Welcome Guest
Log In | Register )
You last visited January 16, 2017, 9:40 pm
All times shown are
Eastern Time (GMT-5:00)

Backtesting and Simulating Lottery Systems

Topic closed. 136 replies. Last post 6 years ago by RL-RANDOMLOGIC.

Page 1 of 10
4.73
PrintE-mailLink

United States
Member #93947
July 10, 2010
2180 Posts
Offline
Posted: April 5, 2011, 5:34 pm - IP Logged

Most active posters on topics related to systems tend to reject rigorous computer backtesting of lottery systems, for the most part, because they think the "human element" is too complex to be modeled in a computer program.  Likewise, suggestions of using computer simulations to augment backtesting efforts tend to be rejected for the same reasons.  In addition, there is significant suspicion among system proponents around the concept of randomness in general, and Random Number Generators in particular.  For this reason, I suggest the following two recently initiated Topics for anyone who shares the skepticism described here.  They are taking a different approach to systems in these threads, one which you might be more comfortable with.

http://www.lotterypost.com/thread/229928

http://www.lotterypost.com/thread/229672

==========================

Here are two articles that illustrate how RNGs have been invaluable in solving some of the worlds toughest problems in Mathematics, Physics, Operations Research, etc.  Based on your responses to these, we can decide where to go from here.

--Jimmy4164

==========================

"Monte Carlo methods are especially useful for simulating systems with many coupled degrees of freedom, such as fluids, disordered materials, strongly coupled solids, and cellular structures (see cellular Potts model). They are used to model phenomena with significant uncertainty in inputs, such as the calculation of risk in business. They are widely used in mathematics, for example to evaluate multidimensional definite integrals with complicated boundary conditions. When Monte Carlo simulations have been applied in space exploration and oil exploration, their predictions of failures, cost overruns and schedule overruns are routinely better than human intuition or alternative "soft" methods.[2]"
 
"You can develop models that capture detailed information about unlikely or worst-case scenarios or obtain approximate solutions to problems that are otherwise intractable or time-consuming to analyze with traditional analytical techniques. Supported capabilities include a wide range of random and quasi-random number generators, parallel computing enabled random number generators, Markov Chain Monte Carlo simulation, and simulation of stochastic differential equations. Financial engineers and actuarial scientists use these capabilities for:
  • Incorporating uncertainty into existing models
  • Modeling interest rates
  • Pricing and valuation of stocks, bonds, options, and derivatives
  • Quantifying operational, market, or credit risk
  • Valuing financial projects, structured products, and real options
  • Assessing insurance and re-insurance risks and value
  • Evaluating financial plans and perform what-if studies"
-------------------------
    RJOh's avatar - chipmunk
    mid-Ohio
    United States
    Member #9
    March 24, 2001
    19892 Posts
    Offline
    Posted: April 6, 2011, 6:40 am - IP Logged

    I've read posts of members claiming to have back-tested pick3 systems but they are only dealing with 1000 possible straight or 220 box combinations.  Picking a winning combination in 10-100 tries might be possible with so few combinations but impossible to do with a system designed for pick5s or jackpot games. 

    Even a system designed for a 5/39 game with only 575,757 possible combinations and 5,781 winning combinations would probably take thousands of tries to pick one of the winning combinations in spite of knowing the five winning numbers. With these systems the best one can do is check past drawings, design the system accordingly and hope their predictions are right for a future drawing. Even a simulation can only be tested on future drawings, since you know the winning numbers you're are likely create a simulation that can't win.

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

      Avatar

      United States
      Member #105312
      January 29, 2011
      435 Posts
      Offline
      Posted: April 6, 2011, 7:48 am - IP Logged

      I've read posts of members claiming to have back-tested pick3 systems but they are only dealing with 1000 possible straight or 220 box combinations.  Picking a winning combination in 10-100 tries might be possible with so few combinations but impossible to do with a system designed for pick5s or jackpot games. 

      Even a system designed for a 5/39 game with only 575,757 possible combinations and 5,781 winning combinations would probably take thousands of tries to pick one of the winning combinations in spite of knowing the five winning numbers. With these systems the best one can do is check past drawings, design the system accordingly and hope their predictions are right for a future drawing. Even a simulation can only be tested on future drawings, since you know the winning numbers you're are likely create a simulation that can't win.

      RJOH:  That last sentence ought to be a sticky headline.

