oops...wrong use of Poisson calculator?

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • Bean
    SBR Rookie
    • 10-26-10
    • 16

    #1
    oops...wrong use of Poisson calculator?
    Bodog had an interesting prop bet this weekend:

    2011 NFL Playoffs - How Many Wild Card Teams will Win in Round 1 of the Playoffs? (Jets, Ravens, Saints, Packers)
    0 +2500
    1 +400
    2 +140
    3 +175
    4 +700

    Using Bodog's Saturday morning spreads for the four games, I came up with these winning probabilities for each of the four WC teams: Saints 84%, Jets 42.6%, Ravens 63.23%, Packers 44.4%. So, the predicted number of WC winners should have been: .84 + .426 + .632 + .444 = 2.342

    Plugged this into the Poisson calculator, which spit out:
    0 winner +940 9.6%
    1 winner +361 22.51%
    2 winner +279
    3 winner +386
    4 winner +728
    Which means that "0" was a great value at +2500 and "1" a good value at +400. So naturally, I took these two bets (and "0" is already dead, thank you Antonio Cromartie and Nick Folk!). It just occurred to me that this logic might be completely flawed, because the 4 events are not identical like interceptions, rushing yards, or pass completions might be. The right way to think about it might be as a sum of 4-way parleys, in which case "0" and "1" were probably horrible bets (haven't done the math). Or maybe the two approaches amount to the same thing???
  • durito
    SBR Posting Legend
    • 07-03-06
    • 13173

    #2
    Yes.

    Take the no vig ml at pinny and then calculate:

    odds of 0 winners = P Colts Win * P Seattle Win * P KC Win * P Phi Win
    odds of 4 winners = P Jets W * P Ravens W * P Saints W * P GB W

    odds of 1,2,3 winner have multiple possibilities compute the probabilities and add them
    Comment
    • Justin7
      SBR Hall of Famer
      • 07-31-06
      • 8577

      #3
      Poisson doesn't work very well when you have this few trials. If you have only 4, I would use a Binomial (or do each outcome manually).

      Consider this problem: You have one wild-card team that is 67% likely to win. What answer does Poison give you for 1 wild-card win, with one trial at 0.67? 0.488, when it should obviously be 0.67.

      A general rule I use on Poissons... You normally want at at least 20 trials with odds of 0.1 or less per trial. Or, N/p >200. So if you have lots of trials, it is ok for higher p's.
      Last edited by Justin7; 01-09-11, 12:17 PM.
      Comment
      • Bean
        SBR Rookie
        • 10-26-10
        • 16

        #4
        Interesting, thanks Justin! You da man!
        Comment
        • Peregrine Stoop
          SBR Wise Guy
          • 10-23-09
          • 869

          #5
          giving the poisson calculator to these fools that don't know what the distribution is or how to use it is like giving the crack pipe to Whitney Houston
          Comment
          • Bean
            SBR Rookie
            • 10-26-10
            • 16

            #6
            Thanks for the complement Peregrine Stoop! I'm just an average Joe, win some/lose some, trying to learn something every day. And look on the bright side: if it wasn't for fools like me to take the wrong side of bets, how would the elites like yourself get rich???

            Namaste!
            Comment
            • jackpot269
              SBR Posting Legend
              • 09-24-07
              • 12842

              #7
              wow wish i had seen this when looking at bodog yesterday
              Comment
              • jackpot269
                SBR Posting Legend
                • 09-24-07
                • 12842

                #8
                Originally posted by Bean
                Thanks for the complement Peregrine Stoop! I'm just an average Joe, win some/lose some, trying to learn something every day. And look on the bright side: if it wasn't for fools like me to take the wrong side of bets, how would the elites like yourself get rich???

                Namaste!
                Hey , I'm not the only one, after all for years I thought everyone on here was a great betting mind except for myself glad to know there are others out there!!!!

                Now you know I meant in each ones opinion of there self !!!!!!!!!!!!
                Comment
                • subs
                  SBR MVP
                  • 04-30-10
                  • 1412

                  #9
                  stattrek maybe can help with the binomial distribution...

                  namaste baba
                  Comment
                  • That Foreign Guy
                    SBR Sharp
                    • 07-18-10
                    • 432

                    #10
                    When accurate odds are available for each constituent event, attempting to approximate it by (mis)using a generalised distribution is foolish.

                    Durito's method is optimal IMO.
                    Comment
                    • uva3021
                      SBR Wise Guy
                      • 03-01-07
                      • 537

                      #11
                      Poisson distributions are reserved for events that are mutually exclusive, extremely rare, and more or less randomly associated with its cause. Such as the weather, the mutation rate of genes, or the gait of a drunkard. Football outcomes scarcely qualify as following a Poisson Distribution.

                      In this case, either calculate the odds manually or use BINOMDIST in excel: [insert "=BINOMDIST(ROW()-1,4,2.34/4,FALSE)" in cell A1 and copy to A5]

                      0 - 2.97% 3271
                      1 - 16.72% 498
                      2 - 35.36% 183
                      3 - 33.23% 201
                      4 - 11.71% 754
                      Comment
                      • simplydusty
                        SBR High Roller
                        • 12-18-08
                        • 229

                        #12
                        Originally posted by Justin7

                        A general rule I use on Poissons... You normally want at at least 20 trials with odds of 0.1 or less per trial. Or, N/p >200. So if you have lots of trials, it is ok for higher p's.
                        Just finished reading this in the props chapter of Conquering Risk. Great read
                        Comment
                        SBR Contests
                        Collapse
                        Top-Rated US Sportsbooks
                        Collapse
                        Working...