Strategy

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • a4u2fear
    SBR Hall of Famer
    • 01-29-10
    • 8147

    #1
    Strategy
    Using a points system, I have come up with a way to estimate each teams points for in each game.

    Since I have every total since 2003 in Excel, I have the luxury of going back through and seeing how my estimates stack up versus the totals set by Vegas. These totals have ZERO trends in them. All based on stats from their prior games to estimate the current week's game.

    The stats below start in 2003 and I do not bet any lines until week 7 (so there is more consistency in team's performances)

    I add up both team's estimated scores and if there is a 4pt differential between my estimated total score compared to the vegas total, there were 81 occurrences since 2003 hitting at 59.3% rate. i.e. If my estimate is 4pts lower than vegas line, bet under, if 4pts higher, bet over.

    For a 7pt differential between my line and vegas, hits at 66.7% (60 occurrences)

    For a 14pt differential between my line and vegas, hits at 79% (28 occurrences)

    For a 20pt differential between my line and vegas, hits at 87.5% (8 occurrences)

    From what I can see, everything here makes sense and is HIGHLY profitable. I will divulge more info if requested but looking for thoughts and/or comments.
  • JMon
    SBR Hall of Famer
    • 12-11-09
    • 9800

    #2
    Interesting...will you be tracking here?
    Comment
    • a4u2fear
      SBR Hall of Famer
      • 01-29-10
      • 8147

      #3
      Originally posted by JMon
      Interesting...will you be tracking here?
      yea I will be tracking it in NFL forum. I tried using the same system for spreads, but I could not develop any strategy worth betting for it. I guess that means a couple of things - the point spreads set by vegas are better/tighter, have more variance (I think there is some saying like a 6 point line has something like a 25% chance of either team winning by more than 2 scores or something like that)
      Comment
      • HUY
        SBR Sharp
        • 04-29-09
        • 253

        #4
        Originally posted by a4u2fear
        Using a points system, I have come up with a way to estimate each teams points for in each game.

        Since I have every total since 2003 in Excel, I have the luxury of going back through and seeing how my estimates stack up versus the totals set by Vegas. These totals have ZERO trends in them. All based on stats from their prior games to estimate the current week's game.

        The stats below start in 2003 and I do not bet any lines until week 7 (so there is more consistency in team's performances)

        I add up both team's estimated scores and if there is a 4pt differential between my estimated total score compared to the vegas total, there were 81 occurrences since 2003 hitting at 59.3% rate. i.e. If my estimate is 4pts lower than vegas line, bet under, if 4pts higher, bet over.

        For a 7pt differential between my line and vegas, hits at 66.7% (60 occurrences)

        For a 14pt differential between my line and vegas, hits at 79% (28 occurrences)

        For a 20pt differential between my line and vegas, hits at 87.5% (8 occurrences)

        From what I can see, everything here makes sense and is HIGHLY profitable. I will divulge more info if requested but looking for thoughts and/or comments.
        Here are some thoughts and comments:

        a) Never accept a model on in-sample performance. It's the only 100% certain way of landing you in the poor house.
        b) 81 occurences since 2003 means 8 bets per year. You call that a significant sample size?
        c) Please divulge some info on your points system. You don't have to give it away if you don't want to, just some hints.
        Comment
        • a4u2fear
          SBR Hall of Famer
          • 01-29-10
          • 8147

          #5
          Sample size is those 81 games were chosen from every game from 2003-12, Weeks 7-17. Of course I'm not going to find a bunch of games every week because the Vegas lines are not that far off. What I think we all know and I sort of showed here is that it is rare to find a Vegas line that is off, but when I do, I can exploit it as you should be able to.

