Announcement

Collapse
No announcement yet.

Further Proof that Sagarin is Idiotic

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by yourfavestoner View Post
    Well, I have no doubt that Boise would beat the piss out of Florida this year. The Gators are absolutely awful in every facet of the game. Wayyy too much talent lost over the last two years.

    I've never understood why cfb fans think you can just replace talent like that and keep on churning. Those Bama and Florida NC teams could stand toe to toe with just about any NC team ever. They were that good.
    I do think Florida would beat BSU this year (see, that's easy).

    Comment


    • #32
      Everything is going to have a bias, no matter what. Period. This includes any algorithm or any human opinion even if hundreds of games were played.

      The real question is whether there is a motive behind these biases (whether or not they're trying to use this bias to push an agenda), and I think the only culprit here is the BCS. The biggest problem I think is that only one of the methods is open (Colley), so there is no way the algorithms can be publicly analyzed.

      There are a lot of computer methods that don't require human polls or previous results, some in the BCS do for some god unknown reason. Sagarin's are phased out as the season goes on until they aren't used anymore at all (the results only actually matter at the end of the season anyways). Billingsley's ratings by design were meant to emulate human polls (total crap).

      I think you're making the computer ratings out to be a lot more evil than they actually are. The vast majority of them give pretty good ratings based on how well they played against their schedule, and that's all that any computer do. Isn't that what ranks are supposed to be? Humans just provide speculation, emotion, and homerism.

      Also, I don't think most of the ratings borrow ELO, see: Colley matrix. Also, the BCS can use more than 'just a handful of geeks', http://masseyratings.com/cf/compare.htm

      Comment


      • #33
        Originally posted by bustabinary View Post
        Everything is going to have a bias, no matter what. Period. This includes any algorithm or any human opinion even if hundreds of games were played.

        The real question is whether there is a motive behind these biases (whether or not they're trying to use this bias to push an agenda), and I think the only culprit here is the BCS. The biggest problem I think is that only one of the methods is open (Colley), so there is no way the algorithms can be publicly analyzed.

        There are a lot of computer methods that don't require human polls or previous results, some in the BCS do for some god unknown reason. Sagarin's are phased out as the season goes on until they aren't used anymore at all (the results only actually matter at the end of the season anyways). Billingsley's ratings by design were meant to emulate human polls (total crap).

        I think you're making the computer ratings out to be a lot more evil than they actually are. The vast majority of them give pretty good ratings based on how well they played against their schedule, and that's all that any computer do. Isn't that what ranks are supposed to be? Humans just provide speculation, emotion, and homerism.

        Also, I don't think most of the ratings borrow ELO, see: Colley matrix. Also, the BCS can use more than 'just a handful of geeks', http://masseyratings.com/cf/compare.htm
        This is great discussion. A fellow nerd!

        Sagarin says his rankings have the data disappear after several games (think it's 7, but I'm not sure). Regardless, I suspect (but can't prove) that it's hogwash. There isn't enough info for the preseason rankings to dissipate. If they wiped out, then they shouldn't be used in the first place and every team should be equal on day one.

        I see nothing inherently wrong in computer algorithms. I just want diversity in opinion. I want hundreds of different algorithms. I also want some human voting influence as well in the hundreds. Right now they only use 6 algorithms. The BCS is biased way too much in the favor of 6 human beings or small math teams.

        I don't think any of them have ulterior motives, but even if you look at complex network theory, I just don't believe that there is enough data to come up with a decent ranking system via formula. The season is too short and the teams are too many.

        Colley's rankings are interesting because they are bias free. They are as "honest" as you could get.

        Here is a nice breakdown of the rankings by a BSU blog post.

        So in summary, diversity in opinion trumps all. Give me a few hundred more computer algorithms to balance this stuff out.

        Comment


        • #34
          That's a good article.

          I agree about the lack of diversity, and it's even worse with the BCS algorithms because 5 of them are black boxes. There are plenty of different methods out there that can be used and people can understand, but the BCS chooses to use the ones that nobody knows the details of and only 6.

          I think the only reason Sagarin starts out using the previous season's results is so that he can publish something early that isn't completely whacked out (I mean, after one week, you could say Troy and Alabama are tied for #1 and nobody wants to read that) -- it's his job to publish these for USA Today. The actual results don't matter until all the games are played anyways. But like you say, we don't actually know this, it's just what Sagarin says.

          I think a 12 game schedule is enough data to make a decent ranking for CFB. One game can make all the difference between a good and bad rank, plus it gives an incentive to schedule good teams. Convergence is no fun, I like my insanity.

          Comment


          • #35
            Originally posted by HindSight View Post
            The fact that a team can rise or fall based on their opponents winning or losing is beyond stupid.
            TCU completely agrees. You shouldn't look at opponents at all.
            **** her in da *****!

            Comment


            • #36
              Originally posted by descendency View Post
              TCU completely agrees. You shouldn't look at opponents at all.
              Nice reading comprehension.

              Comment


              • #37
                Originally posted by HindSight View Post
                Nice reading comprehension.
                Nice trying to have your cake and eat it to.

                The reason why TCU's SOS is so poor is they play teams that lose a lot traditionally.

                So, playing teams that lose weakens your SOS. What's so hard to understand about that?

                And he (OSU's president) said that weaker SOS's don't belong in the National Championship. So, that's exactly why this metric did. It dropped OSU for their SOS getting weaker.
                **** her in da *****!

                Comment

                Working...
                X

                Debug Information