History Differences filter

[Jerzy]

Stan,

I am experimenting with History differences filter. It is a new and very good filter, and I feel that its full potential for filtering out unwanted tickets and saving the jackpot ticket has not been developed yet.

In my experiments I use the values for default segments. I imported the segments history to Excel and obtained the statistical values, including median value, for the 1st parameter (Number count) of this filter. I am looking for possible correlations between various parts of data.

For example: There are the following segments for 14/07/07 (Australian Saturday Lotto - 6/45):

2-1-3, 3-1-2, 2-4-0, 4-1-1, 4-1-1, 4-0-2, 3-3-0, 3-1-2, 5-0-1, 4-1-1, 1-3-2

When I count the numbers in the first segment in all columns, in the 2nd segment in all columns, and finally in the 3rd segment in all columns I get:

Number count 1st segment all columns = 35
Number count 2nd segment all columns = 16
Number count 3rd segment all columns = 15

Next I connect those values to get Distribution of Count of Numbers in the segments: 35-16-15. That's what I am interested in.

It happens that 1 of 3 components of Distribution of Count of Numbers is repeated from the previous draw, but I have not noticed a drawing with all 3 components repeated at the same time.

I would like to filter out the tickets which result from the Distribution of number count 35-16-15 but it is impossible. Currently, we can filter out only one Distribution component separately, without linking it to other components.

We need triple (having three parts) History differences filter, with each part linked to other parts and working interdependently.

There are even more interesting findings when we examine Odd/Even Pattern of Distribution of Number Count. 35-16-15 becomes O-E-O pattern.

In the history of my game the following O/E Patterns of Distribution have the following occurrences:

E-E-E approximately 25%
E-O-O approximately 25%
O-E-O approximately 25%
O-O-E approximately 25%

It means that approximately 75% of Distribution of Number Count have one even and two odd components, although at different positions.
But if I want to filter out uncommon E-E-E patterns, I can't do it.

In the game where 6 numbers are drawn there are no patterns: O-O-O, O-E-E, E-O-E, and E-E-O. Creating such patterns is impossible. The maximal possible count of numbers for all columns is 66, which is an even number, and you can't get the even number by adding 3 odd numbers or adding 2 even and 1 odd number.

In the game where 5 numbers are drawn the maximal possible count of numbers is 55 and the situation with O/E patterns is reversed.

I feel that O/E Pattern of Distribution of Number Count filter looks attractive. The name of the filter can be shortened or changed.

What do you think, Stan, of the filters that I described, and about their implementation?

My best regards.

Jerzy

[Jerzy]

Stan,

Previously, I described the Distribution of Count of Numbers.

There is another finding. Statistical analysis of my 6/45 game (1151 drawings), using Excel, shows that the 1st component of the Distribution of Count of Numbers has always the gretest value of all, and the 3rd component has most commonly the smallest value.

However, in 16.52% of drawings the 2nd component has the smallest value. And in 5.72% of drawings the 2nd and 3rd components are equal.

Filtering out the Distribution of Count of Numbers which has the smallest value at the 2nd position, and/or filtering it if the values at the 2nd and 3rd positions are equal, may be a good bet.

Best regards.

Jerzy

[Joe]

Hi Jerzy,

Your idea is Good, and I think some what Its similar to my Idea that I have posted in "Nice to Have" thread http://www.expertlotto.com/en/forum/for ... age=3#4795

I Hope stan already has this idea in his "ToDo" list,

You can check my post with more ideas that I have been described in that Link, give me suggestions on it and expecting your views on it too,

thanks Jerzy and Good Luck,

[Jerzy]

Joe,

Yes, my suggestions are based on your idea of counting the occurrences of winning numbers across all 11 columns horizontally. History differences filter seems to be the first implementation of this idea. The filter is good but it has limitations which stimulated my thinking to try to find ways for further development.

I read all your posts and I will review them again. More thinking and experimentation is needed to evaluate everything related to segments and number counts. Probably in a couple of days I will be able to discuss some of the ideas presented by you.

Thank you. My best regards.

Jerzy

[Joe]

thanks Jerzy, Happy to see your post, I am really getting confidence on your supports and ideas, Like to see your ideas and should be worked out with different ways with segments, My statement is "Segments is powerful tool".
I am on your side. Good luck Jerzy

[scenium]

This is a good filter, but my thinking is that although we can specify the total number of the 1st segment, the segment combination is not fixed.

for eg.. 4-1-1...can be at any history difference position(0 -10 ). This will
still leave a lot of tickets.
Furthermore, my feeling is that, the current segment combination filter
will serve the same purpose.

[Jerzy]

Hi scenium,

All filters serve the same purpose: reduction of the package of tickets to the manageable and affordable size. And it is a common situation that some filters, when they are used together, overlap each other's action which decreases their expected effectiveness. There are correlations between filters. For example: Sum and Number Movement filter - when the numbers move, the sum changes.

