What math could I use for this

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • Stocks
    SBR Wise Guy
    • 11-01-10
    • 569

    #1
    What math could I use for this
    I was looking at last years hockey results using percent of goals scored verses win% and I thought this was kind of interesting. Its like the results went up like a curve.

    Heres some result to show what I'm talking about

    First number is a teams percent of goals scored in a game like if there was 5 goals scored in a game and team A scored 3 then they would have scored 60% .

    Second number is win%

    58% = 66%
    54% = 59%
    50% = 50%
    47% = 43%
    43% = 33%

    I did it for all the teams last season and the results were similar so it difinitly shows that both are relavent.

    Something interesting the top team Washington scored 58% of the goals in games they played and had a 66 win % while the worst team was Edmonton who only scored 43% of goals in games they played and had a 33 win % thats almost the exact opposite.

    Anyway what kind of math or method could I use here to get a win% for a game if I can already predict a score.

    Thanks
  • subs
    SBR MVP
    • 04-30-10
    • 1412

    #2
    i'm pretty new so sorry but i don't get it - if you can predict a score then why do you need a win %, you already know. but this is common sense if you score more goals in games then you win more games. but it does not mean that you can predict future scores with more accuracy then the people setting the line.

    what you need to do is back test to see if this system will win you money (good luck with this). find the closers and see if you would have won or not over a season.

    you may wish to add
    home ice advantage
    strength of schedule
    injuries
    just to name a few.
    BOL
    Last edited by subs; 11-17-10, 11:53 PM.
    Comment
    • Stocks
      SBR Wise Guy
      • 11-01-10
      • 569

      #3
      What I'm looking at here is if I can predict a score and then turn that into a win% then I can turn it into a moneyline.

      So if I predict a score to be

      Washington 3.8
      Montreal 2.7

      Washington scored 58% of the goals so that would make them 66% change of winning and then we could turn that into a moneyline -200

      And we can account for any other factors like home ice advnage and what not into the predicted score.

      What I'm saying about the relationship between % of goals scored in a game verses win% is that its like a sliding scale where the higher the % of goals scored the winning percent goes up exponentially.

      So is pythagrom the way to go here?

      pythagrom gets very close to win% based off the goals scored I can mess with the exponent a bit.


      On to the predicted score see if I got this right.

      If the league average is 2.77 goals per game.

      Boston scores 2.9 goals per game while giving up 2.2 per game

      And they're facing

      Montreal scores 2.4 goals per game while giving up 2.6 per game

      This is not going to take into account home ice or current form or anthing else

      To find the predicted score do I find the difference between each teams goals scored for and against verses the league average.

      Boston scores .13 goals more and gives up .57 goals less then the league average
      Montreal scores .37 goals less and gives up .17 goals less then the league average

      So for Bostons score you would add they're goals scored and montreal goals against to the league average and do the same for montreal.

      Boston 2.77 + .13 -.17 = 2.73
      Montreal 2.77 -.37 -.57 = 1.83

      Predicted score

      Boston 2.73
      Montreal 1.83

      Using pythagrom

      Boston 68.9%
      Montral 31.1%

      Moneyline

      Boston -222
      Montreal +222
      Comment
      • subs
        SBR MVP
        • 04-30-10
        • 1412

        #4
        ok since no 1 else is helping and if i screw this up some1 might want to flame me and help stocks out, i'll give my $0.02 worth here.

        there are 2 ways of looking at this:

        1/ (OffBos * DefMon)/LeagueAverage

        (2.9 * 2.6)/2.77 = 2.72 BOSTON

        (2.4 * 2.2)/2.77 = 1.9 MONTREAL

        2/ OffBos + DefMon - LeagueAverage which is what you did. i think both are right. hopefully some1 better than me can tell us which is better but maybe an average of the 2 numbers would be best?

        2.9 + 2.6 - 2.77 = 2.73 BOSTON
        2.4 + 2.2 - 2.77 = 1.83 MONTREAL

        So this is where i prolly get it wrong but i used the 2 variable poisson calculator in the tools.
        ok using 2 variable poisson distribution i type 2.73 into event1 expectation and 1.83 into event 2 expectation and i get.

        57.0% for win -133
        25.1% for loss +298
        17.8% for push +462

        hope this is right and i'm not confusing the issue more. these do not add to 100% because of rounding and i approximated the lines from the %. please remember this is the no-vig line.

        hope this helps. having said all this i'm not sure what you have found here. yea for sure the team that scores more points than its opponents wins more games and vice versa. how this is helpful not sure. if you test this going back to seasons before i am pretty sure you will find a net loss but i'm a rookie really so please check it all out for yourself.

        good luck buddy
        Comment
        • Stocks
          SBR Wise Guy
          • 11-01-10
          • 569

          #5
          Thanks Subs

          I'm not sure about which of the 2 methods of predicting the score is right, I like yours better as it would fit into excel a lot better but I'll porobably end up trying both unless someone else wants to chime in on the situation.

          As for the next part I dont know if poisson will work or not but I have been looking into Pythagorean which is showing a lot of promise. I got a few ideas to test this that I'll try tomorrow.

          My theory for all this is if you can accurately predict a score of a hockey game its very much relavent to win% which you can then convert into a moneyline and find value.

          I'll post results tomorrow

          Thanks man
          Comment
          • subs
            SBR MVP
            • 04-30-10
            • 1412

            #6
            i know nothing about hockey but i know that in college basketball a guy called ken pomeroy does free public modeling, which is really deep. i have been told that these are not great but can be very useful for futures bets. how many games will a team win over a season type stuff. apparently this is an easier market to attack. it has lower limits to reflect this.

            good luck Stocks.
            Comment
            • Stocks
              SBR Wise Guy
              • 11-01-10
              • 569

              #7
              Heres some results using last season stats

              I compared a team win% to goal% using Pythagorean with an exponent of 2 for the goals.

              The win% was equal to or off by just 1 or 2 percent of the goal% for most teams.

              An example of what I'm talking about is

              Washington won 66% of they're games while scoring 3.82 goals per game and allowing 2.77 goals per game and using pythagorean equals 66%

              This was the kind of results for most teams. A couple were way off like Phoenix and Nashville

              Phoenix won 61% of its games 54% for goals but looking at the results from last season they ended the season on a nice hot streak going 13-5 with 9 games going to a shootout so they won a lot of 1 goal games and thats the reason why they're numbers are off same with Nashville. Now this could be corrected atleast a little bit by using current form as part of your predicted score.

              I also looked at 20 teams results just using home game stats and although the results showed promise they were'nt as good as the overall stats. There were still a lot of teams that were right on like Detroit who won 61% of they're home games and got 61% goals. Phoenix and Nashville were off aswell but other teams like Washington and Chicago who were exact using overall preformance but were off using just the home stats.

              I still think this has some promise I may start testing it this season at around the 20 game mark and see what happens.

              By the way to get the predicted score I would use

              Strength of schedule
              Home/away goals
              Overall goals
              Last 5 games goals
              Starting goalie
              Comment
              • uva3021
                SBR Wise Guy
                • 03-01-07
                • 537

                #8
                find the correlation between goals per shots and winning percentage, maybe from last 5-10 years of data, then you can extrapolate winning percentage from where a team's goals per shots falls on the least squares fit line

                use the excel function "Forecast"
                Comment
                SBR Contests
                Collapse
                Top-Rated US Sportsbooks
                Collapse
                Working...