Combinations & % Error Challenge

 Hi,

  Below are addition filtering methods available in commercial software packages. Quite a few of the following methods have been recycled constantly here at the LP ever since I joined. That is why my thinking is that there should be some standardized measure (ie- amount of combinations being reduce versus the amount of error being incurred.

Any thoughts?

Lottery Filter

Consecutives -  A filter that eliminate combinations with N consecutive numbers. For example, the combination 21,22,23,30,33,42 has 3 consecutive numbers - 21,22,23.

Evens - A filter that eliminate combinations with N even numbers. For example, the combination 20,22,24,26,33,43 has 4 even numbers - 20,22,24,26.

Odds- A filter that eliminate combinations with N odd numbers.
For example, the combination 1,3,11,19,30,40 has 4 odd numbers - 1,3,11,19.

Space - A filter that eliminate combinations with N equally spaced numbers. For example, the combination 5,10,15,20,23,28 has 4 equally spaced numbers - 5,10,15,20.

Span ( narrow ) - A filter that eliminate combinations with a span of N or lower. For example, the combination 1,2,5,9,11,14 has a span of 14-1=13. There for applying a filter of Span ( narrow ) of 13.

Span ( wide ) - A filter that eliminate combinations with a span of N or higher. For example, the combination 1,2,5,9,11,14 has a span of 14-1=13.

Sums ( low ) - A filter that eliminate combinations with a sum of N or lower. For example, the combination 1,2,5,9,11,14 has a sum of 1+2+5+9+11+14=42.

Sums ( high ) - A filter that eliminate combinations with a sum of N or higher. For example, the combination 1,2,5,9,11,14 has a span of 1+2+5+9+11+14=42.

Repeats - Use this filter to eliminate combination that contain numbers, which already came up in history draws.

1.      frequency, consistency,

2.      repeating, repeating vs. nonrepeating,

3.      proportions of numbers: high, low, odd, even,

4.      sums of winning combinations,

5.      skip draws, sums of skip draws,

6.      pair tables,

7.      unofficial results from the last 84 draws,

8.      hot, due, overdue numbers,

9.      hot trail, consistency #2,

Consecutive Filter - filters out lines with more than two or three consecutive numbers
Sum of Line Filter - filters out lines where the sum of all the numbers is over (or under) a certain number
Last Digit Filter - filters out lines according to their last digit
Width of Line Filter - filters lines according to the difference between the lowest/highest number in each line
Odd/Even - filters out lines with more/less odd numbers than even

• Sum [Total, Odd, Even, Small, Big, Congruence, Digit, Root]
• Sum Odd/Even Patterns [Total, Odd, Small, Big, Congruence, Digit, Root]
• Mass [Odd, Small, Alternation O/E, Common,Dif. Digits]
• First digit [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] [Jackpot games]
• Last digit [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
• Overall digits [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] [Equal to Last digit in Pick games]
• Difference [all possible differences]
• Consecutive [all possible consecutive ranges]
• Number position
• Pattern Odd/Even
• Pattern Low/High [Pick games only]
• Linear symmetry [all numbers in field]
• In/Out [all possible layouts]
• Middle-transverse/Out [all possible layouts]
• Horizontal overlap [all possible layouts]
• Vertical overlap [all possible layouts]
• Overall rows [all possible layouts]
• Overall columns [all possible layouts]
• [S] - congruence [all shapes ranging from 3 - 9]
• Congruence [9] / [8] / [7] / [6] / [5] / [4] / [3]

• First Digit
• Last digit
• Overall digits
• Difference
• Consecutive
• Linear symmetry
• In/Out
• Middle-Transverse/Out
• Horizontal overlap
• Vertical Overlap
• Overall rows
• Overall columns
• Congruence [3-9]

Unless you narrow the parameters of each filters, most of the combinations you generate will pass through and narrowing them down and running previous winners will eliminate most of them.

I try to set filter parameters so that 70% of the previous winners would have passed and once a combination is accepted, it is used to block other combinations that passed the filters but match three or more of the numbers in one of the combinations already accepted and I record them in a reject file.

While I have never eliminated a big winner because I have never generated a big winner, I have rejected some combinations that would have matched 3 or 4.  My reject files usually contain 500 or more combinations and my play file usually have 10-20 combinations.

@RJOh, have you ever changed the order the combos are filtered in?

I wrote in another thread, the p5 is very sensitive to the initial conditions. You can have the big winner in your list but if it's "too far down the list", you run out of budget before you get to it. Which (besides being very annoying) makes a very big difference in the profit/loss column.

@JKING, yes, the % of combos removed by those various filters can be calculated easily enough. You may find the totals to be disappointingly small, though. e.g. "consecutives" http://www.lotterypost.com/thread/173783/1017708

They make good filters because they don't provide very big targets.

@RJOh, have you ever changed the order the combos are filtered in?

I don't generate a bunch of combinations to be filtered later, so it makes no different.   I filter the combinations as they are generated and once a combination is accepted, it blocks any other similar combinations.  For example if the combination 1 2 3 4 5 makes it by the filter and is accepted then any other combinations with three or more of those numbers are rejected.  So a combination like 2 4 5 9 10 would be rejected and if the winning combination turns out to be 5 9 10 13 15 which was never generated, the chance to have matched 3 is in the reject file.