        Avatar

        United States
        Member #105312
        January 29, 2011
        435 Posts
        Offline
        Posted: April 6, 2011, 8:46 am - IP Logged

        A person might partially get around the problem by 'backing into' it.  Using persistent or strongly negative test results as a target for analysis in hopes of gleaning whatever it tells about the tests, the system and the possible directions for modifications leading to something to test next.

          garyo1954's avatar - garyo
          Dallas, Texas
          United States
          Member #4549
          May 2, 2004
          1823 Posts
          Offline
          Posted: April 6, 2011, 2:14 pm - IP Logged

          I've read posts of members claiming to have back-tested pick3 systems but they are only dealing with 1000 possible straight or 220 box combinations.  Picking a winning combination in 10-100 tries might be possible with so few combinations but impossible to do with a system designed for pick5s or jackpot games. 

          Even a system designed for a 5/39 game with only 575,757 possible combinations and 5,781 winning combinations would probably take thousands of tries to pick one of the winning combinations in spite of knowing the five winning numbers. With these systems the best one can do is check past drawings, design the system accordingly and hope their predictions are right for a future drawing. Even a simulation can only be tested on future drawings, since you know the winning numbers you're are likely create a simulation that can't win.

          RJOH,

          You hit the nail on the head shooting from the hip! Kudos!

          Monte Carlo simluations are not designed to test the likelihood of past events.

          Monte Carlo Simulations test the the life of your washing machine, the possibility of earthquake, the breakdown temperature of the oil in your vehicle, the potential for radiation contamination when your nuclear reactor fails, future valuations on investment portfolios, the growth potential for business, etc.

          It does so reasonably well using historical data and current trends and/or known factors to establish some baseline characteristics by which it proceeds to 'know the future.' Simple example, I want to know what from what low sum to what high I need to play to hit the Pick 3 say, 80% to 85% of the time. A Monte Carlo Simlatution says:

            garyo1954's avatar - garyo
            Dallas, Texas
            United States
            Member #4549
            May 2, 2004
            1823 Posts
            Offline
            Posted: April 6, 2011, 2:28 pm - IP Logged

            The chart above should work for any Pick 3 game in the world within +/-3% depending on the historical data available. Those numbers were derived over 5000 draws. Not to say a game with less won't show comprarable results, but it might take a a time to level out to these results. 

            Now, we could also call this a backtest if we derived those numbers from a specific game. And we could write a much more complicated test that produced not only the sums we had to play, but the combinations, the amount we would have to spend, the projected earning, the actual earnings, etc.

            However, no matter how complicated we make this program there is no way it could make the necessary decisions we might make at some point in the past. 

            Short example, if one used it to play the Mega-Millions challenge as Jimmy outlined the program would run fine. Yet where the program would continue to run in any instance, in reality we would rather quickly decide to change strategy when we found ourselves starving to death and homeless in the freezing rain along the Ohio river because we were spending (what was it?) $12,400 a month buying lottery tickets. 

            Monte Carlo doesn't care. Monty Hall doesn't care. Monty Python would laugh.

            Of course, we're not all named Monte.


              United States
              Member #93947
              July 10, 2010
              2180 Posts
              Offline
              Posted: April 6, 2011, 4:57 pm - IP Logged

              RJOH:  That last sentence ought to be a sticky headline.

              Don't be fearful, Joey.  Any simulation I propose in this thread will include the source code for your scrutiny.

                garyo1954's avatar - garyo
                Dallas, Texas
                United States
                Member #4549
                May 2, 2004
                1823 Posts
                Offline
                Posted: April 6, 2011, 6:05 pm - IP Logged

                Don't be fearful, Joey.  Any simulation I propose in this thread will include the source code for your scrutiny.

                I'm having Dick Dastardly moments here.


                  United States
                  Member #93947
                  July 10, 2010
                  2180 Posts
                  Offline
                  Posted: April 7, 2011, 1:50 am - IP Logged

                  I've read posts of members claiming to have back-tested pick3 systems but they are only dealing with 1000 possible straight or 220 box combinations.  Picking a winning combination in 10-100 tries might be possible with so few combinations but impossible to do with a system designed for pick5s or jackpot games. 

                  Even a system designed for a 5/39 game with only 575,757 possible combinations and 5,781 winning combinations would probably take thousands of tries to pick one of the winning combinations in spite of knowing the five winning numbers. With these systems the best one can do is check past drawings, design the system accordingly and hope their predictions are right for a future drawing. Even a simulation can only be tested on future drawings, since you know the winning numbers you're are likely create a simulation that can't win.

                  RJOh,

                  Thanks for your input RJ.  So far, yours is the only comment worthy of a reply.

