France
Member #92,431
June 7, 2010
43 Posts
Offline
Hi,
I will like to know if somebody is seriously studying or using artificial intelligence for lotto, I know there are some old posts about it and some explanations but has somebody got good results using artificial intelligence software or algorithms?
Is important to me because I am currently working on this kind of software, using backpropagation and it will be very helpful to get some feedback from your experience.
United States
Member #59,352
March 13, 2008
5,626 Posts
Offline
Quote: Originally posted by themagician on Jun 14, 2010
Hi,
I will like to know if somebody is seriously studying or using artificial intelligence for lotto, I know there are some old posts about it and some explanations but has somebody got good results using artificial intelligence software or algorithms?
Is important to me because I am currently working on this kind of software, using backpropagation and it will be very helpful to get some feedback from your experience.
Thank you ;-)
themagician
If you find something then please pass it on. I too have been working with AI and NN but as yet have
nothing to share. Have you ever looked into Brain Maker, if not then google it and read some of the
examples. I was most interested in the example of the binary srting prediction that was above 90%
after being trained. It is a bit high $$$ but I might give it a shot. backpropagation I think would be the
way to go concerning the lottery but I think before any real headway could be made you would need
to use other data instead of the numbers. Think of it this way, the numbers are the final result of the
lottery drawing. Try to reverse engineer the process looking for markers along the way. Much data
would be needed to find high quality markets but I think it might lead to something. I did find a old
program written in basic years ago of a AI or NN that gave you two choices and asked you to press
a key after you had decided. It would then make a prediction as to which choice you had made. Next
you would press the key for correct or incorrect . The more you played the better it got. If I can find
it again I will pass it on. At the time I never had the time to check the code and see how it worked.
I have been 'tinkering' around with converting drawn numbers into binary patterns eg if the number 3 was drawn once over 5 draws the pattern may look like this F,T,F,F,F where T = drawn and F = not drawn. It is possible to find patterns that repeat and some are more frequent than others. But as always the problem knowing when the patttern will occur again.
France
Member #92,431
June 7, 2010
43 Posts
Offline
Ok, I will take a look at Brain Maker, a little bit expensive though..
It is true you cannot input "raw" numbers to the network,and there are some aproaches here, either taking the whole thing as a 1-0 matrix, or to scale all values before input, ah, I keep on testing.. and I do everything on my sparetime, last time I've got 18-20-36.. the comming out numbers beeing 19-21-36.., and it is good, getting closer, I want to get some more
United States
Member #59,352
March 13, 2008
5,626 Posts
Offline
Quote: Originally posted by themagician on Jun 23, 2010
Ok, I will take a look at Brain Maker, a little bit expensive though..
It is true you cannot input "raw" numbers to the network,and there are some aproaches here, either taking the whole thing as a 1-0 matrix, or to scale all values before input, ah, I keep on testing.. and I do everything on my sparetime, last time I've got 18-20-36.. the comming out numbers beeing 19-21-36.., and it is good, getting closer, I want to get some more
TM
Here is what I done. I made 10 groups of 4 numbers for my 5-39. Group 10 only had 3 numbers
because it is a 5 of 39. Next each time any of the numbers from a group hit it would count as a
"1" if the group missed it would be a "0". each draw was converted into a binary string that looked
like this.
1 0 0 1 0 0 1 1 0 1 = 03-15-26-31-39
1 1 0 1 0 0 1 0 0 0
0 0 1 1 0 0 0 1 0 1
1 0 0 1 0 0 0 0 1 1
0 0 1 0 1 1 0 1 1 0
I done this for all draws in my database. Next I read the data from top to bottom placing
each columb in a string. I now had 10 strings of 0's and 1's the lenth of my drawfile. I then
searched each of the 10 long strings for patterns. Many repeating patterns can be found
I would like to run these strings using Brain maker. Lets say that it said that the next bit
in the first columb should be a "1" then I would know that the first number, if correct would
be one of "01-02-03-04" some draws would have more then 1 number that hit within a group
so to make this more accurate one could shuffel the numbers in the groups and then rerun.
I done much with this but only got about 60 to 70 percent hit rate for a couple of strings and
much lower for the rest. I used a program that I wrote to do this and did not use brainmaker.
BrainMaker web page shows a sample for predicting the next digit of a random generated string
with a very high hit rate. You could use 2 numbers in each group and have a 20 bit long string.
France
Member #92,431
June 7, 2010
43 Posts
Offline
Quote: Originally posted by RL-RANDOMLOGIC on Jun 24, 2010
TM
Here is what I done. I made 10 groups of 4 numbers for my 5-39. Group 10 only had 3 numbers
because it is a 5 of 39. Next each time any of the numbers from a group hit it would count as a
"1" if the group missed it would be a "0". each draw was converted into a binary string that looked
like this.
1 0 0 1 0 0 1 1 0 1 = 03-15-26-31-39
1 1 0 1 0 0 1 0 0 0
0 0 1 1 0 0 0 1 0 1
1 0 0 1 0 0 0 0 1 1
0 0 1 0 1 1 0 1 1 0
I done this for all draws in my database. Next I read the data from top to bottom placing
each columb in a string. I now had 10 strings of 0's and 1's the lenth of my drawfile. I then
searched each of the 10 long strings for patterns. Many repeating patterns can be found
I would like to run these strings using Brain maker. Lets say that it said that the next bit
in the first columb should be a "1" then I would know that the first number, if correct would
be one of "01-02-03-04" some draws would have more then 1 number that hit within a group
so to make this more accurate one could shuffel the numbers in the groups and then rerun.