          The points system is a points per stat system. For example, if I think passing yards tells a much more meaningful story of how team's perform each game, I would take all of the points scored by a team up to week 7 (i.e. 200), and all of their passing yards up to that week (i.e. 1750), I estimate that for each passing yard that team gets, they score .114 points. What IS good about using points per categories is that it does tell the story that the final score does not: does a team get a lot of yards but not score often (has a lot of giveaways or is not a good redzone team), or does a team not get a lot of yard but does score (doesn't have the ball often because their defense is weak, or they are a big play team, or they could have a good special teams).

          Once I have these point per stat categories, I compare the team's offensive passing yards per point to their opponent's defensive passing yards per point. Both of these are also compared to the league median to justify whether their passing points per game are better or worse than average. I also take into account home/away for these instances.

          Since I also like to bet teasers, and I haven't done this for every category, but for a line that is off by 7, adding 6 points to the play now hits at 84% (need two bets of at least 73% to break even on a -110 teaser) and adding 13 points to the play now hits at 99% (need four bets of at least 89% to break even on a -150 teaser). Maybe this does not help anyone else, but I also have the stats on weekly categories that fit into the 13 point teaser model. Now having a leg that hits at 99% some weeks, will only increase my odds of hitting.
          Comment
          • brettd
            SBR High Roller
            • 01-25-10
            • 229

            #6
            Dude you're trying to manually regress various team stats to points. This type of analysis has been done (properly, not by hand) for a long time. Also you do not have out of sample result set.

            IMHO, you're not going to find any advantage doing this.
            Comment
            • sploofdogg
              SBR Sharp
              • 01-20-13
              • 335

              #7
              It is definitely intriguing though. I'd love to see that spreadsheet of every game since '03...
              Comment
              • a4u2fear
                SBR Hall of Famer
                • 01-29-10
                • 8147

                #8
                Originally posted by brettd
                Dude you're trying to manually regress various team stats to points. This type of analysis has been done (properly, not by hand) for a long time. Also you do not have out of sample result set.

                IMHO, you're not going to find any advantage doing this.
                I'm not regressing anything. If what I did can't work, then how could anyone ever estimate a team's points per games? It's the NFL and there are 16 games. Of course you will never have a large enough sample and you HAVE to estimate a team's points and compare them to the league average because there isn't a better way to compare the competition faced.

                I have not developed it using any trends just simple facts from each game played.

                How am I not going to find any advantage doing this? I have 10 years of proven data with it and the league has changed quite a bit in those ten years with it changing multiple rules to allow the offense to score more points. While you may say Vegas lines would eventually catch up with more points being score, I would tell you it doesn't matter because my system is not based off prior year's stats. It's based off how a team has performed and difference between my line and vegas line. If Vegas' line keeps edging higher for certain teams who score a lot of points, then my model will show that difference isn't there, and I shouldn't take the bet.
                Comment
                • a4u2fear
                  SBR Hall of Famer
                  • 01-29-10
                  • 8147

                  #9
                  Its week 7, the first week I always bet my model and its telling me that the line for Denver and Indy game is maybe not "way off" for halving the money, but that the game is going to go well over the posted total.

                  The estimate of my total vs the vegas line fits into a category where the model has performed at 87.5% from 2003 to 2012.

                  It helps we have a strong Manning and Bronco team, also playing his former team, but according to my model, it won't matter.

                  Den/Ind over 55.5

                  Good luck to all
                  Comment
                  • mikemca
                    SBR Posting Legend
                    • 03-10-10
                    • 10047

                    #10
                    Isn't this kind of what Team Rankings(Denver 31.6 - Indy 24.9) and Numberfire (Den 28.3 - Indy 22.2)do?
                    Comment
                    • Louisvillekid1
                      SBR Aristocracy
                      • 10-17-07
                      • 52143

                      #11
                      Is this just for NFL?
                      Comment
                      • a4u2fear
                        SBR Hall of Famer
                        • 01-29-10
                        • 8147

                        #12
                        I don't know what you're talking about, but the numbers above you show the total falling on or near the total. I'm predicting it's going well over
                        Comment
                        • a4u2fear
                          SBR Hall of Famer
                          • 01-29-10
                          • 8147