                  My reason for posting the two articles above is primarily to lay some groundwork for a possible simulation of one or more aspects of a popular lottery system.  You're correct in alluding to the fact that many observations of draws and bet selections are required to pass judgement on a system playing games with as many possibilities as a (5,39) Lotto.  In most cases, the games don't have sufficient history to provide the necessary data.  This is where the beauty of Monte Carlo Methods lies!  To be convinced that these techniques can accomplish what I see you are skeptical of, you must first become confident that a random number generator can be designed to select sets of [in our example] 5 numbers that, for all practical purposes, are indistinguishable from the sets emerging from flawlessly functioning mechanical ball machines.  Once you can effectively generate sets like these, you are then free to test your system over hundreds, if not thousands of years, even though the actual game may have only been around for a decade or less.  Please have another look at the articles above before giving up on the idea.

                  Also, did you take note of the $30,000 Roulette Challenge I posted over in Gaming?  Be sure to read that through to the end.  It deals with these issues quite well.  In the meantime, I'm working on some computer code that we will hopefully be able to turn loose here soon.  Anyone who had a problem with the 3,168 Powerball ticket purchases in the 2010 MadDog Powerball Challenge should really, REALLY, read this roulette challenge!

                  http://vegasclick.com/gambling/betting-system-challenge.html

                  --Jimmy4164


                    RJOh's avatar - chipmunk
                    mid-Ohio
                    United States
                    Member #9
                    March 24, 2001
                    19892 Posts
                    Offline
                    Posted: April 7, 2011, 12:29 pm - IP Logged

                    RJOh,

                    Thanks for your input RJ.  So far, yours is the only comment worthy of a reply.

                    My reason for posting the two articles above is primarily to lay some groundwork for a possible simulation of one or more aspects of a popular lottery system.  You're correct in alluding to the fact that many observations of draws and bet selections are required to pass judgement on a system playing games with as many possibilities as a (5,39) Lotto.  In most cases, the games don't have sufficient history to provide the necessary data.  This is where the beauty of Monte Carlo Methods lies!  To be convinced that these techniques can accomplish what I see you are skeptical of, you must first become confident that a random number generator can be designed to select sets of [in our example] 5 numbers that, for all practical purposes, are indistinguishable from the sets emerging from flawlessly functioning mechanical ball machines.  Once you can effectively generate sets like these, you are then free to test your system over hundreds, if not thousands of years, even though the actual game may have only been around for a decade or less.  Please have another look at the articles above before giving up on the idea.

                    Also, did you take note of the $30,000 Roulette Challenge I posted over in Gaming?  Be sure to read that through to the end.  It deals with these issues quite well.  In the meantime, I'm working on some computer code that we will hopefully be able to turn loose here soon.  Anyone who had a problem with the 3,168 Powerball ticket purchases in the 2010 MadDog Powerball Challenge should really, REALLY, read this roulette challenge!

                    http://vegasclick.com/gambling/betting-system-challenge.html

                    --Jimmy4164


                    I've tracked Ohio Rolling Cash5(5/39) since its beginings when it replaced Buckeye5(5/37) which has had 2235 drawings so far but as a daily game I now have too much information to process before there's another drawing.  I prefer the games that have 2-3 drawings a week which gives me a couple of days to process the latest data before playing.

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


                      United States
                      Member #93947
                      July 10, 2010
                      2180 Posts
                      Offline
                      Posted: April 8, 2011, 12:19 am - IP Logged

                      I've tracked Ohio Rolling Cash5(5/39) since its beginings when it replaced Buckeye5(5/37) which has had 2235 drawings so far but as a daily game I now have too much information to process before there's another drawing.  I prefer the games that have 2-3 drawings a week which gives me a couple of days to process the latest data before playing.

                      Your post here is not typical of you, RJ.

                      Given my post that you commented on, it leaves me nonplussed...

                        RJOh's avatar - chipmunk
                        mid-Ohio
                        United States
                        Member #9
                        March 24, 2001
                        19892 Posts
                        Offline
                        Posted: April 8, 2011, 8:25 am - IP Logged

                        Your post here is not typical of you, RJ.

                        Given my post that you commented on, it leaves me nonplussed...

                        You shouldn't have been if you noticed my posts on the prediction board are always for one drawing.  I've never been the type of player to pick numbers and stick with them for any period of time, I prefer to evaluate all the data up to the most recent drawing and make my picks for the next drawing.

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

                          Coin Toss's avatar - shape barbed.jpg
                          Zeta Reticuli Star System
                          United States
                          Member #30470
                          January 17, 2006
                          10389 Posts
                          Offline
                          Posted: April 8, 2011, 10:31 am - IP Logged

                          jimmy4164,

                          "Most active posters on topics related to systems tend to reject rigorous computer backtesting of lottery systems, for the most part, because they think the "human element" is too complex to be modeled in a computer program."