When I use the existing Segments, Segments Combinations and History Differences filters, each of them applied lightly, one after another, I can see that each of them reduces the package. Since I avoid aggressive filtering total reduction of the package is moderate, and it is very likely that different segment filters can remove the same tickets.

Those filters are currently available and you can experiment with them.

The extensions of History differences filter that I propose would work differently, not only vertically in columns like existing filters, but most importantly horizontally, across the columns. As there are 11 columns, what happens in a single column has a relatively small effect on the Number Count obtained horizontally. And the statistical analysis of various situations correlated with Number Count and Distribution of Number Count indicates that the new extensions would work as effective filters.

In addition to the extension that I already described other improvements are possible.

For example, the Distribution of Number Counts in Segment is a three number concept, a triplet (35-16-15 or similar).

For the games where 6 numbers are drawn, the sum of three numbers is always 66, and some combinations of three numbers which give the sum of 66 (for example: 64-1-1) are extremely unlikely. For games with 5 numbers drawn, the sum of three numbers in the Distribution of Number Counts in Segment is always 55.

I believe a kind of triplet filter for Distribution of Number Count, to eliminate the uncommon triplets, should be the next consideration.

Best regards.

Jerzy

[Jerzy]

Joe,

I tested your idea of replacing the segment configurations with numbers to create other filters - starting with sum filters. I imported my game's segment data to Excel to estimate the results of using such filters.

It does not really matter, whether most commonly occurring segments are replaced with highest or lowest value numbers. Replacing the most common segments with highest values increases the sum values and shifts the curve up on the chart, but practically it does not affect filtering power of the filter. It would change the proportion of Low/ High numbers but I did not examine it.

The segment configuration 0-0-6 has never occurred in my game, so I used only the numbers from 1 to 27. I replaced the most common segments with higest numbers and the least common segments with lowest numbers.

The useful sum range, which I would leave while filtering, is 193 - 273(Median of 233 +/- 40). For this range, the curve on the sum chart is flat and almost horizontal, and covers 92.27 % of my game's past drawings.

I suggest creation of the filter for maximal occurrence of the same segment value. In my data occurrences from 2 to 4 are the most common and should not be filtered out.

Similarly, Standard Deviation filter for the segment values in all columns should leave the Standard Deviation range from 3.0090 to 9.9936

Regards

Jerzy

[Joe]

yeah, seems good,

let me check your Excel sheets again and get back to you,

Nice work Jerzy,

[stan]

[quote=Jerzy:1185342727]
I would like to filter out the tickets which result from the Distribution of number count 35-16-15 but it is impossible. Currently, we can filter out only one Distribution component separately, without linking it to other components.
[/quote]

why is that? just apply the history differences three times (number count 35-35, history differences 1-6 then number count 16-16, history differences 7-12 then number count 15-15, history differences 13-1000)

[Jerzy]

Stan,

Yes, applying the History Differences filter 3 times to filter out the selected number count values from 3 segments is the right way to do it.

I was thinking of a possibility of filtering three history differences ranges (segments) at the same time but I did not express myself clearly.

Sorry and thank you.

[Jerzy]

Stan,

Let me explain it again. I want to be able to remove only unique triplet combinations of number counts for three segments. And it cannot be achieved by applying the existing History Differences filter, no matter how you try, without removing some other triplets.

If I follow your advice, the History Differences filter will remove not only the unique triplet combination (35-16-15 was the example), but also many more triplets; also the triplets which I want to leave.

In addition to the triplet 35-16-15 many other triplets will be removed. For example the following and similar triplets:


34-17-15
33-18-15

36-16-14
34-16-16

So I was thinking of the possibility of filtering all three components of the number count triplets at the same time. However, I do not know whether implementation of it would be easy or possible.

They are interesting triplets. The sum of their components is always 66 for the 6/xx game.

Best regards

Jerzy

[stan]

[quote=Jerzy:1186052293]
Stan,

Let me explain it again. I want to be able to remove only unique triplet combinations of number counts for three segments. And it cannot be achieved by applying the existing History Differences filter, no matter how you try, without removing some other triplets.

If I follow your advice, the History Differences filter will remove not only the unique triplet combination (35-16-15 was the example), but also many more triplets; also the triplets which I want to leave.

In addition to the triplet 35-16-15 many other triplets will be removed. For example the following and similar triplets:


34-17-15
33-18-15

36-16-14
34-16-16

So I was thinking of the possibility of filtering all three components of the number count triplets at the same time. However, I do not know whether implementation of it would be easy or possible.

They are interesting triplets. The sum of their components is always 66 for the 6/xx game.

Best regards

Jerzy
[/quote]

it should work (unless i didn't understand what you're after)
the problem is that if you want to filter *out* a specific 'triplet' you need to apply history differences filter three times with 'leave' option selected and then removed the remaining tickets from the original package.
or you can setup a compound filter (3 times wnh differences filter with leave option connected with 'and') and then select matching tickets 'remove' but you'd have to edit this filter each time you pick a different 'triplet'

this might be a nice exercise for joe to create such filter plugin:)