Welcome Guest
Log In | Register )
You last visited January 17, 2017, 5:54 am
All times shown are
Eastern Time (GMT-5:00)

Nonlinear regression to predict lotteries

Topic closed. 119 replies. Last post 2 years ago by lottoswe.

Page 1 of 8
53
PrintE-mailLink
Avatar
Tahiti- Polynesia
Tuvalu
Member #34524
March 4, 2006
54 Posts
Offline
Posted: September 6, 2014, 5:20 am - IP Logged

Hi,

I have used nonlinear regression for a long time. Two times, one just 3 drawings ago, the program gave the 5 main balls for Euromillions. More surprising, the two times it hit, the 5 numbers were in the order  they popped out the machine.

Others times, I got a lot of 4 numbers out of 15 numbers and 3 numbers as well.

Has anybody experienced such results using nonlinear regression?

Thanks,

 

Bob

    Avatar
    Tahiti- Polynesia
    Tuvalu
    Member #34524
    March 4, 2006
    54 Posts
    Offline
    Posted: September 6, 2014, 5:34 am - IP Logged

    Hi,

    I'm not kidding. I just want to discuss with experienced members who use nonlinear regression. This method is used in some lottery software. I don't say I have won the jackpot. I just got those surprising results when backtesting my game.

    And I have nothing to sell.

     

    Bob

      RJOh's avatar - chipmunk
      mid-Ohio
      United States
      Member #9
      March 24, 2001
      19894 Posts
      Offline
      Posted: September 6, 2014, 6:09 pm - IP Logged

      Hi,

      I'm not kidding. I just want to discuss with experienced members who use nonlinear regression. This method is used in some lottery software. I don't say I have won the jackpot. I just got those surprising results when backtesting my game.

      And I have nothing to sell.

       

      Bob

      You get all kinds of results when you back test that can't be duplicated when you forward test by actually playing.  Has you won anything playing combinations you picked that way?

       * you don't need to buy more tickets, just buy a winning ticket * 
         
                   Evil Looking       

        Avatar
        Tahiti- Polynesia
        Tuvalu
        Member #34524
        March 4, 2006
        54 Posts
        Offline
        Posted: September 6, 2014, 9:47 pm - IP Logged

        You get all kinds of results when you back test that can't be duplicated when you forward test by actually playing.  Has you won anything playing combinations you picked that way?

        Hi RJOh,

         

        My standard method of playing is to use nonlinear regression to predict numbers.
        I choose 15n and use a conditional wheel to get my lines which I filter with usual filters as sums, AC, Last digits, Decades.
        I am let with about 10 lines.

        I won many prices, like 2/5, 3/5 or 4/5. The fifth number is always difficult to get.
        Last weeks, I was working with WinHunter to find numbers to eliminate and didn't played. On the 26 aug 2014, when I checked my method, I saw that I hit 5/5, the very first numbers of the 15n pool. The same thing happened in November 2010. Two drawings ago, it gave me 4/5, 3 numbers beeing the 3 first numbers of the 15n pool.

        So, with nonlinear regression, if you choose the right parameters, it will work for the future. Now, I know which parameters work. You just have to adjust one variable parameter which is the time delay. Dimension is 3 and bandwidth is 1. That's all. The programmer who coded my nonlinear program also coded one program to find the time delay.

        Now, the program will not hit at any time. It's rather good at finding 4 numbers out of 15, but generally the better results come from the very first 6 to 10 numbers. I use 15n to try to get that elusive 5th number.

        Bob

          Avatar
          South Carolina
          United States
          Member #18322
          July 9, 2005
          1743 Posts
          Offline
          Posted: September 7, 2014, 12:36 am - IP Logged

          Hi RJOh,

           

          My standard method of playing is to use nonlinear regression to predict numbers.
          I choose 15n and use a conditional wheel to get my lines which I filter with usual filters as sums, AC, Last digits, Decades.
          I am let with about 10 lines.

          I won many prices, like 2/5, 3/5 or 4/5. The fifth number is always difficult to get.
          Last weeks, I was working with WinHunter to find numbers to eliminate and didn't played. On the 26 aug 2014, when I checked my method, I saw that I hit 5/5, the very first numbers of the 15n pool. The same thing happened in November 2010. Two drawings ago, it gave me 4/5, 3 numbers beeing the 3 first numbers of the 15n pool.

