Welcome Guest
Log In | Register )
You last visited December 5, 2016, 3:33 pm
All times shown are
Eastern Time (GMT-5:00)

Neural Net Lottery Picker

677 replies. Last post 9 days ago by dr san.

Page 26 of 46
511
PrintE-mailLink
lottolot's avatar - Sphere animated2.gif
Tanhauser Gates
Holy See (Vatican City State)
Member #139281
February 18, 2013
173 Posts
Offline
Posted: January 5, 2016, 8:36 am - IP Logged

Perhaps time to check out other ways to use the number outputs from the NNLP?

For example let it output a lot of bets and find out which numbers appear most often and use them in a wheel.

I am planning to do this at a later stage myself.

Hi MillionsWanted:Smile

Thanks for your strategie,but I´ll try more deeply my theory of filtering the output of NNLP.Smash

But ,again, it´s a good idea for work in it.I Agree!

Thanks and Regards.Thumbs Up

    Avatar
    New Member
    Austin, texas
    United States
    Member #158905
    September 7, 2014
    1 Posts
    Offline
    Posted: January 16, 2016, 11:43 am - IP Logged

    Has there been any thought to applying NN to scratch off games? My thoughts are that lotto games are random drawings of balls, whereas scratch off jackpot tickets have to be controlled by the lottery in some way to make sure the same retailer is not getting the jackpot tickets. The lottery has to know where all of the jackpot tickets are going to be placed in advance. By tracking the retailer location of the previous jackpot tickets we might to be able to find out where the future jackpots tickets are located. It can also be broken down by game type/cost. Any thoughts?????

      MillionsWanted's avatar - 24Qa6LT

      Norway
      Member #9517
      December 10, 2004
      1272 Posts
      Online
      Posted: January 21, 2016, 12:09 pm - IP Logged

      Won 3 out of 6 on Viking Lotto yesterday. Which was enough to cover the cost of playing. Played 10 lines.

      Created  25 lines originally and removed the most unlikely lines and was left with 10 to play.

        lottolot's avatar - Sphere animated2.gif
        Tanhauser Gates
        Holy See (Vatican City State)
        Member #139281
        February 18, 2013
        173 Posts
        Offline
        Posted: January 23, 2016, 6:33 pm - IP Logged

        Won 3 out of 6 on Viking Lotto yesterday. Which was enough to cover the cost of playing. Played 10 lines.

        Created  25 lines originally and removed the most unlikely lines and was left with 10 to play.

        Congrats!!MillionsWanted.

        A win is a win,even is little.

        Best regards.

          MillionsWanted's avatar - 24Qa6LT

          Norway
          Member #9517
          December 10, 2004
          1272 Posts
          Online
          Posted: January 24, 2016, 11:37 am - IP Logged

          Congrats!!MillionsWanted.

          A win is a win,even is little.

          Best regards.

          Thanks, but it's perhaps time to find out how to win better prizes. 3 out of 6 is not really hard to achieve.

            lottolot's avatar - Sphere animated2.gif
            Tanhauser Gates
            Holy See (Vatican City State)
            Member #139281
            February 18, 2013
            173 Posts
            Offline
            Posted: January 24, 2016, 8:50 pm - IP Logged

            Thanks, but it's perhaps time to find out how to win better prizes. 3 out of 6 is not really hard to achieve.

            I Agree!Thumbs UpCheers.

            Best Regards

              Avatar

              United States
              Member #98539
              October 9, 2010
              57 Posts
              Offline
              Posted: January 27, 2016, 8:59 pm - IP Logged

              Hi MillionsWanted,

               

              I can't give you recommended settings for this tool because I haven't done enough testing.

              While testing last week, this tool I gave me a match 5 (back-testing 10 (upcoming) draw results, pick 6/45) and sometimes a match 4, 3 and many times a match 2.

              Here some match results (back testing the latest 10 draw results, pick 6/45. see image below the settings I used ):

              I used the following settings (these settings gave me last week a match 5 while testing for my game):

              When after a while no new results are found by the tool you can stop the training en start it again to collect lottery picks (and do some analysis).

              As I stated earlier, I'm not asserting this tool produces better picks than other techniques (more tests are required).

              I developed this tool for fun and was curious how to apply an ANN algorithm for different lottery games on a more generic an easy way (compared to other products).

              Regarding the Perceptlotron software (version 1.0) quoted above:

               

              Results: 990

               

              Just how long did Stoopendaal run the software for to get 990 results?