Unless at least 50% of the possible combinations are generated then the chances of generating a combinations with all the winning numbers are remote with my system.  The chances of generation a combination that could win one of the lower tier prize are better but if the total numbers of combinations accepted to play is limited to 20 or less then there's a good chance such combinations will be rejected too.

I could open up the parameters of the filters to accept more similar combinations but the limit of 20 combinations would be reached faster with no guarantee of a better results and I'm not interested in playing more than 20 lines.

@JKING: Are you working your system on paper or by code? That will make a big difference in how sophisticated you can work these filters.

My system uses a RNG to pick numbers which are compared with previous drawings before being checked using the filters parameters since I'm trying to avoid using combinations that have already won.

I recently added a feature that stops the program when ever it picks a combination that has all the winning numbers of a previous drawing and the program seldom stops.  This tells me that the odds of picking a winning combination to filter out is the same as picking a winning combination without filters, all filtering is going do is allow you to eliminate those combinations that you don't think look like a winner.  The odds of picking a winner will not change.

I tested this theory by using the LP RNG to pick some lines to post on the prediction page and they did a well as any I had picked using my program.

Hi,

 Yes time*treat, I write code.

 The point of calculating the combinations for each filter type is to select the best combination of filters. Basically getting the "Biggest Bang For Your Buck". That would be reducing the most combinations possible with the least amount of error. Sometimes I feel that I am the only one that understands that error is multiplicative. And that you can filter yourself out of winning anything if you don't pay attention to the amount of error in your system.

  So, can anyone say that they are using the best combinations of filters possible without having a standard measure to compare one filter type against another?

Getting the "Biggest Bang For Your Buck" depends on the game payouts too. I compared different games that I follow by calculating the amount I could expect to win if I played 10 lines during 100 drawings. For my calculations I used the winning numbers from the 10 previous drawings during the last 100 drawings.  I used these numbers because they were probably as random as any I could have picked.

OHIO CLASSIC LOTTO 6/49 TOTAL MATCHES

                       MATCH  0 = 441     
                       MATCH  1 = 388     
                       MATCH  2 = 152     
                       MATCH  3 = 18            $ 36
                       MATCH  4 = 1             $ 70
                       MATCH  5 = 0       
                       MATCH  6 = 0               0 JP
 
            TOTAL PRIZES = $ 106 + 0 JACKPOTS
     checked 100 records with 10 previous records = 1000

OHIO ROLLING CASH5 5/39 TOTAL MATCHES

                       MATCH  0 = 481     
                       MATCH  1 = 389     
                       MATCH  2 = 112           $ 112
                       MATCH  3 = 18            $ 180
                       MATCH  4 = 0       
                       MATCH  5 = 0               0 JP

            TOTAL PRIZES = $ 292 + 0 JACKPOTS
     checked 100 records with 10 previous records = 1000

MEGAMILLIONS 5/56+1/46  TOTAL MATCHES
                                         BONUS#
                       MATCH  0 = 631      15    $ 30
                       MATCH  1 = 310      6     $ 18
                       MATCH  2 = 53       2     $ 20
                       MATCH  3 = 6        0     $ 42
                       MATCH  4 = 0        0
                       MATCH  5 = 0        0

 
            TOTAL PRIZES = $ 110 + 0 JACKPOTS
     checked 100 records with 10 previous records = 1000

POWERBALL 5/55+1/42     TOTAL MATCHES
                                         BONUS#
                       MATCH  0 = 618      20    $ 60
                       MATCH  1 = 319      6     $ 24
                       MATCH  2 = 57       0
                       MATCH  3 = 6        0     $ 42
                       MATCH  4 = 0        0
                       MATCH  5 = 0        0

 
            TOTAL PRIZES = $ 126 + 0 JACKPOTS
     checked 100 records with 10 previous records = 1000

W. VA CASH25 6/25       TOTAL MATCHES

                       MATCH  0 = 159     
                       MATCH  1 = 399     
                       MATCH  2 = 326     
                       MATCH  3 = 101           $ 101
                       MATCH  4 = 15            $ 150
                       MATCH  5 = 0       
                       MATCH  6 = 0       
 
            TOTAL PRIZES = $ 251 + 0 JACKPOTS
     checked 100 records with 10 previous records = 1000

It appears the games with the lower potential jackpots give you more bangs for your bucks so if you going after the larger jackpots you've got to be willing to win less.  Winning a jackpot or 2nd prize is a random event that most hope to win but fail unless they get lucky.

The way I "grade" some of my filters is to see how they perform based on number of drawings back. I can select a block size of N drawings-per-block and X number-of-blocks and then see the percentage of success over N *X drawings, in groups of N. That tells me if a filter is any good and for how long. The result is a simple chart with multiples of N in col 1 and a % in col 2. I can't say I have the "best" combination of filters, but for certain ones, I have a standard way of comparing their effectiveness.

As for a bigger bang - if a method generates the winner say 20% of the time, but only creates 10% of the combos - I'm giving it a closer look.