Guess who's back?
My attempt to use Towerian Logic to predict the rating for the new Long Winters album was surprisingly effective. I was only 0.2 points off, bringing my cumulative point differential to 11.7 after nine albums reviewed (average of 1.3 per album in case you're keeping track). My confidence is nearing an all-time peak! I will hold my head high today, and possibly part of tomorrow as well.
posted by Trellick Tower at 8:17 AM
4 Comments:
Great idea, but ...
Methinks that a real statistician should weigh in on this. I'd say that many -- if not most -- Pitchfork reviews probably fall within a 3-point range (let's say 6.5 to 9.5). So is an average error of 1.3 points per album really all that good? As long as your prediction is within that point range (6.5 to 9.5), you're probably gonna be within a point or two of what Pitchfork actually gives a disc.
I don't have the skills to figure this out scientifically, but I'd say the anecdotal evidence is strong.
While I truly appreciate your interest, it should be noted that all types of in-depth scientific analysis are strictly forbidden at That's Mr. Tower.
That said, the point of That's Mr. Tower is not to attain a good average prediction rate. The point (if there is one) is to try to closely predict each album individually, in the hopes of one day getting one right.
And also of course have a bit of fun along the way, while making use of some of the otherwise useless musical knowledge that has taken over my brain.
Word. I suggest this, then: Celebrate those individual successes. Instead of keeping cumulative point differential stats and broadcasting them to the world, you should keep a Top 10 of your closest guesses. Put it on the right rail, where it can always be seen. I'd check it once in awhile, for sure.
Just tryin' to help.
Thanks jw, I appreciate it! I am planning to put a Top 5 or 10 list on the side bar, I just haven't had time to figure out how to do it.
Post a Comment
<< Home