undyingking: (Default)
undyingking ([personal profile] undyingking) wrote2007-05-22 10:31 am
Entry tags:

Football prediction results

Back in August I invited people to predict which teams were going to finish as the top 5 in the Premiership. And they did!

This is what actually happened:
  1. Man Utd
  2. Chelsea
  3. Liverpool
  4. Arsenal
  5. Spurs

So, the top two changed places, but otherwise the same as last year. These were the predictions:

[livejournal.com profile] dr_bob
  1. Chelsea
  2. Arsenal
  3. Man Utd
  4. Spurs
  5. Liverpool

[livejournal.com profile] colonel_maxim
  1. Arsenal
  2. Liverpool
  3. Chelsea
  4. Man Utd
  5. Spurs

[livejournal.com profile] mrlloyd (as he now is)
  1. Chelsea
  2. Liverpool
  3. Arsenal
  4. Spurs
  5. Man Utd

[livejournal.com profile] undyingking
  1. Chelsea
  2. Liverpool
  3. Arsenal
  4. Man Utd
  5. Spurs

So everyone got the right top 5... well, OK, maybe that was the easy part. According to the by-now-familiar [livejournal.com profile] undyingking's Rank Correlation Coefficient, that gives scores of (small is good):

Well done again [livejournal.com profile] dr_bob, the only one not to wildly underestimate Sir Alex's merry men. But let's hope that next year things get shaken up a bit more...
chrisvenus: (Default)

[personal profile] chrisvenus 2007-05-22 11:49 am (UTC)(link)
And being a sad mathematician I've even reverse engineered your rank correlation. And I'm not covinced its really yours though my memory is too hazy to say for sure. :)
chrisvenus: (Default)

[personal profile] chrisvenus 2007-05-22 12:00 pm (UTC)(link)
Oh no, yours differs from the one I was thinking of by some constants.
I think I was looking back at the Spearman rank coefficient which produces a number in the range -1 to 1. for just working out which ranks best though yours is teh win. :)

[identity profile] undyingking.livejournal.com 2007-05-22 12:05 pm (UTC)(link)
I was thinking about using Spearman's and also looked at Kendall's, but I didn't quite like either of them. I expect a statistician would scream in horror at mine, but hopefully none is reading this LJ ;-)
chrisvenus: (Default)

[personal profile] chrisvenus 2007-05-22 12:09 pm (UTC)(link)
Spearman's is pretty much the same as yours. If you'd divided by four and subtracted it from 1 I think you'd have had spearman's. If you are doing no further analysis than you did then this is suitable since the ordering will be the same as spearman, you just won't be able to do as useful things as saying -1 is revere related (ie opposite order, etc.

Oh, actually I think you square rooted as well didn't you. Ah well, the ranking is definitely the same anyway. :)