Artist Poll, followed by Genesis and Roxy Music.
-
- Deadline is January 30th.
-
- Below is a discography of Peter Garbriel's solo albums. Please feel
- free to enter any item I've forgot to include, but as always please
- no bootlegs, videos and interview albums.
-
- Please, use the discography as a template for your ballot. Particularly
- observe how I'm referring to his four first unnamed albums. (The fourth
- album was called "Security" only in the US.) I've included one song from
- each as a comment if you don't know which is which. I've also included
- the German versions of the 3rd and 4th albums, although they are very close
- to the original albums.
-
-
- &BEGIN
- Peter Gabriel <Solsbury Hill>
- Peter Gabriel (2) <On The Air>
- Peter Gabriel (3) <Intruder>
- Ein Deutsches Album (3)
- Peter Gabriel (4) <Shock the Monkey>
- Ein Deutsches Album (4)
- Peter Gabriel Plays Live
- Birdy
- So
- Passion
- Shaking the Tree
- Us
- &END
-
- If you are new to these polls, please read the instructions below.
- If you interested in reading about the algorithm I use to compute
- the results, there is a descpription following the instructions.
-
- How to vote
- -----------
-
- Use the list above as a template for your ballot. Sort the list
- so that your favourite comes right after the &BEGIN and your
- least favourite just before the &END. Delete albums you are not
- familliar with. One album by line. No numbering. Please delete
- quote characters like ">" from your ballot.
-
- If you can't tell two or more albums apart you can declare a
- tie. You do this by preceding the albums with &TIE. The tie
- lasts until the next &END. The number of &END on the ballot
- must be equal to the number of &TIE + one for the &BEGIN.
-
- If you feel that there is a big gap between two albums you
- can mark this with a filler. A filler has the format &FILLERn
- where n is a positive number, default is 1. Notice that a very
- big value of n make little sense; the net effect is the same
- declaring an &TIE on the albums separated by the &FILLER.
-
- To participate you need to know at least two albums, so you don't
- have to be shy. Ballots are weighted, so the more you albums you
- know, the more influence do your ballot get.
-
- For further details on the algoritm, see below if you are interested.
-
- *MAIL* your vote ballot to me, sommar@enea.se. I try to acknowledge
- all votes, so if you do not receive an ack after a few days, please
- mail me again.
-
- If your mailer leave a decent return address *and* your ballot mail
- has a non-empty subject line, you will get the results in your mailbox.
- If you want the results to another address, please state so, but no
- guarantees.
-
- You are welcome to send in a vote for a friend who does not have net
- access. Just indicate that the vote is for someone else so I know that
- it's not a duplicate.
-
- Have fun!
-
- The Algorithm
- -------------
-
- This an example ballot for an imagninary group: (The numbers in the
- right are NOT part of the ballot as you write it, but are used for
- the description below.)
-
- &BEGIN
- She Comes in the Morning 0.5
- Next Day 0.4
- &TIE
- Looks Like Trouble 0.25
- Futuralized 0.25
- &END
- On the Way Down 0.1
- Cram it! 0.0
- &FILLER4
- Last Really Crappy Album -0.500
- &END
-
- The algorithm to compile the results of a polls consists of a
- number of steps.
-
- Step 1:
- For each individual ballot, the intervall from -0.5 to 0.5 is sliced
- up at (N + F + 1) equidistant points, where N = number of entries,
- and F = number of fillers. In the example ballot N = 6 and F = 4.
- The top album gets 0.5 points, next album ((N+F) / (N+F+1)) - 0.5
- points, third best gets ((N+F-1) / (N+F+1)) - 0.5 points etc, down
- the last which gets -0.5 points. If two or more albums are tied,
- they get the average of the involved points. The net result for
- the example ballot is illustrated by the number on the right side.
- (Side note: if we forget the fillers, the method corresponds to the
- one you use for pairs competitions in bridge.)
-
- Step 2:
- In order to avoid distortion from voters who only know the 2-3
- most popular albums, the ballot is weighed so that the points
- from step 1 are multiplied with sqrt(N).
-
- Step 3:
- All ballots are added together and for each album we get a number
- of points P = ((p1 + p2 ... pM) * 1000 / M) * C + 500, where M =
- the number of ballots the album appears on, and C is a poll constant.
- Thus, the result for an album is indepentent on how well-known it is,
- However there is a cut-off rule, so albums which only appears on one
- or two ballots are excluded from the final result.
-
- The Poll Constant:
- This constant is sqrt(B/T) where B where B is the number of ballots
- in the poll and T is the total number of entries on these ballots.
- The idea with the poll constant is to remove the effect of the weighing
- factor from the range of the points, so we can compare polls with
- each other. If an album tops all ballots, it will get 1000 and 0 if
- it is ranked last by all voters. Note however that this compensation
- is a rough model. If an album tops five ballots of ten, and is not
- mentioned on the other five, its result can be bother higher and
- lower than 1000 depending on the weight of these ballots.
- --
- Erland Sommarskog - ENEA Data, Stockholm - sommar@enea.se
-