          So, with nonlinear regression, if you choose the right parameters, it will work for the future. Now, I know which parameters work. You just have to adjust one variable parameter which is the time delay. Dimension is 3 and bandwidth is 1. That's all. The programmer who coded my nonlinear program also coded one program to find the time delay.

          Now, the program will not hit at any time. It's rather good at finding 4 numbers out of 15, but generally the better results come from the very first 6 to 10 numbers. I use 15n to try to get that elusive 5th number.

          Bob

          Perhaps you should explain exactly what nonlinear regression is, and how it works.  Give us an example from a lottery game.

            Avatar
            Tahiti- Polynesia
            Tuvalu
            Member #34524
            March 4, 2006
            54 Posts
            Offline
            Posted: September 7, 2014, 1:38 am - IP Logged

            Perhaps you should explain exactly what nonlinear regression is, and how it works.  Give us an example from a lottery game.

            Hi destinycreation,

            I'll try to keep it simple. 

            Linear or nonlinear regression is a mathematic tool to predict time series, i.e a serie of numbers,in many domains as meteorology, economy,etc... Lotto numbers are a time series but scientists think that lottery cannot be forecasted with regression.

            If you create a graph with your lottery numbers, you'll see a sawtooth like graph. Regression consists in finding the formula of the graph that can represent your numbers, the closest possible. We say that we model the curb. Having the formula for your lotto, you can then use it to predict the future, if the formula is simple. If not prediction will be difficult.

            There are time series which can fit relatively easily when variations between numbers are small. That is the case when you play a lotto say with numbers between 1 and 10. You can use linear regression in that case. With Euromillions 5/50, with numbers between 1 and 50, variations are large and the curb is chaotic. Prediction is more difficult. We use nonlinear regression.

            This matter is difficult for a non-mathematician to explain. I suggest that you visit the many sites on the web where you'll grasp more easily the subject. Looking at the graphs will help.

            There is a lot more to say about regression but I cannot develop more for now. I hope that my explanations will help.

            I hope somebody with mathematical background will come and explain better than I.

             

            Bob

              Avatar
              Tahiti- Polynesia
              Tuvalu
              Member #34524
              March 4, 2006
              54 Posts
              Offline
              Posted: September 7, 2014, 3:09 am - IP Logged

              Fitting a curb to the number graph curb is the first step of regression. Regression itself (regression meaning going back) consist in going back in history to find similar patterns as those it find in the last drawings and look what pattern follow. It then uses this information to make prediction. The process is more complicated but I hope this will give a general idea of what is regression.

               

              Bob

                RJOh's avatar - chipmunk
                mid-Ohio
                United States
                Member #9
                March 24, 2001
                19894 Posts
                Offline
                Posted: September 7, 2014, 10:37 am - IP Logged

                Fitting a curb to the number graph curb is the first step of regression. Regression itself (regression meaning going back) consist in going back in history to find similar patterns as those it find in the last drawings and look what pattern follow. It then uses this information to make prediction. The process is more complicated but I hope this will give a general idea of what is regression.

                 

                Bob

                I'll be following your predictions on the prediction board.  Good luck.

                 * you don't need to buy more tickets, just buy a winning ticket * 
                   
                             Evil Looking       

                  MillionsWanted's avatar - 24Qa6LT

                  Norway
                  Member #9517
                  December 10, 2004
                  1279 Posts
                  Offline
                  Posted: September 7, 2014, 11:14 am - IP Logged

                  Hi destinycreation,

                  I'll try to keep it simple. 

                  Linear or nonlinear regression is a mathematic tool to predict time series, i.e a serie of numbers,in many domains as meteorology, economy,etc... Lotto numbers are a time series but scientists think that lottery cannot be forecasted with regression.

                  If you create a graph with your lottery numbers, you'll see a sawtooth like graph. Regression consists in finding the formula of the graph that can represent your numbers, the closest possible. We say that we model the curb. Having the formula for your lotto, you can then use it to predict the future, if the formula is simple. If not prediction will be difficult.

                  There are time series which can fit relatively easily when variations between numbers are small. That is the case when you play a lotto say with numbers between 1 and 10. You can use linear regression in that case. With Euromillions 5/50, with numbers between 1 and 50, variations are large and the curb is chaotic. Prediction is more difficult. We use nonlinear regression.