                          There are those who come up with theories and systems and check them out and admit that they don't work, and then there are those who come up with theories and systems and if the results prove them wrong they ignore the results.

                          Those who run the lotteries love it when players look for consistency in something that's designed not to have any.

                          Lep

                          There is one and only one 'proven' system, and that is to book the action. No matter the game, let the players pick their own losers.


                            United States
                            Member #93947
                            July 10, 2010
                            2180 Posts
                            Offline
                            Posted: April 8, 2011, 11:13 am - IP Logged

                            jimmy4164,

                            "Most active posters on topics related to systems tend to reject rigorous computer backtesting of lottery systems, for the most part, because they think the "human element" is too complex to be modeled in a computer program."

                            There are those who come up with theories and systems and check them out and admit that they don't work, and then there are those who come up with theories and systems and if the results prove them wrong they ignore the results.

                            "There are those who come up with theories and systems and check them out and admit that they don't work, and then there are those who come up with theories and systems and if the results prove them wrong they ignore the results."  Coin Toss

                            I think you're absolutely right Coin Toss.  For these reasons I'm questioning whether it makes any sense for me to devote any more time to this.  I was planning to back-test and/or simulate ONE component of the "Digit System," namely the first filter applied which rejects all sets from the (5,39) matrix that have less than 5 or more than 6 unique digits.   The results showing that selecting from this reduced set of possibilities provides no statistically significant advantage to the player would probably be rejected for at least 3 reasons:  one, the general distrust of the use of RNGs in the analysis, crucial to Monte Carlo Methods; two, the leap of faith required (for many) to see that the result was useful in isolation from the other filters applied in practise; and, of course, the missing "human element."

                            Another way to state your observation is to say, "Anyone who would accept the results of a Monte Carlo simulation didn't need to see it in the first place, and all those who need to see the results, would not accept them!"

                            What do you think of that dude putting up $30K to challenge betting systems? Smile

                              RJOh's avatar - chipmunk
                              mid-Ohio
                              United States
                              Member #9
                              March 24, 2001
                              19892 Posts
                              Offline
                              Posted: April 8, 2011, 3:32 pm - IP Logged

                              "There are those who come up with theories and systems and check them out and admit that they don't work, and then there are those who come up with theories and systems and if the results prove them wrong they ignore the results."  Coin Toss

                              I think you're absolutely right Coin Toss.  For these reasons I'm questioning whether it makes any sense for me to devote any more time to this.  I was planning to back-test and/or simulate ONE component of the "Digit System," namely the first filter applied which rejects all sets from the (5,39) matrix that have less than 5 or more than 6 unique digits.   The results showing that selecting from this reduced set of possibilities provides no statistically significant advantage to the player would probably be rejected for at least 3 reasons:  one, the general distrust of the use of RNGs in the analysis, crucial to Monte Carlo Methods; two, the leap of faith required (for many) to see that the result was useful in isolation from the other filters applied in practise; and, of course, the missing "human element."

                              Another way to state your observation is to say, "Anyone who would accept the results of a Monte Carlo simulation didn't need to see it in the first place, and all those who need to see the results, would not accept them!"

                              What do you think of that dude putting up $30K to challenge betting systems? Smile

                              I was planning to back-test and/or simulate ONE component of the "Digit System," namely the first filter applied which rejects all sets from the (5,39) matrix that have less than 5 or more than 6 unique digits. 

                              No need to run a simulation to get those figures, it's part of my regular routine to pick parameters for my predictions.  This is part of the results for Ohio Rolling Cash5(5/39) with 2236 drawings so far. I use 95% for my parameters.  I can create a text file with all the information but I can only load 800 lines to sort and analyze.  I would have to rewrite the program to use virtual memory since my programs are in GWBasic, which I may do sometime in the future or learn a new Basic and rewrite all my programs so they will work in Window7.  Years ago when I wrote this program as a utility, 800 lines were plenty and still is for the other games I play. They have more than 800 total records but their latest matrix have less than 800 drawings.

                              Different Digits
                                0: 0     10: 0
                                1: 0     11: 0
                                2: 0     12: 0
                                3: 14    13: 0
                                4: 222   14: 0
                                5: 763   15: 0
                                6: 892   16: 0
                                7: 319   17: 0
                                8: 26    18: 0
                                9: 0     19: 0

                              95% DD = 4-7

                              Total Digits
                                0: 0     10: 552
                                1: 0     11: 0
                                2: 0     12: 0
                                3: 0     13: 0
                                4: 0     14: 0
                                5: 0     15: 0
                                6: 22    16: 0
                                7: 140   17: 0
                                8: 557   18: 0
                                9: 965   19: 0

                              95% TD = 7-10 

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

                                 
                                Page 1 of 10