Predicting Increase or Decrease in Columns

[sliver]

Hello all EL users,

I've spend ALOT of time trying to accurately predict which columns in winning numbers history will increase or decrease. From experience it's probably too much to hope to predict all 11 correct, so I've focused on developing a strategy that will predict as many as possible of the 11 correct.

I find that the simulator is not very relibale on this, and sometimes has up to 6 of the columns wrong. I have also tested the value predictor and have not been able to get any consistent results in predicting the increase or decrease of columns

Today I think I may have stumbled on a fairly good method, but it still needs to be tested and for now I can only say it will predict at the very least 7 of 11 columns correctly, and sometimes up to 9 of 11 columns correctly. In my lottery if I filter the full set of tickets for incr/decr and setting "7 to 11 columns must pass", I am able to reduce tickets by almost half so for me this is a good start.

I have to give credit to PadawanLotto, as this method is based on some strategy he posted earlier which I worked on and now, well it seems to have good results. So in summary the strategy is this -

NOTE: I'm working on the UK Daily Play 7/27 , so you will have to adapt this to other lotteries

1. I create a package of random numbers "max overlap 0" from the package generator. Each time I run this only 3 tickets are created. I run this several times and keep appending tickets until I got a package of about 24 tickets. I then run statistics on this package of 24 to ensure that the lottery numbers are somewhat evenly spread in occurence.If they are not, you may have to delete some tickets and generate new ones to achieve this. There's probably an easier way to do this, but this is the only way I know

2. With the newly created max overlap 0 package, I go to win history sums filter and run "lookup min/max ranges".

3. For each column I predict if it will increase or decrease based on which is the higher "absolute" min or max value. So if COl 0 shows min = -25 and max = 14 , then I predict col 0 will decrease. If Col 2 shows min = -15 and max = 18 then I predict Col 2 will increase. I do this for all columns.

4. Based on this method at least 7 of the columns will be predicted correctly and sometime up to 9 columns. I have also noticed that when there is a wrong prediction it is usually a "decrease" prediction that is wrong, not an "increase" prediction. The attached excel file shows how I have tracked this.

5. I have only done this with a package of max overlap 0, but I suspect that using max overlap 1,2,or 3 packages may also have some useful results, but then again maybe not.

Another bonus of this method is that the min/max range generated from the max overlap 0 package, seems to be the correct ranges for the winning numbers history sums filter for the full package! The range is wide but still reduces the full package considerably. Try this out too.

So I hope this makes sense, if not ask questions and I'll try and explain. I hope someone can help improve this method also.

Good luck folks

[Clem9403]

[quote=sliver]
Hello all EL users,

I've spend ALOT of time trying to accurately predict which columns in winning numbers history will increase or decrease. From experience it's probably too much to hope to predict all 11 correct, so I've focused on developing a strategy that will predict as many as possible of the 11 correct.

I find that the simulator is not very relibale on this, and sometimes has up to 6 of the columns wrong. I have also tested the value predictor and have not been able to get any consistent results in predicting the increase or decrease of columns

Today I think I may have stumbled on a fairly good method, but it still needs to be tested and for now I can only say it will predict at the very least 7 of 11 columns correctly, and sometimes up to 9 of 11 columns correctly. In my lottery if I filter the full set of tickets for incr/decr and setting "7 to 11 columns must pass", I am able to reduce tickets by almost half so for me this is a good start.

I have to give credit to PadawanLotto, as this method is based on some strategy he posted earlier which I worked on and now, well it seems to have good results. So in summary the strategy is this -

NOTE: I'm working on the UK Daily Play 7/27 , so you will have to adapt this to other lotteries

1. I create a package of random numbers "max overlap 0" from the package generator. Each time I run this only 3 tickets are created. I run this several times and keep appending tickets until I got a package of about 24 tickets. I then run statistics on this package of 24 to ensure that the lottery numbers are somewhat evenly spread in occurence.If they are not, you may have to delete some tickets and generate new ones to achieve this. There's probably an easier way to do this, but this is the only way I know

2. With the newly created max overlap 0 package, I go to win history sums filter and run "lookup min/max ranges".

3. For each column I predict if it will increase or decrease based on which is the higher "absolute" min or max value. So if COl 0 shows min = -25 and max = 14 , then I predict col 0 will decrease. If Col 2 shows min = -15 and max = 18 then I predict Col 2 will increase. I do this for all columns.

4. Based on this method at least 7 of the columns will be predicted correctly and sometime up to 9 columns. I have also noticed that when there is a wrong prediction it is usually a "decrease" prediction that is wrong, not an "increase" prediction. The attached excel file shows how I have tracked this.

5. I have only done this with a package of max overlap 0, but I suspect that using max overlap 1,2,or 3 packages may also have some useful results, but then again maybe not.