                          #13
                          Originally posted by Louisvillekid1
                          Is this just for NFL?
                          I'm working on it for hockey too but not far enough yet
                          Comment
                          • Louisvillekid1
                            SBR Aristocracy
                            • 10-17-07
                            • 52143

                            #14
                            8 times since 2003 and now this week we have the 9th occurrence?
                            Comment
                            • a4u2fear
                              SBR Hall of Famer
                              • 01-29-10
                              • 8147

                              #15
                              Originally posted by Louisvillekid1
                              8 times since 2003 and now this week we have the 9th occurrence?
                              You are correct, so, in this case small sample size, but only because the estimated line and Vegas line are so far off, which is highly unusual.

                              makes sense as the broncos offense being so good is also very unusual.
                              Comment
                              • a4u2fear
                                SBR Hall of Famer
                                • 01-29-10
                                • 8147

                                #16
                                Where there is a 20 pt difference in the line, now is 89% or 8-1 with the colts and broncos over last night
                                Comment
                                • a4u2fear
                                  SBR Hall of Famer
                                  • 01-29-10
                                  • 8147

                                  #17
                                  For this week of a line of 56 or 57 in Denver/Washington, it looks like it will fall into the 14 pt model with it being the 29th occurrence falling at a rate of 79%
                                  Comment
                                  • a4u2fear
                                    SBR Hall of Famer
                                    • 01-29-10
                                    • 8147

                                    #18
                                    Originally posted by a4u2fear
                                    For this week of a line of 56 or 57 in Denver/Washington, it looks like it will fall into the 14 pt model with it being the 29th occurrence falling at a rate of 79%


                                    the 29th occurrence is a winner making the 14 pt difference still 79%. I think it was 78.5% rounded up and is now 79.3% (23-6).

                                    Game fell at 69 pts and the system estimated 71 pts. Last week the total fell at 72 pts and system estimated 79 pts
                                    Comment
                                    • VegasInsider
                                      SBR Posting Legend
                                      • 12-12-10
                                      • 14593

                                      #19
                                      Very good stuff, fear. I'm looking forward to seeing your posts for next week (if any qualify)
                                      Comment
                                      • a4u2fear
                                        SBR Hall of Famer
                                        • 01-29-10
                                        • 8147

                                        #20
                                        Originally posted by VegasInsider
                                        Very good stuff, fear. I'm looking forward to seeing your posts for next week (if any qualify)
                                        yea, will let you know. With Denver off, its unlikely.

                                        In regards to games that do not include certain teams, is dead nuts on and other times not.

                                        It accurately predicted the GB/MIN would be the second highest scoring game of the week.
                                        It predicted a 51-20 game for the Was/Den.
                                        It predicted 25-15 for NE/Mia
                                        It predicted 38-17 for NO/Buf
                                        It predicted 34-10 for Jax/SF
                                        It predicted 39-19 for Cin/NYJ

                                        Those were the closest models predicted and the others were not as close. I haven't found a great way to use the model for estimating lines/spreads even though what I posted above shows its very close. I started on totals specifically and stayed there, without focusing hard on spreads.
                                        Comment
                                        • Philyerglesion76
                                          SBR Rookie
                                          • 11-18-11
                                          • 16

                                          #21
                                          I agree with others, this is definitely intriguing! I'm interested in seeing more picks. Have you done it for NCAAF, or would it be too difficult with such a difference in scheduling?

                                          Comment
                                          • a4u2fear
                                            SBR Hall of Famer
                                            • 01-29-10
                                            • 8147

                                            #22
                                            Originally posted by Philyerglesion76
                                            I agree with others, this is definitely intriguing! I'm interested in seeing more picks. Have you done it for NCAAF, or would it be too difficult with such a difference in scheduling?

