Game odds, win probabilities, power rankings

Discussion in 'Vols Football' started by Tar Volon, Aug 9, 2016.

  1. Tar Volon

    Tar Volon Me Blog @RockyTopTalk.com

    Last edited: Aug 9, 2016
  2. TennTradition

    TennTradition Super Moderator

    Awesome. I really like the idea of taking betting odds vs historical win percentage.
     
  3. bigpapavol

    bigpapavol Chieftain

    But the historical data is generally worthless.
     
  4. TennTradition

    TennTradition Super Moderator

    In a realm where otherwise you are just grabbing a percentage out of your rear, I like the idea of trying to use data to generate the percentage chance of each win, though.

    You're probably right but it isn't completely clear why that's the case. I'm guessing the biggest reasons are 1) opening odds vs closing odds and 2) odds are also set to move money not necessarily set the believed odds of a win.

    If the historical data is crap, then I'm guessing the regression coefficient would suck. Does it?
     
  5. bigpapavol

    bigpapavol Chieftain

    I suspect the regression is decent because of the repeated success of the bigger programs. Makes the anomaly at UT even more annoying.

    That said, it's a spot where I don't like the predictive ability of the data, even though it's probably fairly accurate. Causality problem.
     
  6. Tar Volon

    Tar Volon Me Blog @RockyTopTalk.com

    I look at it more of a way of trying to set expectations rather than predict what will happen
     
  7. IP

    IP Super Moderator

    And that makes sense. That is what is truly being captured both with odds and past events.
     
  8. kmf600

    kmf600 Energy vampire

    WTF does all of this mean? Are we winning the SEC championship or not?
     
  9. IP

    IP Super Moderator

    It means expectations are such that it wouldn't be absolutely shocking if we did.
     
  10. GahLee

    GahLee Director of Conspiracy Theories, 8th Maxim

    Were you listening to the Dude's story? So you have no frame of reference. You're like a child who wanders in to a theatre...
     
  11. kmf600

    kmf600 Energy vampire

    I was bowling.
     
  12. kmf600

    kmf600 Energy vampire

    Is this science?
     
  13. IP

    IP Super Moderator

    Social science. So no.


    (burn)
     
  14. kidbourbon

    kidbourbon Well-Known Member

    How so?
     
  15. kidbourbon

    kidbourbon Well-Known Member

    I wouldn't even go with the term "historical data". It's just about having a large enough sample size to make the correlation meaningful. With large enough sample size available, one can plot a pretty little graph with a line, and then one can assign an equation to said line, and then one can make a chart -- using said eequation -- that matches up winning percentage and point spread. With that in hand, compare it against future seasons. It'll match up. Methinks it not crappy.

    I understand that win probability and point spread aren't gonna skip together down the street hand in hand. Which is to say, point spreads don't dictate who wins the games. There's no causal connection.* But there is a correlation, and one that is described quite nicely by the aforesaid data. That's not quite the same as having your own personal Nostradamus, but it's useful information.


    *This is why I gamble.
     
    Last edited: Aug 10, 2016
  16. bigpapavol

    bigpapavol Chieftain

    The people playing all change. Our 2005 result against UGA has nothing at all to do with the upcoming 2016 game.
     
  17. TennTradition

    TennTradition Super Moderator

    I can think of other reasons why the regression might be questionable, but not this one.

    The regression correlated Vegas odds to win percentages of teams with those odds, over a number of years of data.

    Vegas obviously adjusts its odds based on personnel changes year to year.
     
  18. kmf600

    kmf600 Energy vampire

    Easy there Sheldon
     
  19. A-Smith

    A-Smith Chieftain

    die
     
  20. fl0at_

    fl0at_ Humorless, asinine, joyless pr*ck

    I thought one should very, very, very rarely use regression for prediction.

    I thought moving averages and such were better for extrapolating futures, and then, only one or two units in the future.
     

Share This Page