Another bonus of this method is that the min/max range generated from the max overlap 0 package, seems to be the correct ranges for the winning numbers history sums filter for the full package! The range is wide but still reduces the full package considerably. Try this out too.

So I hope this makes sense, if not ask questions and I'll try and explain. I hope someone can help improve this method also.

Good luck folks
[/quote]


Interesting!! I will definitely give it a try on a 5/39 game and let you know how it went! Good work!

[Clem9403]

[quote=Clem9403:1180217967]
[quote=sliver]
Hello all EL users,

I've spend ALOT of time trying to accurately predict which columns in winning numbers history will increase or decrease. From experience it's probably too much to hope to predict all 11 correct, so I've focused on developing a strategy that will predict as many as possible of the 11 correct.

I find that the simulator is not very relibale on this, and sometimes has up to 6 of the columns wrong. I have also tested the value predictor and have not been able to get any consistent results in predicting the increase or decrease of columns

Today I think I may have stumbled on a fairly good method, but it still needs to be tested and for now I can only say it will predict at the very least 7 of 11 columns correctly, and sometimes up to 9 of 11 columns correctly. In my lottery if I filter the full set of tickets for incr/decr and setting "7 to 11 columns must pass", I am able to reduce tickets by almost half so for me this is a good start.

I have to give credit to PadawanLotto, as this method is based on some strategy he posted earlier which I worked on and now, well it seems to have good results. So in summary the strategy is this -

NOTE: I'm working on the UK Daily Play 7/27 , so you will have to adapt this to other lotteries

1. I create a package of random numbers "max overlap 0" from the package generator. Each time I run this only 3 tickets are created. I run this several times and keep appending tickets until I got a package of about 24 tickets. I then run statistics on this package of 24 to ensure that the lottery numbers are somewhat evenly spread in occurence.If they are not, you may have to delete some tickets and generate new ones to achieve this. There's probably an easier way to do this, but this is the only way I know

2. With the newly created max overlap 0 package, I go to win history sums filter and run "lookup min/max ranges".

3. For each column I predict if it will increase or decrease based on which is the higher "absolute" min or max value. So if COl 0 shows min = -25 and max = 14 , then I predict col 0 will decrease. If Col 2 shows min = -15 and max = 18 then I predict Col 2 will increase. I do this for all columns.

4. Based on this method at least 7 of the columns will be predicted correctly and sometime up to 9 columns. I have also noticed that when there is a wrong prediction it is usually a "decrease" prediction that is wrong, not an "increase" prediction. The attached excel file shows how I have tracked this.

5. I have only done this with a package of max overlap 0, but I suspect that using max overlap 1,2,or 3 packages may also have some useful results, but then again maybe not.

Another bonus of this method is that the min/max range generated from the max overlap 0 package, seems to be the correct ranges for the winning numbers history sums filter for the full package! The range is wide but still reduces the full package considerably. Try this out too.

So I hope this makes sense, if not ask questions and I'll try and explain. I hope someone can help improve this method also.

Good luck folks
[/quote]


Interesting!! I will definitely give it a try on a 5/39 game and let you know how it went! Good work!
[/quote]

Hey Sliver. I've been working with your strategy for the past couple of days. Although I don't get the same results as you as far as using the higher absolute value, I do see about the same results as you doing the following:

1) As you did, I generated tickets according to step one above, Only I did't generate 24 tickets. I only generate one set (which by the way, is 7 tickets for a 5/39 game).

2) I did as you did in step 2; ran my lookup min/max ranges for those 7 tickets.

3) What I did in this step was, I added up the two sums, the min and the max and I devided by 2. I used that sum as my gauge. If it was larger than the current sum, I increased; if it was smaller, I decreased. When I did it this way, I got similar results as you. At least 7 of 11 cols passed and sometimes 9 of 11 passed. I've only back tested this about 4 draws back, but it seems promising.

And as you stated as a 'bonus', I too used the the min/max ranges generated from the max overlap package on a full package and it seemed correct. Wide, yes, but it was still correct. I will back test some more tomorrow when I get off work. I also have some ideas that may make this a better strategy. I just hope they work out!

[Joe]

Me too worked out his idea on predicting,. Some what its good, Nice try,
Let me see your ideas clem, Hope both your ideas will be good on predicting clues on Columns. Try your best.

[Clem9403]

[quote=Joe]
Me too worked out his idea on predicting,. Some what its good, Nice try,
Let me see your ideas clem, Hope both your ideas will be good on predicting clues on Columns. Try your best.
[/quote]

I will do my best!! LOL

[sliver]

Thanks for testing this out, I think it is possible to make this even better. I'll let you know if I manage to take this further, and looking forward to your findings on this