               

              Does anyone have an educated guess.

               

              I find in using the Perceptlotron version 1.6 software that the software completes its calculation and finishes training rather instantly.

               

              Thanks.

                MillionsWanted's avatar - 24Qa6LT

                Norway
                Member #9517
                December 10, 2004
                1272 Posts
                Online
                Posted: January 27, 2016, 9:54 pm - IP Logged

                Regarding the Perceptlotron software (version 1.0) quoted above:

                 

                Results: 990

                 

                Just how long did Stoopendaal run the software for to get 990 results?

                 

                Does anyone have an educated guess.

                 

                I find in using the Perceptlotron version 1.6 software that the software completes its calculation and finishes training rather instantly.

                 

                Thanks.

                990 is not the results from the prediction.

                990 is the number of lottery draws used for the predictions.

                  Avatar

                  United States
                  Member #98539
                  October 9, 2010
                  57 Posts
                  Offline
                  Posted: January 27, 2016, 10:35 pm - IP Logged

                  Hi jlg69,

                   

                  "what could be the aim (advantage)  to get more hidden neurons at your point of view .?"  There a different rules-of-thumb to determine the number of neurons in the hidden layer to optimize the neural network for your data (in our case to get better predictions Smile)

                  Here some rules you can test to find the best number of neurons in the hidden layer:

                  • 1

                  In sum, for most problems, one could probably get decent performance (even without a second optimization step) by setting the hidden layer configuration using just two rules: (i) number of hidden layers equals one; and (ii) the number of neurons in that layer is the mean of the neurons in the input and output layers.

                  The ANN in perceplotron has 2 input neurons (and 1 hidden layer) and 1 output neuron. So the number of neurons in the hidden layer

                  should be (2+1) / 2  = 2 (when round up)  according this rule.

                  • 2

                  One additional rule of thumb for supervised learning networks, the upperbound on the number of hidden neurons that won't result in over-fitting is:

                  Nh=Ns(alpha(Ni+No))

                  Ni = number of input neurons.
                  No = number of output neurons.
                  Ns = number of samples in training data set.
                  alpha = an arbitrary scaling factor usually 2-10.

                  Others recommend setting alpha to a value between 5 and 10, but I find a value of 2 will often work without overfitting. As explained by this excellent NN Design text, you want to limit the number of free parameters in your model (its degree or number of nonzero weights) to a small portion of the degrees of freedom in your data. The degrees of freedom in your data is the number samples * degrees of freedom (dimensions) in each sample or Ns(Ni+No) (assuming they're all independent). So alpha is a way to indicate how general you want your model to be, or how much you want to prevent overfitting.

                  For an automated procedure you'd start with an alpha of 2 (twice as many degrees of freedom in your training data as your model) and work your way up to 10 if the error for training data is significantly smaller than for the cross-validation data set.

                   

                  For example you train the ANN with the (latest) 80 draw results (= samples in training data set)

                  For this training set the number of neurons in your hidden layer should be :  80 / (2*(2+1))  = 80/6 = 13.333 = 13  (for alpha = 2 in this example)

                   

                  source:  http://stats.stackexchange.com/questions/181/how-to-choose-the-number-of-hidden-layers-and-nodes-in-a-feedforward-neural-netw

                   

                  Good luck!

                  This is regarding the Perceptlotron software.   I have version 1.6 running on my computer.

                   

                  The alpha you mention in the quoted text is the same as Learning Rate.   You say that you can set the value to between 5 and 10 but that you find that a value of 2 will often work without overfitting.

                   

                  I am a bit confused.   In using the Perceptlotron software I find that the values of the Learning Rate (alpha) go from the lowest value or setting of 1E-05 to the highest value of 1.   There is no value higher than 1.

                  That means I cannot set the value of the Learning Rate (alpha) to be higher than 1.

                   

                  Does anybody have an answer for the discrepancy between what I find in using the software with the maximum value of 1 for the Learning Rate and what Stoopendaal says is the highest value of 10?

                  Any light that is shed on this would be helpful. Thanks.

                    Avatar

                    United States
                    Member #98539
                    October 9, 2010
                    57 Posts
                    Offline
                    Posted: January 27, 2016, 10:47 pm - IP Logged

                    I find it rather curious that 990 or 976 draws is what is used for the predictions as Stoopendaal used in his examples that he posted earlier on in the forums.   I notice that Adamcustom used 5619 drawings to get his prediction results (see page. 13 in this forum).

