Welcome Guest
Log In | Register )
You last visited January 23, 2017, 7:51 pm
All times shown are
Eastern Time (GMT-5:00)

Neural Net Lottery Picker

700 replies. Last post 4 days ago by lottolot.

Page 9 of 47
511
PrintE-mailLink
paul762's avatar - lion

United Kingdom
Member #3002
December 11, 2003
478 Posts
Offline
Posted: September 4, 2015, 3:35 am - IP Logged

Finding the optimum settings for each game is the biggest problem Confused

Life's Tragedy is that we get old to soon and wise too late - Benjamin Franklin.
    Avatar
    Krakow
    Poland
    Member #86302
    February 2, 2010
    892 Posts
    Offline
    Posted: September 4, 2015, 7:17 am - IP Logged

    Finding the optimum settings for each game is the biggest problem Confused

    Good settings are everything in such smart random set generators. The best I've seen in 6/49 game was 1 x 4 of 6.

      Avatar
      Krakow
      Poland
      Member #86302
      February 2, 2010
      892 Posts
      Offline
      Posted: September 4, 2015, 7:23 am - IP Logged

      stoopendaal

      Once you've started to improve it, there's no stopping. You will see a never ending wish list of features to add to the program. I think what it really needs is to prove that the program can do better than random in some 20-50 plays. If it does then it's good and that's more than people should expect.

       

      Adam

        MillionsWanted's avatar - 24Qa6LT

        Norway
        Member #9517
        December 10, 2004
        1281 Posts
        Online
        Posted: September 4, 2015, 7:30 am - IP Logged

        stoopendaal

        Once you've started to improve it, there's no stopping. You will see a never ending wish list of features to add to the program. I think what it really needs is to prove that the program can do better than random in some 20-50 plays. If it does then it's good and that's more than people should expect.

         

        Adam

        Agree. At some point he should stop adding features,  and too many options in NN Settings will confuse users.

        He's doing a great job. Considering it's freeware we should be happy he want to continue working on NNLP.

          Avatar
          Krakow
          Poland
          Member #86302
          February 2, 2010
          892 Posts
          Offline
          Posted: September 4, 2015, 8:04 am - IP Logged

          I've found these settings to be the best for 6/49:

          settings

            stoopendaal's avatar - archer

            Netherlands
            Member #3476
            January 24, 2004
            212 Posts
            Offline
            Posted: September 5, 2015, 4:55 am - IP Logged

            Some of Millionswanted requested features are added to the application:

            • stop training at x picks found (Page 'Generate')
            • play a signal when a pick has been found (optional) (Page 'Generate')

            Also false generated picks (missing number(s)) are filtered out by the application (will not be listed).

             

            @goodlukman: "I'm really anticipating the epoch addition. This will put your program model in line with similar programs. Keep up the great work."

             I just did some tests with high and low epochs and you're right, the generated picks were terrible Smile

            @lottolot: "..numbers and not only combinations predicted,would be a great feature!" Yes, I'm working on another ANN tool using lottery number statistics as input.

            For each number a ranking value (probability value) is calculated (as output of the ANN tool) so you are now able to select the top x ranking numbers as output.

            I did some tests and the results were very impressive...of course with te right settings Smile. I can't say when I'll release this tool, maybe I'll add it to NNLP on a separate page.

            Version 2.1 is available for download.

             

            Good luck!

              MillionsWanted's avatar - 24Qa6LT

              Norway
              Member #9517
              December 10, 2004
              1281 Posts
              Online
              Posted: September 5, 2015, 7:34 am - IP Logged

              Some of Millionswanted requested features are added to the application:

              • stop training at x picks found (Page 'Generate')
              • play a signal when a pick has been found (optional) (Page 'Generate')

              Also false generated picks (missing number(s)) are filtered out by the application (will not be listed).

               

              @goodlukman: "I'm really anticipating the epoch addition. This will put your program model in line with similar programs. Keep up the great work."

               I just did some tests with high and low epochs and you're right, the generated picks were terrible Smile

              @lottolot: "..numbers and not only combinations predicted,would be a great feature!" Yes, I'm working on another ANN tool using lottery number statistics as input.

              For each number a ranking value (probability value) is calculated (as output of the ANN tool) so you are now able to select the top x ranking numbers as output.

              I did some tests and the results were very impressive...of course with te right settings Smile. I can't say when I'll release this tool, maybe I'll add it to NNLP on a separate page.

              Version 2.1 is available for download.

               

              Good luck!

              Thanks!

                MillionsWanted's avatar - 24Qa6LT

                Norway
                Member #9517
                December 10, 2004
                1281 Posts
                Online
                Posted: September 5, 2015, 8:51 am - IP Logged

                Just noticed: I'm no longer able to set Threshold/Error at 0.01. Instead I must set it to 0.010048.