                                            I have 0 interest in NCAAF and don't have prior years worth of data in excel so it would be impossible. Plus, it would be so difficult to compare performances since there are so many teams, conferences, and good/bad teams
                                            Comment
                                            • YourAllAmerican
                                              SBR Rookie
                                              • 06-25-09
                                              • 47

                                              #23
                                              Two questions:

                                              1. How did you go about forming your formula? For instance, did you take all games from the 2003-2012 seasons and use those stats to create your formula? (Not so much interested in the formula itself, but rather just how it was made.)

                                              2. After creating the formula, how did you backtest it?
                                              Comment
                                              • a4u2fear
                                                SBR Hall of Famer
                                                • 01-29-10
                                                • 8147

                                                #24
                                                Originally posted by YourAllAmerican
                                                Two questions:

                                                1. How did you go about forming your formula? For instance, did you take all games from the 2003-2012 seasons and use those stats to create your formula? (Not so much interested in the formula itself, but rather just how it was made.)

                                                2. After creating the formula, how did you backtest it?
                                                The formula was created after reading some articles on points per yard. I found it very intriguing because other factors can be seen in the points per yard that you wouldn't think are in it. Such as turnovers. There are only a few a game and some can be in more important places on the field than others. If you consistently drive down the field and then either a) miss FGs, b) turn the ball over, or c) have a bad redzone offense (more FGs than TDs), this will all be displayed in the points per yard. As wel as if you have a good special teams. You will possibly need less yards to gain points because your field position is better etc etc.

                                                Points per yard isn't all of it, but it is the foundation. It was not developed using prior games or data or anything. I've been developing formulas for 3 years now and back testing them, so I have a generally good idea of what will or won't work.

                                                So I created the formula and estimation of teams points and back tested it to the 2003 season. I wanted to make sure all 32 teams that are current were playing, and I can't remember why I decided 2003 vs 2002, maybe it was 10 years. A reason I also did not want to go back further was because I felt the style of play had changed quite a bit in those 10 years. Hits to the head were diminishing, QBs were protected much more. Games were higher scoring due to the increased amount of defensive holding/interference and guaranteed first downs because of those plays. The team was moving from a more running league or 50/50 to a more dominant passing league.

                                                I did find out that the system began to test more poorly in weeks 6 and back; mainly because teams do not have a sufficient amount of home and away games.
                                                Comment
                                                • cmaulsby
                                                  SBR Sharp
                                                  • 03-28-12
                                                  • 261

                                                  #25
                                                  Very interested in NHL totals - please post. With the NFL, do you have team totals in your model or an aggregate score based on yardage?
                                                  Comment
                                                  • YourAllAmerican
                                                    SBR Rookie
                                                    • 06-25-09
                                                    • 47

                                                    #26
                                                    Originally posted by a4u2fear
                                                    So I created the formula and estimation of teams points and back tested it to the 2003 season.
                                                    The data you used to create the formula, though, what seasons did it come from?

                                                    On a side note, I very often use 2002 as the cutoff for much of my own tests. That's the season where the current bye-week scheduling came into play.
                                                    Comment
                                                    • a4u2fear
                                                      SBR Hall of Famer
                                                      • 01-29-10
                                                      • 8147

                                                      #27
                                                      Originally posted by cmaulsby
                                                      Very interested in NHL totals - please post. With the NFL, do you have team totals in your model or an aggregate score based on yardage?
                                                      I haven't used this system for NHL totals, but tried briefly with shots-unsuccessfully. I DO believe I have just recently found an NHL system to work. For picking teams not the total. I found what I believed was the single most important stat for the outcome of every game. So I went back and calculated it for each team and starting game 20 thru 80 using it and comparing it to its opponent, it was +110 units (at 1 a game) over 6 years. Not much back testing but amazing results. Again I just found it and it could be misfortune, but at first glance it looks very promising.

