kind of nifty

http://lsufootball.net/articles/2009-preseason-polls.htm

5/11/2009 8:28:25 PM

But mostly pointless

5/11/2009 8:31:48 PM

It was until I legitimized it by making a thread.

5/11/2009 8:33:12 PM

Huh, they ran our numbers wrong.

Our average is 25.53, not 26.
And average appears to be the actual rank-determining statistic.


...and upon closer inspection, they ran most of the numbers wrong if the Average stat is supposed to be across every one of the ranks (Sum divided by 13). It's like they rounded on some of them, didn't round on others, and rounded to some arbitrary decimal number on others (see ECU's).

v Oh... well, that's... interesting, I guess.

[Edited on May 11, 2009 at 8:38 PM. Reason : .]

5/11/2009 8:34:58 PM

if a team wasnt mentioned in a poll they got assigned a 26

5/11/2009 8:36:45 PM

Well that's retarded, why not just rank them in order of "others receiving votes"?

Or, better yet, calculate their share based on vote percentage.

Dumb

5/11/2009 8:39:13 PM

because no one gives a fuck outside of the top 5-10 at this point...

5/11/2009 8:40:46 PM

I guess that makes sense considering their being voted in the top 5 four months before the season starts assures those will be the only teams with a chance to play for the national title.

5/11/2009 8:42:37 PM

It wouldn't be that bad a system at all if it weren't for

Quote :

"To get the average, the highest and lowest rankings are thrown out."

What the fuck is the statistical reasoning behind that? I mean, yeah, if they're outliers, but just tossing out the min/max seems stupid.

5/11/2009 8:42:58 PM

That's pretty standard practice to normalize the data. There are lots of other reasons why this is dumb, though.

5/11/2009 8:44:54 PM

^^I forget what the name of the technique is called, but it isnt to often used. They usually take off the bottom and top 10% of your data when using that technique.

5/12/2009 12:32:13 PM

.

[Edited on May 12, 2009 at 12:32 PM. Reason : double post]

5/12/2009 12:32:13 PM