                     

                    I have a hard time understanding this since Stoopendaal keeps saying that for the best results, only 10 to 70 draw results are needed for his game of 6 /45.

                    He has said this in the forums and in his email replies to me.

                    He said that if a person uses more results, the more noise the neural network gets resulting in bad predictions in his experience.

                     

                    So I am now wondering what the right answer is and whether I should use more than 100 drawings or even use the entire game history to load into the software.

                     

                    It would be very much appreciated if anybody can shed any light on this matter.

                     

                    Thanks.

                      MillionsWanted's avatar - 24Qa6LT

                      Norway
                      Member #9517
                      December 10, 2004
                      1272 Posts
                      Online
                      Posted: January 28, 2016, 6:16 am - IP Logged

                      I find it rather curious that 990 or 976 draws is what is used for the predictions as Stoopendaal used in his examples that he posted earlier on in the forums.   I notice that Adamcustom used 5619 drawings to get his prediction results (see page. 13 in this forum).

                       

                      I have a hard time understanding this since Stoopendaal keeps saying that for the best results, only 10 to 70 draw results are needed for his game of 6 /45.

                      He has said this in the forums and in his email replies to me.

                      He said that if a person uses more results, the more noise the neural network gets resulting in bad predictions in his experience.

                       

                      So I am now wondering what the right answer is and whether I should use more than 100 drawings or even use the entire game history to load into the software.

                       

                      It would be very much appreciated if anybody can shed any light on this matter.

                       

                      Thanks.

                      Perhaps I should correct myself a little.

                      He was loading 990 draws into Percep-Lotron, but he only used the latest 15 draws for prediction.

                      Se "Use Latest: 15". Adamcustom only used the latest 27 out of 5619 draws.

                        Avatar

                        United States
                        Member #98539
                        October 9, 2010
                        57 Posts
                        Offline
                        Posted: January 28, 2016, 2:45 pm - IP Logged

                        Dear Millions Wanted:

                         

                        Thanks.

                         

                        Kosh

                          Avatar

                          United States
                          Member #168877
                          September 20, 2015
                          43 Posts
                          Offline
                          Posted: February 1, 2016, 1:14 pm - IP Logged

                          I am using NNLP - I find it to be a very powerful tool for any one to play with. It's just a matter of learning how to set the right parameters, thus to bring the output very close to Winning numbers.

                          Based on most posts I have read, I find it that many of us, if not all, are setting incorrect parameters hence we're not getting the results we look for.

                          I wanted to introduce a course I am reading which depicts and fills many gaps we may have in machine learning. If interested, search in google for "Neural Networks and Deep Learning" - it should be the first link that shows up. Understanding in statistics and calculus may be recommended but not necessary. Machine learning is a subject which can only be learned with months of study, if not years.

                            lottolot's avatar - Sphere animated2.gif
                            Tanhauser Gates
                            Holy See (Vatican City State)
                            Member #139281
                            February 18, 2013
                            173 Posts
                            Offline
                            Posted: February 1, 2016, 2:44 pm - IP Logged

                            I am using NNLP - I find it to be a very powerful tool for any one to play with. It's just a matter of learning how to set the right parameters, thus to bring the output very close to Winning numbers.

                            Based on most posts I have read, I find it that many of us, if not all, are setting incorrect parameters hence we're not getting the results we look for.

                            I wanted to introduce a course I am reading which depicts and fills many gaps we may have in machine learning. If interested, search in google for "Neural Networks and Deep Learning" - it should be the first link that shows up. Understanding in statistics and calculus may be recommended but not necessary. Machine learning is a subject which can only be learned with months of study, if not years.

                            Hello Osmannica2001:

                            Thanks for your info!

                            Although ,still,I am not getting prizes of 4 or + sucesses,I am happy with my little prizes.

                            To get good settings perhaps needs a lot of" trial and error".

                            Thanks and regardsSmileThumbs Up

                              MillionsWanted's avatar - 24Qa6LT

                              Norway
                              Member #9517
                              December 10, 2004
                              1272 Posts
                              Online
                              Posted: February 6, 2016, 5:48 pm - IP Logged

                              I've been working new ways to play using Percep-Lotron and will do a bunch of new back tests on KENO.

                              The method I am going to test on KENO, will work on lotto Pick-5, Pick-6 and Pick 7 also.

                              I will tell you about the method later. Then you can do back tests on lotto if you want.

                              I will do a small real test on KENO with five bets/lines tomorrow Sunday.

                                 
                                Page 26 of 46