Common Appearance and other filters

Discussion on strategies that can be developed within Lotto Architect
Post Reply
Site Admin
Posts: 112
Joined: Thu Jan 17, 2008 4:22 am
Location: Melbourne, Australia

Common Appearance and other filters

Post by draughtsman » Sun Jan 27, 2008 5:40 am


The following describes how to remove previous draws from your tickets.
You'll use "Common appearances" rejection filter and some draw filters for this.

How to filter past draws
"Common appearances" filter displays directly comparisons up to the last 250 past draws. If you want to filter out any previous draw up to 250 depth, you have to create all filters displayed in "Common appearances" category. Each of those filters should be set to "Yes" for all values 0-x and "No" for the maximum possible value. So, if your game is 6/49, then values 0-5 should be set to "Yes" and value 6 to "No" for all columns in order to filter past draws up to 250 backwards. Then, you have to create these filters and store them in a session. To filter those draws, proceed to stage 3 calculations and use that session you've stored these filters in. Note that if you save these filters in a project, they can be used as is (no modification required) for any future draws. So this is a do once task. Note that the draw filter must be "by 1" for this system to properly filter out the last 250 draws.

Expand filtering of past draws beyond 250
The procedure is the same as above. You have to make "Common appearances" rejection filter to compare draws beyond 250. To do so, create a draw filter with parameters "Pattern=1, Skip=250" and use it with "Common appearances" filter. This will effectively compare draws drawn 251-500 times ago. Create those filters as described above.
To remove tickets drawn 501-750 draws before, create a draw filter with "Pattern=1, Skip=500" and so on and use it in "Common appearances".

How many past draws to remove?
We'll answer to this question using classic probability.
The total combinations for a game eg 6/49 is 49C6=13983816.
The chance to have a draw to be the same with the previous one is 1/13983816=0.000007% approx.
This rate is rather low of course and this is why we want to remove previous draws! So, how many past draws to remove?
- 1) Determine the total combinations for your lottery, here A=49C6=13983816
- 2) Determine the % of the error you want to accept. eg E=0.1=10%
- 3) If X is the past draws that can removed to have at most E% error, then X=E*A

Example for 6/49 game: We want error E=1%=0.01 at most. Then X=0.01*13983816=139838 past draws.

Lotto Architect,

Many thanks for this advice on filtering related to past draws and particularly the statistics in respect of the chance of a previous drawn number set appearing again. One can really, with quite some confidence, filter out these previously drawn number sets from your big wheel should they be occurring in your wheel. To set the various filters to Yes and No as you have suggested I understand that I must set the Algorithm to Custom/User and then proceed to set all filters in the table, the 250 horizontal cells representing the last 250 draws and the vertical cells in each column as per your advice of the last vertical cell set to no and all the previous vertical cells set to Yes. Is it necessary to 'lock' these cells that I set?

Thanks for this. I am beginning to see what Intelligent Filtering is all about. No point spending money on tickets where the numbers have already appeared in previous draws


Quite right relowe!
Assume your game is 6/49. Then, each column in the "Common appearances" filter should look like this:

Value 0 : Yes
Value 1 : Yes
Value 2 : Yes
Value 3 : Yes
Value 4 : Yes
Value 5 : Yes
Value 6 : No

You have to set Algorithm to "Custom/User" to manually set these values. No algorithm can be used to set these values to this formation. I plan to add a specific filter in a future version to filter past draws without the need to create all those filters.
Lock status is not necessary if you don't plan to further analyse those filters.

The case for Pick 3 games is different: we have only 1000 combinations in total and having E=1%=0.01 error at most we can remove only 1000*0.01=10 past draws!

Lotto Architect,

Seemed like a big task but it was achieved quite quickly. Highlighted all the horizontal cells as in a spreadsheet, brought up the menu under the mouse and converted all the last row to No in one go. Did likewise then in saving the filters to a Session - with the only real effort being hitting the enter key some 500 times!

To prove the point of this filter to myself I ran the filter set over the entire ticket set of the lottery and then viewed the rejected tickets. And there they were all the number sets equivalent to the previous draws. Quite clever in my opinion.

Can I ask what happens if you set one of the other values to No ie say all of line 3 to No? Is there any definite group of number sets that will be eliminated? They obviously will bear some relationship to the drawn numbers.



Filtering past draws is one use of "Common appearances" rejection filter.

You probably will not create many "Common appearances" filters (except for filtering a lot of past draws as already described).

The major use of "Common appearances" is to find relationships between the next draw to come and all its previous draws. For the following, assume draw filter is "by 1".
What each column show? Lets see the column "1". This shows how many common numbers found comparing each draw in the history with the previous one. Of course, Draw #1 does not have a previous draw!

So, if Value 0 says eg 500 (Statistics+values formation), this means that 500 draws found in the history that have no common numbers (Value 0) with their previous draw.
If Value 1 says eg 300, this means that 300 draws found in the history that have exactly 1 common number with the previous draw and so on.
If you see column "2", this compare each draw with the draw before the previous draw.

So, four your question on Value 3, if you set it to No only, then the program will remove any tickets that have exactly 3 common numbers with each of the previous draws.
To get a better understanding of the process, I'll use an example. You have set your filter (Column "1", reject value 3 only=No, the rest of values=Yes).
The last draw in the history is rg 03 05 17 30 31 38 (6/49 game).
When the program test a ticket, eg 01 03 18 30 33 38, here you can see that this ticket has exactly 3 common numbers with the previous draw. So, the effect of your created filter will be to reject that ticket.
Use the statistics and the features in the Rejection Filters window to determine what values are good to reject for each displayed column. This requires some effort of course but you'll remove a lot of tickets "safely" enough. You can use the algorithms too, they provide good success ratios with "Common appearances".