                                                      For NFL, I have team totals, if you look a few posts back, it came very close to the real scores from last weeks NFL.. I have not focused much on its winning pct with spreads because usually the outcome in a game can swing 10-20 pts from one team to another, but the totals seem to be "somewhat" steady.
                                                      Comment
                                                      • a4u2fear
                                                        SBR Hall of Famer
                                                        • 01-29-10
                                                        • 8147

                                                        #28
                                                        Originally posted by YourAllAmerican
                                                        The data you used to create the formula, though, what seasons did it come from?

                                                        On a side note, I very often use 2002 as the cutoff for much of my own tests. That's the season where the current bye-week scheduling came into play.
                                                        I created the formula not using prior years data. But I did need stats from each game to formulate the estimate for the current week - aka i needed 2003 weeks 1-6 stats (yds for/against etc) to estimate a team's week 7 performance to test my formula.
                                                        Comment
                                                        • HUY
                                                          SBR Sharp
                                                          • 04-29-09
                                                          • 253

                                                          #29
                                                          You still haven't cleared up whether your system really makes only 8 picks per year. If so, why are we discussing it? It's impossible to evaluate a system that makes so few plays.
                                                          Comment
                                                          • a4u2fear
                                                            SBR Hall of Famer
                                                            • 01-29-10
                                                            • 8147

                                                            #30
                                                            Originally posted by HUY
                                                            You still haven't cleared up whether your system really makes only 8 picks per year. If so, why are we discussing it? It's impossible to evaluate a system that makes so few plays.
                                                            Sir, it's the NFL. What system of any kind exists that has many plays during an NFL season? It does not exist and is impossible because Vegas and $ are much bigger than you and I. There are only 17 weeks, and not always 16 games per week with bye weeks. It's impossible to estimate exactly how each team performs because they play so few games and each game is at a differently talented opponent which is just as equally difficult to address. If you're looking for a system with tons of selections you will grow old trying

                                                            the more picks, the worse the outcome. My system determines when the Vegas line is further off than it should be and exploits it. That should not be often as we all know.

                                                            Ifyou dont believe it, dont follow it, but everyone is a non-believer before they believe. I'm not here as a tout or looking for fandom. I'm here to post what I think is a very good lean on a system that could make a difference.
                                                            Comment
                                                            • chunk
                                                              SBR Wise Guy
                                                              • 02-08-11
                                                              • 808

                                                              #31
                                                              Intriguing thread to this point. I am looking forward to see if there may be something of value for all which would be a rare occurrence indeed.......especially when attacking the NFL which is one tough nut to crack.
                                                              Comment
                                                              • floridagolfer
                                                                SBR MVP
                                                                • 12-19-08
                                                                • 2757

                                                                #32
                                                                I wish you all the luck in the world, especially in the belief that something that happened 10 years ago might have any impact whatsoever on a game played in 2013.
                                                                Comment
                                                                • a4u2fear
                                                                  SBR Hall of Famer
                                                                  • 01-29-10
                                                                  • 8147

                                                                  #33
                                                                  Originally posted by floridagolfer
                                                                  I wish you all the luck in the world, especially in the belief that something that happened 10 years ago might have any impact whatsoever on a game played in 2013.
                                                                  You're way off and haven't read the thread. Only thing from 2003 is that it was back tested as everything needs to be. The % is from all of the last years it was tested. There are confidence intervals and success rates
                                                                  Comment
                                                                  • a4u2fear
                                                                    SBR Hall of Famer
                                                                    • 01-29-10
                                                                    • 8147

                                                                    #34
                                                                    Never once say it is a trend or anything close to it.
                                                                    Comment
                                                                    • a4u2fear
                                                                      SBR Hall of Famer
                                                                      • 01-29-10
                                                                      • 8147

                                                                      #35
                                                                      No system plays this week.

                                                                      I haven't tracked the following plays in the past, but system likes the over in:
                                                                      GB/CHI
                                                                      DAL/MIN
                                                                      Comment
                                                                      Search
                                                                      Collapse
                                                                      SBR Contests
                                                                      Collapse
                                                                      Top-Rated US Sportsbooks
                                                                      Collapse
                                                                      Working...