                Not sure if there are any practical differences.

                  Elizabeth03's avatar - cat anm.gif
                  Nova Scotia
                  Canada
                  Member #9934
                  December 27, 2004
                  891 Posts
                  Offline
                  Posted: September 5, 2015, 2:25 pm - IP Logged

                  stoopendaal, when i've tried generating new numbers, to try for the lotto 649, for Canada, it won't give me new number's, just the same old one's?

                    MillionsWanted's avatar - 24Qa6LT

                    Norway
                    Member #9517
                    December 10, 2004
                    1281 Posts
                    Online
                    Posted: September 5, 2015, 5:41 pm - IP Logged

                    I was playing Saturday lotto with 10 lines which was predicted with a history result list where the numbers was placed in draw order(13 draws).

                    I won a 4/7 prize which gave me my money back. Smile

                    I also played 40 lines on Saturday lotto with a "normal" 42 draws history results list and won nothing.


                    I'm starting to suspect fewer draws are better. I've run back tests with 30, 40 and 50 draws. 30  to 40 is usually better.

                    Using a history results list with 10-30 draws might be even better. I'm going to run new back tests with 15 - 30 draws in history.

                    Type Thud


                      United States
                      Member #168172
                      August 19, 2015
                      18 Posts
                      Offline
                      Posted: September 5, 2015, 6:15 pm - IP Logged

                      I was playing Saturday lotto with 10 lines which was predicted with a history result list where the numbers was placed in draw order(13 draws).

                      I won a 4/7 prize which gave me my money back. Smile

                      I also played 40 lines on Saturday lotto with a "normal" 42 draws history results list and won nothing.


                      I'm starting to suspect fewer draws are better. I've run back tests with 30, 40 and 50 draws. 30  to 40 is usually better.

                      Using a history results list with 10-30 draws might be even better. I'm going to run new back tests with 15 - 30 draws in history.

                      Type Thud

                      I get pretty good results with 50 draws.

                        lottolot's avatar - Sphere animated2.gif
                        Tanhauser Gates
                        Holy See (Vatican City State)
                        Member #139281
                        February 18, 2013
                        192 Posts
                        Online
                        Posted: September 5, 2015, 7:28 pm - IP Logged

                        LOLI Agree!Thumbs UpBang Head

                        Regards

                          lottolot's avatar - Sphere animated2.gif
                          Tanhauser Gates
                          Holy See (Vatican City State)
                          Member #139281
                          February 18, 2013
                          192 Posts
                          Online
                          Posted: September 5, 2015, 7:35 pm - IP Logged

                          Some of Millionswanted requested features are added to the application:

                          • stop training at x picks found (Page 'Generate')
                          • play a signal when a pick has been found (optional) (Page 'Generate')

                          Also false generated picks (missing number(s)) are filtered out by the application (will not be listed).

                           

                          @goodlukman: "I'm really anticipating the epoch addition. This will put your program model in line with similar programs. Keep up the great work."

                           I just did some tests with high and low epochs and you're right, the generated picks were terrible Smile

                          @lottolot: "..numbers and not only combinations predicted,would be a great feature!" Yes, I'm working on another ANN tool using lottery number statistics as input.

                          For each number a ranking value (probability value) is calculated (as output of the ANN tool) so you are now able to select the top x ranking numbers as output.

                          I did some tests and the results were very impressive...of course with te right settings Smile. I can't say when I'll release this tool, maybe I'll add it to NNLP on a separate page.

                          Version 2.1 is available for download.

                           

                          Good luck!

                          Thanks!!!WinkThumbs Up

                          Regards

                            stoopendaal's avatar - archer

                            Netherlands
                            Member #3476
                            January 24, 2004
                            212 Posts
                            Offline
                            Posted: September 6, 2015, 7:41 am - IP Logged

                            Elizabeth03,

                             

                            "stoopendaal, when i've tried generating new numbers, to try for the lotto 649, for Canada, it won't give me new number's,

                            just the same old one's?"

                            • How many draw results of your game did you load into the tool?
                            • Can you give me an example of the draw results you use?
                              stoopendaal's avatar - archer

                              Netherlands
                              Member #3476
                              January 24, 2004
                              212 Posts
                              Offline
                              Posted: September 6, 2015, 7:47 am - IP Logged

                              MillionsWanted,

                              "Just noticed: I'm no longer able to set Threshold/Error at 0.01. Instead I must set it to 0.010048. "

                              I changed the range of  threshold slightly  so now the lowest value can be set to 0. This change causes the different threshold value with the previous version of te tool. I'll try to fix this in the next release of the tool.

                                 
                                Page 9 of 47