Based on the above, you probably understand why when we set Max value to No only, we effectively filter past draws.

Lotto Architect,

If I may continue the thinking on Common Appearances filter - when I set the Data Values Table to 'Values' and Table Information to 'Statistics' I can then see in the table the actual occurrences of each event ie the number of times previous draw number sets have appeared in future draws. It is quite fascinating to see (at least in the lottery I am observing) that there are virtually no occurrences of previous draw number sets occurring beyond a match =3 (there are some at match =4 perhaps 30 single occurrences in 500 draws) but nothing beyond that. Thus I believe I can use the power of this filter to do a very clean sweep in filtering by setting the values 0 to 3 to Yes and 4,5,6 to No. This I presume will reject every set beyond a match = 3 but retain everything up to and including a match =3.

I presume it also possible to examine other filters in the same way ie observe Statistics and Values and so set cells to Yes and No accordingly, although I am now a bit intrigued with Expectation and Delays and the colour scheme. If you are able to elaborate on these points it would be appreciated.



The matter of filtering is more complex than this. I'll list some theoretical thoughts here to better understand the complexity of the problem and some advice for good filtering:

Each value in your lottery has a chance to appear. The less, the easier to remove this value confidentially. For example, a value that has a hit rate 1:1000 draws or above, can be "safely" removed. Or values that never appeared in the history before can be "safely" removed" too.

Right? No!

The BIG problem is that this value will possibly appear in the future (it has a chance although very small). We wish it will never appear but if you check it out, you'll find that if you set eg 100 rejection filters, 95-98 of them will be set correct and the future draw will return 1-2 "irregular" values which you've filtered out because of this very low chance; so you lose the major prize .
The problem is even bigger because there is no way to find out what these filters might be in advance (and obviously don't use them!).

The above illustrates why it is wrong to filter only by observing the occurrences (statistics) of a value.

What can be done? The program includes the "Expectation" and delays feature to have a better view of values that are due to come. Obviously, values that are due, should be selected. Can we remove "safely" values that are not due (they drawn eg within the last draws)? The answer is "Yes" and "No". Surely, it provides a better guidance of what to "remove" or not.

A good solution is to combine both statistics & delays. Values that are "cold" (displayed as blue at the delays coloring scheme) that have a low chance to appear are the best values to remove. This is because statistically we do not expect them to appear soon due to their cold delay status and they are statistically infrequent values. Also good values to remove are those that have a cold delay and they have been appeared much more than statistically expected at a short range of the last draws (eg 30-50; use last draws x for this).

So, regarding your "Common appearances" approach to filter all Match=4,5,6 although it looks obvious to do so, the above explanation tells you why you'll remove the major prize too! Try to use my suggestion on those filters that provide the suggested properties: cold delay & statistics

Lotto Architect,

I hope I don't have too many questions!! But using the program offers so many options or methods which are worth exploring.

If I work through a filter using the technique you describe above where I set cell values to yes and /or no according to expectation and even statistical value, does this then in effect mean that the filter has been 'optimised' and will do all the work it is capable of doing with the setup you have designed into it? By this I mean is it then also necessary to work through the other algorithms that can be used in conjunction with the filter to develop more filter approaches? So does the above hand crafted yes/no setup simply cause the most optimum filtering and so it is not then necessary to work through the algorithms?

thanks again


First, questions are welcome as this enables understanding of the system better. Keep in mind that this program is designed to allow deep search on every property of a lottery game; the ultimate goal is to provide the best possible and accurate filtering of course! This is to contrast almost all other programs that provide an "this is only" approach with limited flexibility. The forum serves the purpose to explain "how-to-do" advanced analysis.

Now, to your question:
Algorithms provide an additional edge to your analysis. Their effect cannot be performed by the user (eg Delays/Differences algorithms) as they go deeply into analysis of statistical data, which the user cannot do by himself.

If you want to combine the approach of delays & statistical analysis described in previous posts, you have to follow the next guidelines:
1) Determine what algorithms to use. Use the report feature to do so. You can create additional algorithms to further analyse the data. The aim is an algorithm to reject as many values as possible and still maintain high hit ratios.
2) Now, go through all your good algorithms. For each one, Lock the values it has rejected. This is needed because if you choose another algorithm, it will erase the rejected values of your previous algorithm. We assume here that if your algorithm has rejected a value, it has a good reason to do so which means we want to keep this value rejected, no matter what the other algorithms might suggest.
3) When all your good algorithms used, we can proceed to our manual approach (delays & statistical analysis) for those values that have not been locked (as all your locked values have already been rejected). You may want to Lock your manually rejected values too if you plan to further analyse your designed filter.
4) Perform any other sort of analysis you want to use (eg observation of rundown statistical data). Don't forget to lock the values you want as if you use an algorithm again, it will affect all non-locked values.

The above procedure is a complete analysis on a rejection filter. This means, you have combined all algorithms and user analysis in one filter only. There is no need to create additional filters of the same type.

Disadvantage: You have to do the same procedure from the beginning for every new draw to come.

Solution: create several filters of the same type; each one will perform a part of the above analysis. The final results will be exactly the same as the above procedure and they are more easy to handle. Also, you have the benefit to disable a filter if for some reason it does not perform as expected and still use the others. Also, you don't need to lock values now.

Post Reply

Who is online

Users browsing this forum: No registered users and 4 guests