                  This matter is difficult for a non-mathematician to explain. I suggest that you visit the many sites on the web where you'll grasp more easily the subject. Looking at the graphs will help.

                  There is a lot more to say about regression but I cannot develop more for now. I hope that my explanations will help.

                  I hope somebody with mathematical background will come and explain better than I.

                   

                  Bob

                  Show us how you do it in a spreadsheet.

                   

                  (Wow! My 1000th post!)

                    SergeM's avatar - slow icon.png
                    Economy class
                    Belgium
                    Member #123700
                    February 27, 2012
                    4035 Posts
                    Offline
                    Posted: September 7, 2014, 2:00 pm - IP Logged

                    I also don't see how you want to use regression.

                    There are functions in Excel for regression.

                    You must select few numbers because it is 2 euros for one combination.

                      Avatar
                      Tahiti- Polynesia
                      Tuvalu
                      Member #34524
                      March 4, 2006
                      54 Posts
                      Offline
                      Posted: September 7, 2014, 4:25 pm - IP Logged

                      Show us how you do it in a spreadsheet.

                       

                      (Wow! My 1000th post!)

                      Hi MillionsWanted,

                      I have never tried to use regression with a spreadsheet. I know there are addins in Excel to do that.

                      I hope somebody on this site could create an example.

                      Sorry.

                      Bob

                        Avatar
                        Tahiti- Polynesia
                        Tuvalu
                        Member #34524
                        March 4, 2006
                        54 Posts
                        Offline
                        Posted: September 7, 2014, 4:35 pm - IP Logged

                        I also don't see how you want to use regression.

                        There are functions in Excel for regression.

                        You must select few numbers because it is 2 euros for one combination.

                        Hi SergeM,

                        Yes, you can use regression with Excel but I don't know how to do it. I use a program coded by a professional coder I rented. It costed me some hundreds $$ bucks!

                        Euromillions cost 3 euros for one combination here in Tahiti. I'm trying to optimize my games. My goal is to play 10 numbers, wheel them, then filter the combinations. It will let me with few lines.

                         

                        Bob

                          Avatar
                          Tahiti- Polynesia
                          Tuvalu
                          Member #34524
                          March 4, 2006
                          54 Posts
                          Offline
                          Posted: September 7, 2014, 4:43 pm - IP Logged

                          I'll be following your predictions on the prediction board.  Good luck.

                          Hi RJOh,

                          I will probably post some predictions for Powerball. I do have the powerball history. I however wanted to know when the last changes occured in that game. Changes for the main balls and changes for the powerball. I want to use a coherent history. Drawings that came before any changes must be eliminated if you want to get the right statistics. Also, I delete the 100 drawings after any change. 

                           

                          Bob

                            Avatar
                            Tahiti- Polynesia
                            Tuvalu
                            Member #34524
                            March 4, 2006
                            54 Posts
                            Offline
                            Posted: September 7, 2014, 6:18 pm - IP Logged

                            Hi,

                             

                            There is more than just use your game file and predict with linear or nonlinear regression.

                            That's not a push button method.

                            1 - you must use your game history in the drawn order. Don't sort it.

                            2 - forget classical statistics as fequency, skips, etc..

                            3 - Regression need that you analyse your file to test it for normality, autocorrelation, stationarity, seasonality.

                            4 - Those results will tell you if the file need to be transformed or not. Transformation will be needed most of the time. Depends upon your lottery. Most lotteries need transformation. This will help getting a more appropriate modeling. I personnaly use the BoxCox transform. The resulting file will look as follow :

                            4.3505932159215
                            11.0771790164803
                            12.6013163423106
                            9.73738708163018
                            14.5223617514636
                            12.8478218642374
                            3.95862233956806

                            Powerball file needs to be transformed.

                            5 - After transformation and prediction, you'll need to inverse transform the results to get the real numbers.

                            I use the "R" software to realize BoxCox transform and inverse transform.

                            I didn't wanted to let you believe that this is an easy method. It needs lot of work. It took me more than 15 years to begin master this system.


                            Bob

                              SergeM's avatar - slow icon.png
                              Economy class
                              Belgium
                              Member #123700
                              February 27, 2012
                              4035 Posts
                              Offline
                              Posted: September 8, 2014, 5:31 am - IP Logged

                              You do regression in Excel with functions or built in macro.