NFL - Overs & Unders ...and more!

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • CFMiles
    SBR Rookie
    • 09-03-10
    • 4

    #1
    NFL - Overs & Unders ...and more!
    This is my first post -- so play nice.

    Using data over the last 5 years, I've found trends for betting the OU on certain games that seem to come out pretty well. I'm really just using thresholds, so it's not even all that complicated.

    Code:
    SEASON    WIN    LOSS    PUSH    PCT
    2005      43       34       0    55.8
    2006      33       22       4    60
    2007      38       23       2    62.3
    2008      30       24       2    55.6
    2009      34       16       1    68
    And broken down by bet...

    Overall, over the last 5 years we show:
    Code:
    Overs           75-56-4     57.25%
    Unders         103-63-5     62.05%
    Totals         178-119-9    59.93%
    Now, here is where I wonder if I am doing some statistical nit-picking.

    Betting suggested overs for the last 5 years in weeks 1 and 2 have a 4-11 record. So I was thinking of just not betting them. Of course, I would like to logically support this. It is easy to say that offenses aren't clicking early in the season and, thus, you can't bet the suggested overs with the expectation that offense will be putting points on the board like they will during other weeks.

    On another note -- not trying to mix leagues, this post is mostly centered around NFL -- I've done similar analysis on college football games and actually come up with an overall of 70%. I'm not too sold on it as it is only 87 plays (4 pushes) over 5 years. And the nit-picking gets a little... crazy. Like "only ever bet a suggested over on a non-conference game after a bye week" (11-1 from 2005-2009)... all other "over" scenarios were either losers or too close for comfort.

    And on yet a last note -- I have a pretty extensive database at my hands with odds data (unfortunately) only going back to 2005. I would be willing to run the numbers to test any specific trends if someone thinks they might have something. I was recently asked to check how often the Gamecocks cover under Spurrier because someone heard that he pays attention to the line for the sake of the football boosters. He's 31/55 over the last 5 years... so just bet on Steve Spurrier and you'll beat the juice! ;-)

    I look forward to some thoughts and honest criticism.

    Cheers:
    Miles
  • Indecent
    SBR Wise Guy
    • 09-08-09
    • 758

    #2
    Originally posted by CFMiles
    Now, here is where I wonder if I am doing some statistical nit-picking.
    Originally posted by CFMiles
    On another note -- not trying to mix leagues, this post is mostly centered around NFL -- I've done similar analysis on college football games and actually come up with an overall of 70%. I'm not too sold on it as it is only 87 plays (4 pushes) over 5 years. And the nit-picking gets a little... crazy. Like "only ever bet a suggested over on a non-conference game after a bye week" (11-1 from 2005-2009)... all other "over" scenarios were either losers or too close for comfort.
    I think your concerns are valid. You'll need to share a little more about your process to actually know, but from what you are describing it does not sound like you are finding trends that will be useful going forward.
    Comment
    • jetsjets1028
      SBR MVP
      • 02-10-10
      • 1234

      #3
      go jets
      Comment
      • CFMiles
        SBR Rookie
        • 09-03-10
        • 4

        #4
        Originally posted by Indecent
        I think your concerns are valid. You'll need to share a little more about your process to actually know, but from what you are describing it does not sound like you are finding trends that will be useful going forward.
        Agreed. I think the college football numbers are much less significant than the NFL numbers.

        I think the NFL numbers actually have some clout as there isn't really any nit-picking going on, things are pretty much just "bet it" across the board.

        I've simply analyzed the over/under numbers over the last 5 years and see that better overs on totals of 35.5 or less and betting unders on totals of 47 or more, show a decent win rate.

        Showing 57% and 62% over 135 bets and 175 bets respectively has got to be worth something... no? What sample sizes should we be shooting for to establish confidence in a system? I think I need to try finding some odds data that goes back further than 2005.

        Thanks for the input.
        Comment
        • Indecent
          SBR Wise Guy
          • 09-08-09
          • 758

          #5
          Originally posted by CFMiles
          Showing 57% and 62% over 135 bets and 175 bets respectively has got to be worth something... no? What sample sizes should we be shooting for to establish confidence in a system? I think I need to try finding some odds data that goes back further than 2005.
          It depends.

          If you developed your hypothesis on one year (say 2005), your results from 2006+ could be used to evaluate what it will do moving forward.

          If you used all of the years to when developing/testing your hypothesis (2005-2009), you need data for more seasons to be able to conclude anything. Anything your model hasn't seen (games from 2004 or before or games from this year that haven't been played yet) will be valid to use for this purpose.

          If you haven't read Wong's book check it out, Chapter 7 goes into detail on how to evaluate W/L records.
          Comment
          • LtDementia
            SBR High Roller
            • 08-22-10
            • 203

            #6
            Originally posted by CFMiles
            Agreed. I think the college football numbers are much less significant than the NFL numbers.

            I think the NFL numbers actually have some clout as there isn't really any nit-picking going on, things are pretty much just "bet it" across the board.

            I've simply analyzed the over/under numbers over the last 5 years and see that better overs on totals of 35.5 or less and betting unders on totals of 47 or more, show a decent win rate.

            Showing 57% and 62% over 135 bets and 175 bets respectively has got to be worth something... no? What sample sizes should we be shooting for to establish confidence in a system? I think I need to try finding some odds data that goes back further than 2005.

            Thanks for the input.
            For quick and dirty data analysis you can try http://www.sundaystrategy.com. He has data back to about 2000 and you can do searches on totals and overs.

            For 2009 he shows 8 games equal to or less than 35.5 with 5-3 O/U. For equal to or greater than 47, he shows 48 games at 19-29 O/U. Which look to be about 60% combined...

            LT
            Comment
            • CFMiles
              SBR Rookie
              • 09-03-10
              • 4

              #7
              @Indecent - Well, I'm going to test it this season with a small unit. So let's hope I have a horseshoe up my ass! ;-)

              @LT - Thanks for the resource. My analysis showed 34-16 last year compared to the 34-22 you mentioned, so maybe we have a different odds provider. Still a good resource, though. Cheers!
              Comment
              Search
              Collapse
              SBR Contests
              Collapse
              Top-Rated US Sportsbooks
              Collapse
              Working...