I done much with this but only got about 60 to 70 percent hit rate for a couple of strings and
much lower for the rest. I used a program that I wrote to do this and did not use brainmaker.
BrainMaker web page shows a sample for predicting the next digit of a random generated string
with a very high hit rate. You could use 2 numbers in each group and have a 20 bit long string.
or 3 numbers for a 13 bit string.
RL
If I understand well you test if a column formed by a group of four digits is true or false for the next drawing. If your second group is true, then one or more of these numbers 5,6,7,8 should be comming...mmm..
One of the approaches I had was taking the entire game as a matrix, or a string, in your game it will be a 1 to 39, lets's say then that number 3 has come out then the third position of my array will be 1 "true", and 0 "false" for all the other numbers that didn't came out. Well I've tested a lot with this approach and yet no good conclusion..
United States
Member #59,352
March 13, 2008
5,626 Posts
Offline
Quote: Originally posted by themagician on Jun 25, 2010
If I understand well you test if a column formed by a group of four digits is true or false for the next drawing. If your second group is true, then one or more of these numbers 5,6,7,8 should be comming...mmm..
One of the approaches I had was taking the entire game as a matrix, or a string, in your game it will be a 1 to 39, lets's say then that number 3 has come out then the third position of my array will be 1 "true", and 0 "false" for all the other numbers that didn't came out. Well I've tested a lot with this approach and yet no good conclusion..
TM
Thats the idea. Hope you can make more of it than I did
NASHVILLE, TENN United States
Member #33,371
February 20, 2006
1,044 Posts
Offline
1 0 0 1 0 0 1 1 0 1 = 589
1 1 0 1 0 0 1 0 0 0 = 840
0 0 1 1 0 0 0 1 0 1 = 197
1 0 0 1 0 0 0 0 1 1 = 579
0 0 1 0 1 1 0 1 1 0 = 182
RL
Another method you might try is to see all those zero's and one's as a binary number. I took the liberty of calculating the decimal equivalent of each row of binary numbers.
You could also track the columns as binary number.
1 0 0 1 0 0 1 1 0 1 = 589
1 1 0 1 0 0 1 0 0 0 = 840
0 0 1 1 0 0 0 1 0 1 = 197
1 0 0 1 0 0 0 0 1 1 = 579
0 0 1 0 1 1 0 1 1 0 = 182
________________________________
26 8 5 30 1 1 24 21 3 22
Comparing the columns (not rows) to the next drawing might yield some worthy results. Let's look at that first column. Taking only the last 5 results yielded a decimal sum of 26. What was the next drawing? Let us imagine the result was "1". Would a result of 26 always yield future "1"? How about most of the time? Or almost never? 50% of the time? You will have to backtrack to find out.
Why use the last 5 drawing results? Why not the last 16? or 20? Woiuld increasing the database result in more accurate data? You will have to backtrack to find out.
Now the next step forward is nothing but quicksand.
France
Member #92,431
June 7, 2010
43 Posts
Offline
Quote: Originally posted by mikemerkury on Jun 29, 2010
Can you explain how you calculate decimal sum on binary ?
Hi,
I think he is not taken in consideration the lowest value in the colum to calculate the sum, and so 26 is: 1 1 0 1 in binary, I think that that's how he does it.
It is another way to find patterns, you need to test on it, it is only a matter of continue to experiment on it,
NASHVILLE, TENN United States
Member #33,371
February 20, 2006
1,044 Posts
Offline
If you are using Windows, you have a calculator which will convert decimal into binary and vise-versa. Click on "Start", then "All Programs", then "Accessories", and find "Calculator".
On the left hand side you will see "HEX", "DEC", "OCT", and "BIN". Click on "DEC", input a decimal number, then click on "BIN". The transition is made.
To go from BINary to DECimal, click on BIN, input your binary number, then click on DEC. The transition is made.
NASHVILLE, TENN United States
Member #33,371
February 20, 2006
1,044 Posts
Offline
Quote: Originally posted by themagician on Jul 1, 2010
Hi,
I think he is not taken in consideration the lowest value in the colum to calculate the sum, and so 26 is: 1 1 0 1 in binary, I think that that's how he does it.
It is another way to find patterns, you need to test on it, it is only a matter of continue to experiment on it,
;-)
Correct me if I am wrong but 1101 in binary would be 13.
United States
Member #81,841
October 31, 2009
856 Posts
Offline
Neuroxl has an add-on for excel. Careful tho, they are in the old russia and have a lousy distributer last time I looked. Many sites for the home company (and names?) and one I think was In the Maldives.
The site that came up on a G search shows a big sale going on. Keep me informed and PM me if you find anything out about this stuff. It looked good and better than most I found or had, just never took the leap. Make a purchase with a visa gift card to limit any bad biz from the distributer, they have a history.
DD
Nice signature RL and Ditto on randomness. The creation was not made in disorder and chaos. GEN 1:1 created is perfection implied.
France
Member #92,431
June 7, 2010
43 Posts
Offline
I was taking a look at Neurolox, very interesting, it has a predictor add-on, for example to predict the result of a dog race it take one dog past data: dog's weight, the time of the day, and for the past output it takes the dog's timing in past races.
As always for artificial intelligence all is a matter of parameters,
but how could I pick up the parameters for a Pick 5 data past drawings?
Let's take number 15 for example, maybe I could take the total number of times it came out, the number of times it skipped, the position by frequency, but then what could be the output?
Past output data is needed for the training,
And then I need to take the actual output for every number and compare all the numbers to try to find the "winners"
Uh, I really need to think about this one ;)